Lotto Problems ----

[SavinDwa]

<DIV>[NOTE: this is posted here because I can't find anywhere else that makes sense to post this message]</DIV> <DIV> </DIV> <DIV>Hopefully a dev will reply and give us some info.  I also suspect we will get a lot of posts from people who don't understand probabilities .. *sigh*</DIV> <DIV> </DIV> <DIV>Let me start with some questions/statements<BR></DIV> <DIV>1) The vast majority of players are under the opinion that the lotto system will result in each memeber of the group having an EQUAL chance to win EACH piece of loot.  This means we assume that there is a ROLL made for each new piece of loot.  For instance, in a group of 4 players there should be a 25% chance that each member of the group wins a particular piece of loot on EACH new piece of loot.  We all hope/assume that:</DIV> <DIV>a) The level or class of the group members make no difference</DIV> <DIV>b) the order in which people joined the group makes no difference, or the adding and removing of members over time makes no difference</DIV> <DIV>c) The person that created the group (or group leader) has no advantage</DIV> <DIV>d) There is not a roll made at group formation time (for instance, the system does not assign a players a number range such as 65-88 meaning the y win on a roll of 65-8<img src="/smilies/b2eb59423fbf5fa39342041237025880.gif" border="0" alt="SMILEY" /> or if this is done it is recalcuated correctly every time a member leaves or joins the group.  It would also be nice to think that the system at least uses 3 significant digits if it is used.</DIV> <DIV>e) there is no difference between the split of loot based on the loot type or the con of the mob.</DIV> <DIV> </DIV> <DIV>2) We assume the system rolls a new random number on EVERY event that happens and that this happens server side to avoid exploitations.</DIV> <DIV> </DIV> <DIV>Now if all of the above is true then we are seeing a problem.  We frequently see results in an evening that are at least 2 standard deviations outisde the norm.  Normally 1-2 players seem to get a lot of the loot and 1-2 get very little.  I have personally never seen an even distribution.  I don't mean exact, I mean even close.  In a session with 50 drops in a group of 5 players one player always seems to get 15+ and 1 will get under 5.  The odds of this happening as often as it does are astromically low ... so somethign is going on somewhere.  BUT we the players can't help unless we get soem simple answers </DIV> <DIV> </DIV> <DIV>SO!!! Can someone from SOE post somewhere how lotto works...  this is not a information that is going to destroy roleplaying.  It won't ruin the game.. we all think we know how it works anyway... if it doesn't tell us.  If you want intelligent feedback you need to arm your players with the facts.</DIV> <DIV> </DIV> <DIV>Something is very wrong at the moment.  I have one character that almost never wins the lotto.  Its not a problem for me since I play with the same guild mates all the time and we just adjust the loot at the end of the night.  But it is strange.  My one character has come out with the least lotto wins [or equal worse] at least 50 times in a row... this is beyond bad luck ... this is statistically impossible or very close.</DIV> <DIV> </DIV> <DIV>We need answers .. and I think we can handle them LOL</DIV> <DIV> </DIV> <DIV>END OF RANT!!!</DIV> <DIV> </DIV> <DIV>BTW there are simialr problems with all random number driven events, such as trade skill events.  But lets stick to one example for now.</DIV> <P>Message Edited by SavinDwarf on <SPAN class=date_text>05-10-2005</SPAN> <SPAN class=time_text>09:18 AM</SPAN></P><p>Message Edited by SavinDwarf on <span class=date_text>05-10-2005</span> <span class=time_text>09:19 AM</span>

[jwdanie]

<P>This has been hashed and rehashed ad nauseum here.  You have to count 'orc meat' and other body loot items in the equation.  Is 'orc meat' the equivalent of Ball of Fire: Master I?  As far as players are concerned, no.  As far as computer code is concerned, yes.  If, in fact, you win 5 pieces of 'orc meat' and another person in the group only wins Ball of Fire: Master I you got 'luckier' and won a greater share of the loot.  Also, if you keep the same group together for an infinitely long adventurer over where you find an infinite number of loot drops, the code will be completely random.  Six people together for 2 hours is not a large enough sample size to accurately determine what is going on.  The actual code to generate a random number is simple enough that most calculators can be programmed to do it, therefore the random number generator is likely not broken.  What it is is psychological.  The system cannot be fair because some people get the two best loot drops on a quest.  In this case, quantity counts more than quality.</P> <P>What they could do would be to assign an additional factor that would take the quality of drops into account and adjust the chance of people getting loot to compensate, but then this is no longer a random system.</P>

[Naggyba]

Yea, it has been discussed 1000s of times. It was in beta also.It boils down to them knowing there is a problem with the lotto system. They know certain people always tend to win and soem people never win, until you change the group around. Then other people will tend to win all the time and others loose all the time.They know it's hosed but, I don't think they either care or know how to fix it.

[SavinDwa]

<P>jwdaniels,</P> <P>I disagree entirely.  As I stated in my original post I have grouped with the same group of people [sometime we get an extra) over 50 times.  We tend to adventure for about 3 hours.  We tend to get about 30-70 loot drops dpenending upon where we are adventuring.  This INCLUDES orc meet and all the silly drops.  My character in that group, which is a Ranger, has NEVER ended up the night with anything but the worst or equal worse loot count.  We adjust it all at the end of the night anyway so I don't care, but its a problem.  In total we are talking about 2000+ loot drops.  Our group varies from 5-6 players most of the time, we may have had one night with 4 players?  Even if we assume that there were 6 of us I should have gotten about 300 loot drops.  I have had more like 130-150.</P> <P>Now I leave it up to other posters to calulate the odds of a player who should be winning 1 in 6 to only get 150 out of 2000  ... but it could be pushing 3 standard deviations from the norm.  Even more important is that I have never in any of the 50 sessions had anything other than the lowest or equal lowest percentage of the loot.  I have another character that tends to always get more than their fair share.</P> <P> </P> <P>Your staement about random number generators is exactly the type of urban legend problems we have.  random number genreators are extremely hard to build... EXTREMELY.  The problem is that there is almost no known proven random source and hence all information dervived from such a source may in the end begin to show a pattern.  In addition, random number generation is compute heavy and there is a temptation to try and reduce the compute load on the servers -- net result is that many systems do not generate random numbers for every event.  If you are an expert of random numbers then I apologize, but if you are then I'm interested what you think they are using for a seed.</P> <P> </P> <P>Please don't tell me it evens out ..... or if you are a mathematicians then show me the error and that the results are actually within one standard deviations.  The sample set is certainly large enough.   I am willing to put money on the fact that the system does not work the way we think it does.  in a 6 player group each player does not have a 1 in 6 chance of winning a loot otem on each and every loot drop.  </P> <p>Message Edited by SavinDwarf on <span class=date_text>05-10-2005</span> <span class=time_text>09:47 AM</span>

[SavinDwa]

<P>Naggybait.</P> <P>Its not just the lotto, it seems to happen almost naywhere a random number is needed.  But in the case of the lotto system did they ever explain how it works?  If we knew the formaul we might be able to spot the pattern. </P>

[Uggli]

<div></div>Have noticed this my self also.  I lead my groups 90% of the time, and I rarely win the lotto'd loot.   But when a guildie leads and I'm in slot 3 or higher I end up winning much more.   The other night I got about 60% of the drops in a 5 man group.  That's 1 person getting 3 times the amount of lotto wins than he should be if it worked correctly.   Their random system is broken, and baddly. Edit: To add that they may use some kind of a number table each time you zone.  I know in EQ1 they used something like this.  I could actually adjust my wizards crit rate with a few observations and then zoning if I was on a "bad" table.   Zone back in and I would crit at the proper %.   Could go 1/2 of a LDoN adv and never of gotten a crit with maxed AA's.  Zone out at the 1/2 way point and come back in and I was up to my 20% crit rate as it should of been.    I don't see them actually doing a true random calculation for each thing that would need it.   The amount of so called "random factors" in a game this size would kill a server.  You have a random calc for each melee swing, nuke cast, hit taken, loot rolled, and so on.  It's a lot of crap, then times that by 1500 active people per server.  They had to of found a way to make it easier and less resource intense for their systems, and some kind of pre-generated table seems to make the most sense. <div></div><p>Message Edited by Uggliey on <span class=date_text>05-10-2005</span> <span class=time_text>12:53 PM</span>

[ugl]

<P>A dev responded in this thread that if its been /bugged thats all you can do about it.   It is the last anyone has heard about it however.   But yes, read this thread and you will see that you should not be using the lotto system, however most dont know the problem, like some posters in this thread.   I think its a "sweep it under the rug" tactic.</P> <P> </P> <P><A href="http://eqiiforums.station.sony.com/eq2/board/message?board.id=tech_support&message.id=41833" target=_blank>http://eqiiforums.station.sony.com/eq2/board/message?board.id=tech_support&message.id=41833</A></P> <P><SPAN class=time_text></SPAN> </P><p>Message Edited by uglak on <span class=date_text>05-10-2005</span> <span class=time_text>10:07 AM</span>

[SavinDwa]

<P>uglak,</P> <P>that might explain it.  Lets say they are just using the numbers 1 to 5 in a group then who ever gets 1 and 5 is going to get about 50% of the loot of the pther three.  This is close to what I think we have been seeing.  Now the question is how do you get your position assigned?  I think I know  .... the first and second players are always the top and bottom in the group .. even on a leader change.  I need to loon at this some more .. it still doesn't explain why I'm always getting the bottom share or equal bottom share.  Maybe each character has an internal character number and its sorts them high to low and mine is always high/low?</P> <P> </P> <P> </P>

[IsleWit]

Thanks for the link. I wish the devs would take the time to fix this. I usually start groups and now i know why i havent been getting my fair share of loot. Its quite frustrating seeing lotto after lotto go by and you winning a fraction of what others win. Thats neither right or fair. Please fix the issue SoE. <div></div>

[SavinDwa]

<P>I just tested this random number problem.  I have not done it extensively yet, but have they fixed the random issue now?</P> <P> </P> <P>My results from 171 rolls of /ran 1 3 were:</P> <P>1: 54</P> <P>2: 40</P> <P>3: 77</P> <P>Now having said that .. its a pretty weird distribution to say the least.</P> <P> </P> <P>Now here is an example of where SOE COULD JUST TELL US THE CODE!!!  it is not a destruction of roleplayign to tell us the code they use for /ran 1 3</P> <P>I was expecting something like </P> <P>1: 43</P> <P>2: 85</P> <P>3: 43</P> <P>its not far from that expect the high number os the 3 not the 2??</P>

[SavinDwa]

<P>Just did another test and got expected results:</P> <P>/ran 1 4</P> <P>1:  70</P> <P>2:  74</P> <P>3:  73</P> <P>4:  71</P> <P> </P> <P>i think they have fixed this problem.</P> <P>But the lotto problem is still there...  WE NEED A DEV's input here PLEASE!!!</P> <P> </P>

[Uggli]

It working some times, and not others goes back to my "tables" idea that's generated each time you zone.  Some tables are fair, others are clearly way way bad. <div></div>

[jwdanie]

Until you can run the simulation somewhere in the realm of 1 million times your data collection is irrelevant.  Unfortunately, without the code it is tough to do it 1 million times.

[Tatali]

<div></div><span><blockquote><hr>SavinDwarf wrote:<p>I just tested this random number problem.  I have not done it extensively yet, but have they fixed the random issue now?</p> <p>My results from 171 rolls of /ran 1 3 were:</p> <p>1: 54</p> <p>2: 40</p> <p>3: 77</p> <p>Now having said that .. its a pretty weird distribution to say the least.</p> <p>Now here is an example of where SOE COULD JUST TELL US THE CODE!!!  it is not a destruction of roleplayign to tell us the code they use for /ran 1 3</p> <p>I was expecting something like </p> <p>1: 43</p> <p>2: 85</p> <p>3: 43</p> <p>its not far from that expect the high number os the 3 not the 2??</p> <hr></blockquote> The /random bug was fixed awhile back after several people took the time to prove it was broken, despite dev's claiming "hey, its random! How can there be a problem? People always think random isn't fair." Well, after several megs of log files, hundreds of thousands of /random rolls logged, the devs swallowed their pride and ninja patched the /random function to fix the bug. I could probably find the date it was fixed if I wanted to bother. Now given that they <i>DID</i> have a bug in /random and fixed it, why is it so hard to believe they had a similar logic error in their lotto code? Odds are the same person wrote the various RNG functions and if they used incorrect function calls for one its very likely they did so becuase they didn't understand how it worked and used the wrong function calls in all their work. *edit: Okay, I bothered to look it up. They fixed the lotto bug in the 2/1/05 patch, was just a normal nightly downtime and they made no mention, on public boards, about the fix. Look in <a href="http://eqiiforums.station.sony.com/eq2/board/message?board.id=combat&message.id=5406&view=by_date_ascending&page=1" target=_blank>this thread</a>, towards the last two pages.</span><div></div><p>Message Edited by Tataline on <span class=date_text>05-10-2005</span> <span class=time_text>06:51 PM</span>

[BlackHa]

<P>Just another reason why people are clamoring for solo content. Another reason why people are asking for better loot drops from solo mobs. Another reason why people are so defiant of a new combat sytem designed to stop small groups from taking epics or solos from taking on heroics. It's all about the loot. </P> <P>When the common perception is that the distribution system is flawed people will work around it. One solution is to solo or small group. So these people were obliged with a heavy dose of solo/small group content. I personally believe that a simple random number generator was influential in many sweeping changes within the game. Don't get me wrong...I like the changes and I like to solo. This whole lotto thing makes grouping even more uninspiring to me.</P> <P> </P>

[Mercuratrayin]

Ive noticed this -random- number problem also oocurs in particularly long drop quests like Legend and Lore updates. Where you will kill 300 mobs and have nothing and then get 2 or 3 of the needed updates in one drop. Lotto is indeed broken and needs to be adressed, particularly before Exchange opens, or there will those who exploit this to their advantage. <div></div>

[Naggyba]

The problem is, the system doesn't keep track of who won last and bell curves tend to take effect. If you drop the person that's winning the most and loosing the most and then reinvite them in a different order, the person winning a lot changes. Our groups have experimented with this and it works everytime.This is why I wish they would simply remove all the crap loot (meats, etc.) and replace it with coins and split it evenly among the group. Then just place that meat on a vendor for purchase.Chest drops, well, they simply should be done in order, like we did in EQ1 when we had no random loot system. Just take turns getting the drop.

[IsleWit]

<span><blockquote><hr>Naggybait wrote:The problem is, the system doesn't keep track of who won last and bell curves tend to take effect. If you drop the person that's winning the most and loosing the most and then reinvite them in a different order, the person winning a lot changes. Our groups have experimented with this and it works everytime.<b> This is why I wish they would simply remove all the crap loot (meats, etc.) and replace it with coins and split it evenly among the group. Then just place that meat on a vendor for purchase.</b>Chest drops, well, they simply should be done in order, like we did in EQ1 when we had no random loot system. Just take turns getting the drop.<hr></blockquote> This post deserves its own thread. Ive bolded the part that should be implemented ASAP.</span><div></div>

[SavinDwa]

<P>jwdaniels,</P> <P> </P> <P>Why do you think </P> <P><BR> </P> <BLOCKQUOTE> <HR> Until you can run the simulation somewhere in the realm of 1 million times your data collection is irrelevant.  Unfortunately, without the code it is tough to do it 1 million times. <BR> <HR> </BLOCKQUOTE> <P>If I were to do the test a 1,000 times I would see the trend with a very high level of certainty.  If I did it 10,000 times I would be using mass overkill.  To do it a million times is just  .... silly.  </P> <P>Unless you know something about the random number environment and EQ2 that precludes standard random numbering sampling?  1000 tests should give me a result that is accurate within 2 stadard deviations with an exceedingly high level of certainty.   </P> <P>In the real world if I randomly select 1000 people and ask them whether they prefer blue or red I will get a result that is within 2 standard deviations of the correct answers with a level of certainty that is high enough to allow a billion dollar corp to make a huge investments in product development and marketing.  AND in that case the fear is that my sample was not random or my question had built in bias.  Here I have no built in bias (unless they were really smoking when they wrote the code) and the only way I could have a problem with the sample being random is if the code was designed to have the random numbers vary based on the player casting the function, or the zone I was in, or the time of day etc.  Assuming none of those are there ... 1,000 will give a very accurate result.  Heck ... 200 would give a pretty good indication.</P> <P>Perhaps you could go to this link and explain what portion of the text that you disagree with and hence conclude that at least 1 million rolls would be required here to obtain a view of the pattern with a high level of confidence:</P> <P><A href="http://www.cofc.edu/~milesd/sampling.htm" target=_blank>http://www.cofc.edu/~milesd/sampling.htm</A><BR> </P><p>Message Edited by SavinDwarf on <span class=date_text>05-11-2005</span> <span class=time_text>04:03 AM</span>

[Subtlekni]

Intersting. I ran two small trials before I saw the post about them fixing random. Setting: Inn room. No zoning, or logging between tests. Method: Log file collection               Using a macro to /rand 1 6               Importing the collected data into excel, and throwing it into a pivot table cause I'm lazy Test 1: 1 - 82 2 - 71 3 -71 4 - 80 5 - 93 6 -68 Total run Test1 465 randoms. Test 2: 1- 101 2- 108 3- 112 4- 114 5- 112 6- 101 Total run Test2 649 randoms. So yeah, /rand 1 6 looks like it currently produces a fairly linear distribution in my two quick tests. FYI these tests were conducted today on a live server.

[SavinDwa]

<P>I went back through some logs I kept of evening sessions (I wish I had kept more *sigh).  It is beginning to look like tw players always get about half the amount of loot as the rest,  Everything else seems to be down to normal distribution curves.  So the problem could be that they have the same problem they used to have in the /ran function.  What I'm having problems with is working out which 2 players are going to get shafted.  I don't really have enough data and I didn't start the logs until combat so I don't really know the order we grouped in.</P> <P>But lets take a stab at the technique they are using.  They probabky assign a number to each player from 1 to the number of players in the group.   But when a players leaves the group they would need to reassign numbers (unless the player was the last number).  So lets assume that the players are reassigned numbers anytime there is a change in group structure (player is added or player leaves).  The might also do this when you change from or to the lotto system.  There is no reason for them to remember the order that players joined in, so lets assume that this is really not a factor.  So what do they use?</P> <P>I bet its one of two things: Alpha based on name or numberic based on some hidden character key that we don't know that exists in the server side database.  Or it could be a list kept in memory for the group. If that were the case a player add would always go on the end and player removal would always just collaspse the list to remove the space the player had in the list.  This would suggest that person that created the group would be one of the people hosed until they left the group.</P> <P>There is another possibility.  They have room for 6 characters in a group.  A new character gets one of the open slots when thye join.  Another list states which is active.  If it works like this then it might explain some of the onconsistency.  But none of this explains why one of my characters is always one of the players that gets hosed on the loot drops.  I have never started the group, I have become leader, I have been added second, third, fourth, fifth and sixth.  My character name is Lydya in this case, so its not an alpha sort since there are players below and above me.  But it could be that I have a low internal number.  This character was created day one by my wife testing the game and I decided to use it since she wasn't playing anymore.  The rest of the members of my group all created new characters or were created much later than Lydya .. so could be be a sort on an internal DB number?</P> <P>Of course all of this assumes the problem is a bad structure to the random code.  It seems to fit the pattern, but it could be another problem entirely.  But after norming the data I can confirm the following:</P> <P>2 player group splits loot pretty evenly (but I only have two logs and the samples are not lrage)</P> <P>3 player group results in one player getting about 50% of the loot -- the other two get 25% each (3 logs but not a large sample)</P> <P>4 player group results in two highs and a low -- but only one sample to look at</P> <P>5 player group seems to have 3 highs and two lows -- but the sample is too small to be sure.</P> <P>I don't have a 6 player log.</P> <P>So I guess I need to start logging.</P> <P>But it would be nice to here from a dev.</P>

[Tatali]

<span><blockquote><hr>SavinDwarf wrote:<p>It is beginning to look like tw players always get about half the amount of loot as the rest,  Everything else seems to be down to normal distribution curves.  So the problem could be that they have the same problem they used to have in the /ran function.  What I'm having problems with is working out which 2 players are going to get shafted.  I don't really have enough data and I didn't start the logs until combat so I don't really know the order we grouped in.</p> <p>But lets take a stab at the technique they are using.  They probabky assign a number to each player from 1 to the number of players in the group.   But when a players leaves the group they would need to reassign numbers (unless the player was the last number).  So lets assume that the players are reassigned numbers anytime there is a change in group structure (player is added or player leaves).  The might also do this when you change from or to the lotto system.  There is no reason for them to remember the order that players joined in, so lets assume that this is really not a factor.  So what do they use?</p> <p>I bet its one of two things: Alpha based on name or numberic based on some hidden character key that we don't know that exists in the server side database.</p><div></div><hr></blockquote>Most likely, the members of the group are stored in an array data type. Each "Slot" for the group is numbered, 0-5, and characters fill those slots as they join the initial group. When someone leaves, their slot is simply left empty to be filled by the next person to join. Due to a logic error in the code, some of those slots have a higher probibility to win. There was another post by someone who spend a few night building groups in a specific order and instructing everyone to lotto on everything. Their findings were that there are indeed "hot seats" in the loot order, and they <i>always</i> fall in the same spot of the initial group construction without exception. If you run a search on the boards you might be able to find it. While you say there isn't a reason to remember the order players join the group, I'd suggest there isn't any reason to shuffle the numbers every time someone else joins. Think of it like this: A restraunt has 6 tables. The first people to enter are given table 0, next group table 1, and so forth until all 6 are full. The people at table 2 are in a hurry to get to the theatre so eat quickly and leave. Does the restraunt manager ask the people at table 3 to move to table 2, table 4 to move to the now vacent 3, and table 5 to move to 4? No, they just clean up table 2 and give it to whoever next comes in. What SOE needs to do is go through all of their "random" code and put it through a test program to make sure it doesn't have the same flaws as the /random prior to Feb... not only lotto, but also their combat code as well. Look at is this way: The biggest problem they are having with their current combat engine is it is possible for a player to reach a level of defence where they become all but invulnerable, right? Now think about that flaw that was in the /random code, where the two far ends of the range where only getting half the probibility to occur. Now with that idea in your mind, go back to combat where a tank is near the very edge of the mob's "to hit" range... maybe the tank is only getting half as often as they should because of a bugged random? First, SOE needs to be willing to accept that there may in fact be a problem with their random code in the game. Unless they do, any other effort is wasted since we can't see the code and other than /random, the other things are rather hard to collect bulk evidance for a normal player. Given that /random <i>was</i> confirmed to be bugged and it was fixed, I'd say that puts reasonable doubt on the rest of the game's RNG code. Second, SOE needs to take one of their coders and ask them to review that code, preferable someone with experience in RNG and who didn't write the original. I would suggest making a small non-eq application and plugging in the existing formulas that prints the results to a text log that can be parsed. Many players have written parsers, so just need to match the text out message to something a parser already understands and that part of your work is done for you. Set this little program to loop about 10,000 times and parse that result, see what the actual loot distribution is for the game for a 6 memeber group. When that comes up bugged, fix the problem and run again. When that second test passes becuase the minor flaw was fixed, run it though 5, 4, 3, and 2 member groups just to make sure those are working too. Once the bug is swashed for one, it probably will run properly for all. Third, check anything else that hits the RNGs... quest updates (L&L anyone?), harvesting, combat, loot tables, skill gains, spawn rates and tables, weather, ... all of it. Some of that might be pinging the same RNG, but it would be good to check it anyway. Fourth, post a small patch note that "Lotto code was optimized to run more smoothly in game" or something that sounds like "well, there wasn't a problem we just made it better" to most people, but to others they know that means you fixed the careless bugs that made the probibilities uneven.</span><div></div>

[SavinDwa]

<P>Tataline,</P> <P> </P> <P>I agree with almost everything you have said.  The one area where I suspect you are wrong is with your analogy to a restaurant.  Code is a lot weirder than that, it unfortunately tends to evolve rather than get designed and it gets optimized for performance in cases like a game engine where poor performance converts to a higher cost per user from a server perspective.</P> <P>Here is my guess.  Each player has some information that is kept in memory on the server about that player.  Some of this information may be written to disk when a significant event takes place or based on time elapsed.  There is also information about a group.  I strongly suspect that the group information only has pointers to specific information about players and that the player specific information is stored with the players other information in memory.  We can also pretty much conclude based on game behavior that all of this information is stored on a server/zone basis.  When you leave a zone it creates new information about the player which is why zoning can fix a lot of problems that occur.  Since the group information exists on all players in all zones my real guess is that there is no group information as a seperate entry, each player has on their player record a copy of the group info and any changes is echoed to the record of every player in the group so that their group info is also updated.</P> <P>OK.. I don't want to beat this to death.  But for this system to work and be effecient in a lotto situation each player needs to be assigned a number from 1 to the number of players in the group.  The little lotto code "rolls" to get a number probably using the same code that /ran x y uses.  The overhead to deal with a skip in the sequence would have to happen on every roll.. much easier to eliminate the skip[ when the event occurs.  So what happens when:</P> <P>1) A player is added to a group?  That player is assigned the a number X where X is equal to the number of players in the group.  this would seem logical, but i think it doesn't work this way.  I will explain why below.</P> <P>2) A player leaves the group.  If it s the player assigned the number X where the number on the group was X then do nothing except reduce the  number of players in the group.  If the player has a number X that is less than the total number of players in the group then reduce the number assigned to all players with a number higher than X by 1.</P> <P>Ok, this works,  But its not the way a programmer looks at stuff.  Its lots of code with if then elses.  Here is the cleaner way to handle this.</P> <P>We add a new feature called Lotto.  The program thinks about this and say "ah ha!!" this is easy as long as I assign a player a number from 1 to X and then I can use the random code we already have.  If this was a subroutine great .. if not then cut and paste is great.  the code is so simple that the overhead of a subroutine is stupid .. so it gets cut and paste.  So the program has the DB guy add a new field to a player record called "player lotto number".  The programmer now knows wher to store the number in memory.  Now the programmer says "OK.. what do I do if they turn lotto on?"  I look at the players in the group and assign each of them a number from 1 to X where X is the number of players in the group.  How does the player handle a player adding or leaving the group?  Easy .. tell the group management programmer than on any change in group formation call the lotto setup program again.  No new code, no special case, works everytime.   But, if I'm right, how does it assign the numbers?  We just don't know... it depends how they store the siz player Ids in memory.  Forgetting the lotto it would be fine to have spot 3 blank.  So a if player 3 leaves, then slot 3 is blank.  If a new player arrives .. it might take the empty slot 3? It might take the empty slot 6? who knows.</P> <P>Anyway....   without a Developer its all guess work unless we reverse engineer it.  And reverse engineering a broken lotto system is not a great way to spend our time LOL </P>

[Kroder]

<DIV>I have noticed a trend, where those who have the greatest exchange rate of damage, seem to have better chances of winning a lotto.</DIV> <DIV> </DIV> <DIV>Here is my theory:</DIV> <DIV> </DIV> <DIV>Players get a higher percentage chance of winning any given lotto, based on most recent fight's activity.</DIV> <DIV>If Player A was a DPS, and did 400 pts of damage to the critter...</DIV> <DIV>and player B was a healer, and healed 300 pts of damage taken by the tank, and 100 taken by the DPS when they caught aggro...</DIV> <DIV>and player C was the Tank, who took 300 points of damage, but dished out 125 points to the creature as well....</DIV> <DIV> </DIV> <DIV>Then A = 400 (damage done to benefit the group)</DIV> <DIV>         B = 400 (300 + 100 healed, for groups benefit)</DIV> <DIV>         C = 425 (Damage taken on behalf of group, + damage done to benefit the group)</DIV> <DIV> </DIV> <DIV>Then player C has a slightly better chance of winning the next lotto they enter into.</DIV> <DIV> </DIV> <DIV>We have seen many parser programs.  These are not complicated, and take up few resources on a system.  I am considering the possibility that there is one built into the game, and it modifies the lotto system based on percieved levels of participation.</DIV> <DIV> </DIV> <DIV>(Buffs will have a presence here too, depending on their effect, and whether they changed the ability to take, or deliver, damage to a creature)</DIV> <DIV>The above example is hardly typical.  Often, one party member has a larger role than his peers, whether it is because they are the only DPS in a group of tanks and priests, or they are the only healer in a group of scouts, etc.</DIV> <DIV>Whoever is having the single greatest impact on a given fight, if I am guessing this system right, will end up winning more often than the others in their group.</DIV> <DIV> </DIV> <DIV>Any ideas?</DIV> <DIV> </DIV> <DIV> </DIV>

[VonStein]

<BR> <BLOCKQUOTE> <HR> Kroder wrote:<BR> <DIV>I have noticed a trend, where those who have the greatest exchange rate of damage, seem to have better chances of winning a lotto.</DIV> <DIV> </DIV> <DIV>Here is my theory:</DIV> <DIV> </DIV> <DIV>Players get a higher percentage chance of winning any given lotto, based on most recent fight's activity.</DIV> <DIV>If Player A was a DPS, and did 400 pts of damage to the critter...</DIV> <DIV>and player B was a healer, and healed 300 pts of damage taken by the tank, and 100 taken by the DPS when they caught aggro...</DIV> <DIV>and player C was the Tank, who took 300 points of damage, but dished out 125 points to the creature as well....</DIV> <DIV> </DIV> <DIV>Then A = 400 (damage done to benefit the group)</DIV> <DIV>         B = 400 (300 + 100 healed, for groups benefit)</DIV> <DIV>         C = 425 (Damage taken on behalf of group, + damage done to benefit the group)</DIV> <DIV> </DIV> <DIV>Then player C has a slightly better chance of winning the next lotto they enter into.</DIV> <DIV> </DIV> <DIV>We have seen many parser programs.  These are not complicated, and take up few resources on a system.  I am considering the possibility that there is one built into the game, and it modifies the lotto system based on percieved levels of participation.</DIV> <DIV> </DIV> <DIV>(Buffs will have a presence here too, depending on their effect, and whether they changed the ability to take, or deliver, damage to a creature)</DIV> <DIV>The above example is hardly typical.  Often, one party member has a larger role than his peers, whether it is because they are the only DPS in a group of tanks and priests, or they are the only healer in a group of scouts, etc.</DIV> <DIV>Whoever is having the single greatest impact on a given fight, if I am guessing this system right, will end up winning more often than the others in their group.</DIV> <DIV> </DIV> <DIV>Any ideas?</DIV> <DIV> </DIV> <DIV> </DIV><BR> <HR> </BLOCKQUOTE><BR> <DIV>Sorry, doesn't work because I've had my Inquisitor, who was doing NO damage (just healing) get the hot seat for lotto ( win a lot )... and also had him on the loosing end, with almost the same type of chars in the group....</DIV>

[Kroder]

<P>Read more carefully.</P> <P> </P> <P>Healing is a type of damage exchange, it erases it.</P> <P> </P> <P>My example clearly stated the healer would be greatly credited for his efforts.  It may not be accurate, but your reason is not the one that would explain why.</P> <P> </P>

[jwdanie]

<P>SavinDwarf,</P> <P>The reason why you would need to do it 1,000,000 times is simple... because the more samples you have the better your sample size.  Sure, you can get accuracy out of 1,000 samples to within 2 standard deviations of the norm, but if I'm the one who might be getting 2 standard deviations below the norm and not the one that might be getting 2 standard deviations above the norm I am not satisfied with that level of accuracy.  Also, in a group of six, if you have 4 rare loots, the same individual is going to win them 1 time in every 216 (doesn't matter who wins the first one, than for that same individual to win the next three is (1/6)*(1/6)*(1/6)=(1/216).  </P> <P>Also, I don't know if you have ever taken a statistics class, but one of the first things that they teach you about probability is that the only way to run a test without any chance of error is to have it becomplletely duplicatable in every way and run it an infinite number of times.  Clearly this is impossible, but 1 million is much closer to infinity than 1 thousand thus you will get a much more accurate depiction of what is really going on.</P> <P>Finally, it isn't that hard to run this simulation 1 million times.  It is simple computer code, and I am sure that it is run well over 1 million times per day already given the population of people that plays this game and the duration of time that they spend playing.  So many people have complained about this issue and it has been responded to by the devs time and time again.  They have the capability to test the code and find out whether or not it is working, and you do not.  They have tested it and said it works, and anyone who complains about it beyond a certain point is starting to really seem like they are whining.</P>

[SavinDwa]

<P>jwdaniels,</P> <P> </P> <P>I really don't know where to start in answering your statement.  It has probably got a high probability to become a waste of ours and everyone else's time.  But let me answer some of your questions:</P> <P><BR> </P> <BLOCKQUOTE> <HR> <P>Also, I don't know if you have ever taken a statistics class, but one of the first things that they teach you about probability is that the only way to run a test without any chance of error is to have it becomplletely duplicatable in every way and run it an infinite number of times.  Clearly this is impossible, but 1 million is much closer to infinity than 1 thousand thus you will get a much more accurate depiction of what is really going on.</P> <HR> </BLOCKQUOTE> <P>hmm..</P> <P>I suspect I must have taken a class or two.  I have a BSc in Pure Mathematics from Churchill Cambridge and graduated in 1974.  I have also had the misfortune to be an Associate Professor Teaching proability Theory....  between the two I think it qualifies me to have a bit of an opinion on the subject, wlthough it in no ways makes me an expert.  But hopefully that answers the first part of your question.  I have to confess to feeling a bit out of date since things must have changed a lot.  I have never had a professor ever teach me that "the only way to run a test without a chance for error......" nor was it ever anything I taugh to any of my students.  Of course I was teaching the theory of probability.  Perhaps you are referring to more of a Business Stats class? But even then .. it seems a strange thing to say?  Using the same principle I could through out DNA as being valid evidence in a  trial since its results are not absolute -- there could be another person with the exact same DNA that is not a twin ... but its beyond the realms of reasonableness.</P> <P>But lets get back to the issue at hand.  If I ran a test on the SOE /ran function for 6 sets of 1,000 and the total result for the 1,000 and the individual results for each of 6 sub sets were all within 1 standard deviation of one another what is the probability that I have recieved a result that is actually not within two standard deviations of what the model should produce in a run of 1 billion times?</P> <P>Does it make a difference if one of my samples is more than a standard deviation from the total set? how much does that reduce the probability that have a recieved a result within two standard deviations of what would be produced from a run of 1 billion?</P> <P>Also, your faith in SOE, and I use the word faith for a reason, is I feel somewhat misplaced.  They are designing a very complex system, it is by its nature full of issues.  The also suffer from having a user base that is for the most part makes statements as if they were fact based on zero testing, so they have to work out which 1% of the posts are real asoppsed to fabricated or exaggerated [eiether deliberately or innocently).  They are getting terrible feedback.  </P> <P>A case in point is the /ran function was clealry not working and once SOE finally listened to the right people they spotted the bug and fixxed it without telling any of us.  </P> <P>So now for the 64 thousand dollar question:</P> <P>Depite the fact that you have a the right to an opinion and the right to free speech do you actually have any empirical data to support your position that the lotto system is working?  If not then at least plase preface your statements with "I really don't knwo if its working or not, but .."  this would help SOe immensely.</P> <P>I went through this on a TS bug.  I tried to report it and everytime I brought up on the boards 100 people would tell me why it wasn't a bug and why they had a bigger problem and to stop whinning.  Eventually I got a messag ewith a question.. followed by a response, followed by a thank you for spotting the problem, followed by the bug being fixed.  </P> <P>AND, finally, can you show me the post where SOE has staed the lotto code is working correctly?  I have never seen it. <BR>  </P>

[Sav]

<span><blockquote><hr>SavinDwarf wrote:<p>jwdaniels,</p> <p>I really don't know where to start in answering your statement.  It has probably got a high probability to become a waste of ours and everyone else's time.  But let me answer some of your questions:</p> <blockquote> <hr> <p>Also, I don't know if you have ever taken a statistics class, but one of the first things that they teach you about probability is that the only way to run a test without any chance of error is to have it becomplletely duplicatable in every way and run it an infinite number of times.  Clearly this is impossible, but 1 million is much closer to infinity than 1 thousand thus you will get a much more accurate depiction of what is really going on.</p> <hr> </blockquote> <p>hmm..</p> <p>I suspect I must have taken a class or two.  I have a BSc in Pure Mathematics from Churchill Cambridge and graduated in 1974.  I have also had the misfortune to be an Associate Professor Teaching proability Theory....  between the two I think it qualifies me to have a bit of an opinion on the subject, wlthough it in no ways makes me an expert.  But hopefully that answers the first part of your question.  I have to confess to feeling a bit out of date since things must have changed a lot.  I have never had a professor ever teach me that "the only way to run a test without a chance for error......" nor was it ever anything I taugh to any of my students.  Of course I was teaching the theory of probability.  Perhaps you are referring to more of a Business Stats class? But even then .. it seems a strange thing to say?  Using the same principle I could through out DNA as being valid evidence in a  trial since its results are not absolute -- there could be another person with the exact same DNA that is not a twin ... but its beyond the realms of reasonableness. </p> <p><font color="#ffff00">Probability theory should include basic statistics.  In basic statistics, you should have taught about various hypothesis testing.  You could have simply made a statement such as: </font></p> <p><font color="#ffff00">H0: When /rand 1 6, each outcome has an equal probability. HA: When /rand 1 6, at least two outcomes have unequal probabilities. </font></p> <p><font color="#ffff00">You could then perform a chi squared test.  In the examples given by jwdaniels, N = 171 or whatever number he used.  The rest you can look up in a basic statistics textbook.  The most common one used in Intro to Statistics is Devore and Peck, <i>Introduction to Probability and Statistics for Engineers and Scientists</i> or Devore and Peck's <i>Introduction to Data Analysis</i>.  </font></p> <p><font color="#ffff00">Take alpha = 0.05 or 0.01 or whatever alpha corresponds to the level of confidence you would like.  Since you have been an Associate Professor (which I find suspect since you only have a BS and Professorships require Ph.D's or at least a Master's degree at a Community College), I'm sure you can run the Chi Squared test yourself.  I can always do it and post the results if you so desire.  I hope I am not sounding too condescending, but I just find your story to be dubious at best. </font></p> <p><font color="#ffff00">While we are throwing around qualifications, I am a graduate student in Physics and graduated from the University of Minnesota with dual BS Physics and Statistics and a minor in Mathematics. You are correct that 1000 is a reasonable sample size.  So is 1 million. </font></p> <p>But lets get back to the issue at hand.  If I ran a test on the SOE /ran function for 6 sets of 1,000 and the total result for the 1,000 and the individual results for each of 6 sub sets were all within 1 standard deviation of one another what is the probability that I have recieved a result that is actually not within two standard deviations of what the model should produce in a run of 1 billion times?</p> <p>Does it make a difference if one of my samples is more than a standard deviation from the total set? how much does that reduce the probability that have a recieved a result within two standard deviations of what would be produced from a run of 1 billion?</p> <p>Also, your faith in SOE, and I use the word faith for a reason, is I feel somewhat misplaced.  They are designing a very complex system, it is by its nature full of issues.  The also suffer from having a user base that is for the most part makes statements as if they were fact based on zero testing, so they have to work out which 1% of the posts are real asoppsed to fabricated or exaggerated [eiether deliberately or innocently).  They are getting terrible feedback.  </p> <p>A case in point is the /ran function was clealry not working and once SOE finally listened to the right people they spotted the bug and fixxed it without telling any of us.  </p> <p>So now for the 64 thousand dollar question:</p> <p>Depite the fact that you have a the right to an opinion and the right to free speech do you actually have any empirical data to support your position that the lotto system is working?  If not then at least plase preface your statements with "I really don't knwo if its working or not, but .."  this would help SOe immensely.</p> <p>I went through this on a TS bug.  I tried to report it and everytime I brought up on the boards 100 people would tell me why it wasn't a bug and why they had a bigger problem and to stop whinning.  Eventually I got a messag ewith a question.. followed by a response, followed by a thank you for spotting the problem, followed by the bug being fixed.  </p> <p>AND, finally, can you show me the post where SOE has staed the lotto code is working correctly?  I have never seen it.   </p> <div></div><hr></blockquote></span><div></div>

[jwdanie]

<P>To address the points that have been made:</P> <P>1: Yes, I am in fact referring to a statistics class for business.</P> <P>2: I do not have any empirical evidence to show that the system is working correctly, because when I am playing the game I am playing the game and not keeping statistical evidence to prove whether or not the lotto system is fair.</P> <P>3: I don't recall saying that SOE has confirmed or denied whether or not the system itself is broken, therefore I cannot quote a post illustrating this.  What I was trying to say was that SOE has access to every result generated by the lotto system, and if they really wanted to test what was going on they easily could.</P> <P>What I can say from my personal playing experience is that sometimes I get a lot of loot, sometimes I don't.  I have an occasional hot streak, and I have an occasional cold streak.  When harvesting, some days I get rares, some days I don't.  What I am trying to say is, from my experience, it seems very random to me.  When I see people complaining about the "broken" lotto system, it's generally because they did not get the loot drop that they wanted.  If you get loot that you don't want, or you don't get loot that you want, either trade it with someone or use the broker system to buy/sell stuff.</P>

[VonStein]

<DIV>I think the problem here is that ( and I have only my experience to back this ) the lotto system is more "streaky" than it should be...</DIV> <DIV> </DIV> <DIV>Yes, it may equal out eventualy if you switch in and out of multiple groups...</DIV> <DIV> </DIV> <DIV>But streaks and inequality seem to be the rule, and not the exception.</DIV>

[Tatali]

<div></div><div></div>It has nothing to do with streaks, its about a flaw that gives unequal probibility to each person. Look at it this way: You and 11 of your friends go to the beach with metal detectors and someone suggests you throw dice for who wins each item found. Each person is assigned a number from 1 to 12 and any time an object is found you roll 2 six sided dice, add the totals, and give the item to the person with that number. Now, who do you want to be? The person with #7 or the person with #1? =p Now, you <i>could</i> roll a 12 sided die to give to folks and <i>THAT</i> would be a fair distribution as everyone has the same chance to win. Yes, streaks would occur which is to be expected, but the distribution over a length of time should even out. In the 2d6 example, the distribtuion would never even out, it would remain a bell curve forever due to unequal chances to win...in fact, person 1 would never win anything as a 1 on 2d6 is impossible. That is roughly what EQ2's lotto code is doing, but their distribution more closely resembles rolling 2d6, divide result by 2, and round up and give to group position that has that result. The only reason I can think of that someone would try a 2d6 method for the lotto code is to make it "more random" by hitting the RNG twice, which is a logic flaw. <div></div><p>Message Edited by Tataline on <span class=date_text>05-12-2005</span> <span class=time_text>04:32 PM</span>

[Sav]

<span><blockquote><hr>Tataline wrote:<div></div><div></div>It has nothing to do with streaks, its about a flaw that gives unequal probibility to each person. Look at it this way: You and 11 of your friends go to the beach with metal detectors and someone suggests you throw dice for who wins each item found. Each person is assigned a number from 1 to 12 and any time an object is found you roll 2 six sided dice, add the totals, and give the item to the person with that number. Now, who do you want to be? The person with #7 or the person with #1? =p Now, you <i>could</i> roll a 12 sided die to give to folks and <i>THAT</i> would be a fair distribution as everyone has the same chance to win. Yes, streaks would occur which is to be expected, but the distribution over a length of time should even out. In the 2d6 example, the distribtuion would never even out, it would remain a bell curve forever due to unequal chances to win...in fact, person 1 would never win anything as a 1 on 2d6 is impossible. That is roughly what EQ2's lotto code is doing, but their distribution more closely resembles rolling 2d6, divide result by 2, and round up and give to group position that has that result. The only reason I can think of that someone would try a 2d6 method for the lotto code is to make it "more random" by hitting the RNG twice, which is a logic flaw. <div></div><p>Message Edited by Tataline on <span class="date_text">05-12-2005</span> <span class="time_text">04:32 PM</span></p><hr></blockquote>What is the empirical basis for this claim?</span><div></div>

[Barneba]

<P>Two phrases popped into my head after reading this thread:</P> <P>1) "They nerfed J-boots!!!"</P> <P>2) "Working as intended"</P> <P> </P> <P>If you never played EQ you might miss the point... So here is a quick refresher:</P> <P>1) After each patch, people would claim that the movement speed on J-boots were nerfed... People were always sure they ran slower after each patch untill the day, that people had been claiming it so much, that it simply became a catch-phrase that everybody started using after each patch message <img src="/smilies/8a80c6485cd926be453217d59a84a888.gif" border="0" alt="SMILEY" /></P> <P>2) In EQ we did have DEVs coming out and telling people that feature x and y was "Working as intended"... All they ever got out of that phrase was more bashing and people started making fun of the phrase (especially the time they claimed it was working as intended and it didnt - that can happend, human error).</P> <P> </P> <P>So why did these 2 phrases pop into my head? They just did! <img src="/smilies/8a80c6485cd926be453217d59a84a888.gif" border="0" alt="SMILEY" /></P>

[Kroder]

<P>Lotto is not fair. Random is not fair.</P> <P> </P> <P>Put in a fair loot system.  Round robin, with exception ability for items by mutual consent.</P> <P> </P> <P>Leave lotto as an option, but give better options.</P> <P> </P>

[Big Da]

<BLOCKQUOTE> <BLOCKQUOTE><SPAN>OMG the lotto code is statistically significant (or was last time I looked at this topic a few months back); if you know about random numbers and the theory behind the concept of random this should not happen. We don’t need somebody to explain the concept of random to us or how perception can make us biased but this is not the case. This is not helpful!!<FONT color=#000000></FONT></SPAN></BLOCKQUOTE></BLOCKQUOTE><p>Message Edited by Big Dave on <span class=date_text>05-13-2005</span> <span class=time_text>02:42 PM</span>

[Sunrayn]

<BR> <BLOCKQUOTE> <HR> Naggybait wrote:<BR><BR><BR>This is why I wish they would simply remove all the crap loot (meats, etc.) and replace it with coins and split it evenly among the group. Then just place that meat on a vendor for purchase.<BR><BR>Chest drops, well, they simply should be done in order, like we did in EQ1 when we had no random loot system. Just take turns getting the drop.<BR> <HR> </BLOCKQUOTE><BR> <DIV>Auto split of coins didnt work right in EQ1 either  <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" /></DIV>

[Kneemin]

<P>I haven't seen any such problem in any of the groups i have been in.</P> <P>either way a lotto is actually a very simple program that can created with a few lines of coding.</P> <P>by using a randomize statement then activating a variable to hold the number that is randomly drawn in the program.</P> <P>the only way the loot is unfairly distributed would be if they did something to tamper with a person's number like mentioned in the above posts.</P> <P> </P> <P>Edit: There would be no such situation with 2d6 or 1d12 and such its just a random drawn number generated by a simple visual basic program.</P><p>Message Edited by EQIIgamer on <span class=date_text>05-13-2005</span> <span class=time_text>07:17 AM</span>

[Subtlekni]

<span><blockquote><hr>EQIIgamer wrote:<div></div> <div></div> <p>Edit: There would be no such situation with 2d6 or 1d12 and such <font color="#ff0000">its just a random drawn number generated by a simple visual basic program.</font></p><p>Message Edited by EQIIgamer on <span class="date_text">05-13-2005</span> <span class="time_text">07:17 AM</span></p><hr></blockquote>Eq servers run on VB?  /smacks head Now I understand! P.S. This post in no waited helped or contributed to the current discussion. Please farrwolf, don't hurt me!</span><div></div>

[TerrorRising]

<DIV> <DIV>Let's get down to business.</DIV> <DIV> </DIV> <DIV>First, this comparison of who is more qualified and has better statistics is nonsense. Leave your qualifications at the door. What is most clear from the exchange is that both of you have gotten away from the real issue here: Is Lotto a truly random (and thus equal) means to award treasure?</DIV> <DIV> </DIV> <DIV>I believe the answer is no, it isn't.</DIV> <DIV> </DIV> <DIV>There is in fact a way to guarantee that you do have a chance at winning, but that chance <EM>is divided equally among those who know how to manipulate lotto</EM>.  Before any of you cry foul, <EM>there is no illegal manipulation to make this happen</EM>. The answer is surprisingly simple, and it is shameful that a system was created that isn't truly random. Come on guys, is it that hard to program an automatic random system?</DIV> <DIV> </DIV> <DIV>==========</DIV> <DIV> </DIV> <DIV>Edit: I've removed my explanation. Honestly, I don't care if you believed me or not, but if you do want a fair reward system, I'd advise going with random, not lotto.</DIV> <DIV> </DIV> <DIV>==========</DIV> <DIV> </DIV> <DIV>With that out of the way, let's go back to the bogus secondary argument that has people comparing their "prowess" with statistics. </DIV> <DIV> </DIV> <DIV> </DIV> <DIV>While one million iterations is absurd, the sentiment is still true. The Law of Averages favors larger numbers. Here is a definition pulled from <A href="http://www.m-w.com" target=_blank>www.m-w.com</A>:</DIV> <DIV> </DIV> <DIV>Law of Averages: the commonsense observation that probability influences everyday life so that over the long term the possible outcomes of a repeated event occur with specific frequencies. </DIV> <DIV> </DIV> <DIV>Which basically means, if you run a test more times, you are more likely to get an "equal" distribution of numbers among the possible outcomes.</DIV> <DIV> </DIV> <DIV>And thank you for the statistics refresher. It was nice to revisit some of the basics.  But please leave out your titles and job descriptions. If you know statistics, show us, don't tell us why you are qualified to show us. </DIV> <DIV> </DIV> <DIV>I don't come here often, so don't expect me to correct a mistake if I made one, and definitely don't expect me to respond to your post. I'd prefer to play the game than to talk about it.</DIV></DIV><p>Message Edited by TerrorRising on <span class=date_text>06-17-2005</span> <span class=time_text>07:32 PM</span>

[jwdanie]

<P>Terror,</P> <P>I am calling your bluff.  If you really knew such a system and were interested in fixing it you would post it here and /bug it to let the devs know about it.  I have heard of several methods that people claim influence the lotto table, including the "designated looter" method of switching group leader, delaying until the last possible moment to roll on an item,  number of group members, order of recruiting, etc.  From what I can tell, they don't work.  I think that most of these theories are a result of streaks that people see happening and assume are the rule.  I also play table top RPGs, and we have a friend who "always rolls either a 20 or a 1" on his 20 sided die.  Does he really?  No, but anytime a big roll comes up it seems that he either succeeds with a 20 or critically fails with a 1.  Still, we know it is all random, but that is not the perception.  People allow there perception to cloud their judgement in all areas, from investing (the market always performs poorly when the AFC wins the SuperBowl) to sports (get it to the man with the hot hand) to the lottery (the number 42 comes up more than any other number).  These examples may sound stupid, but people do practice them in real life.</P> <P>To close, I am tired of arguing the same thing over and over, so this will be my last post on this subject, but random number generation is an extremely easy thing to program.  I am not a computer programmer, but I know how to generate a random number simulator to roll between 1 and 6 (technically between 0 and 5) and it just isn't that hard to do.  If more groups just practiced need before greed and switched items around to the person that could use them, we wouldn't even be having this conversation right now.  </P> <P> </P>

[Ceruline]

<DIV>I spent some Excel time just now figuring this out.  These are the theoretical probabilities of winning AT LEAST x lottos in 1000 tries, assuming a six person group with a 1/6 chance of winning each lotto.</DIV> <DIV> </DIV> <DIV>The calculation is as follows (Please double check my combinatorics, those who know enough of it.  I haven't done this for a while, so I may have messed up, but all of my probabilities for each individual result do add up to 1)</DIV> <DIV>Probability of winning exactly x times out of n = P(win)^x * P(loss)^(n-x) * xCn</DIV> <DIV> </DIV> <DIV>Plain language - The probability of winning the lotto x times out of n is equal to the probability of winning once to the power of the number of wins, times the probability of losing once to the power of the number of losses, times the number of combinations of that number of wins that you can take from the total number of trials.</DIV> <DIV> </DIV> <DIV>To get the at least probabilities, I just added the probabilities of a given result and every result higher than it.</DIV> <DIV>At least 326 wins:</DIV> <DIV> 3.61 * 10^ -35 </DIV> <DIV>At least 300 wins:</DIV> <DIV>1.35 * 10^ -25</DIV> <DIV>At least 250 wins:</DIV> <DIV>1.41 * 10^ -11                </DIV> <DIV>At least 200 wins:</DIV> <DIV>.3188%</DIV> <DIV>At least 190 wins:</DIV> <DIV>2.7%</DIV> <DIV>At least 180 wins:</DIV> <DIV>13.8%</DIV> <DIV>At least 170 wins:</DIV> <DIV>40.15%</DIV> <DIV>At least 160 wins:</DIV> <DIV>72.6%</DIV> <DIV>At least 150 wins:</DIV> <DIV>92.8%</DIV> <DIV>At least 140 wins:</DIV> <DIV>99.06%</DIV> <DIV>At least 130 wins:</DIV> <DIV>99.94%</DIV> <DIV>At least 120 wins:</DIV> <DIV>99.998%</DIV> <DIV>At least 110 wins:</DIV> <DIV>99.99998%</DIV> <DIV> </DIV> <DIV>As you can see, even with only 1000 trials, the probability dropoff from the average (166 roughly) is very steep.  Getting a result that is a significant outlier ought to be VERY rare.  Upshot of this is that 1000 trials ought to be a more than sufficient sample size, provided that the conditions remain consistant throughout.  </DIV> <DIV> </DIV> <DIV>Not making any claims here, since I have no data myself, but this is how it SHOULD work with a perfect RNG.</DIV>

[CheezeBurgerAnimal]

<BLOCKQUOTE> <HR> SavinDwarf wrote: <DIV>[NOTE: this is posted here because I can't find anywhere else that makes sense to post this message]</DIV> <DIV> </DIV> <DIV>Hopefully a dev will reply and give us some info.  I also suspect we will get a lot of posts from people who don't understand probabilities .. *sigh*</DIV> <DIV> </DIV> <DIV>Let me start with some questions/statements </DIV> <DIV><STRONG><FONT color=#ffff00>1) The vast majority of players are under the opinion</FONT></STRONG> that the lotto system will result in each memeber of the group having an EQUAL chance to win EACH piece of loot.  This means we assume that there is a ROLL made for each new piece of loot.  For instance, in a group of 4 players there should be a 25% chance that each member of the group wins a particular piece of loot on EACH new piece of loot.  We all hope/assume that:</DIV> <DIV>a) The level or class of the group members make no difference</DIV> <DIV>b) the order in which people joined the group makes no difference, or the adding and removing of members over time makes no difference</DIV> <DIV>c) The person that created the group (or group leader) has no advantage</DIV> <DIV>d) There is not a roll made at group formation time (for instance, the system does not assign a players a number range such as 65-88 meaning the y win on a roll of 65-8<img src="/smilies/b2eb59423fbf5fa39342041237025880.gif" border="0" alt="SMILEY" /> or if this is done it is recalcuated correctly every time a member leaves or joins the group.  It would also be nice to think that the system at least uses 3 significant digits if it is used.</DIV> <DIV>e) there is no difference between the split of loot based on the loot type or the con of the mob.</DIV> <DIV> </DIV> <DIV>2) We assume the system rolls a new random number on EVERY event that happens and that this happens server side to avoid exploitations.</DIV> <DIV> </DIV> <DIV>Now if all of the above is true then we are seeing a problem.  We frequently see results in an evening that are at least 2 standard deviations outisde the norm.  Normally 1-2 players seem to get a lot of the loot and 1-2 get very little.  I have personally never seen an even distribution.  I don't mean exact, I mean even close.  In a session with 50 drops in a group of 5 players one player always seems to get 15+ and 1 will get under 5.  The odds of this happening as often as it does are astromically low ... so somethign is going on somewhere.  BUT we the players can't help unless we get soem simple answers </DIV> <DIV>SO!!! Can someone from SOE post somewhere how lotto works...  this is not a information that is going to destroy roleplaying.  It won't ruin the game.. we all think we know how it works anyway... if it doesn't tell us.  If you want intelligent feedback you need to arm your players with the facts.</DIV> <DIV>Something is very wrong at the moment.  I have one character that almost never wins the lotto.  Its not a problem for me since I play with the same guild mates all the time and we just adjust the loot at the end of the night.  But it is strange.  My one character has come out with the least lotto wins [or equal worse] at least 50 times in a row... this is beyond bad luck ... this is statistically impossible or very close.</DIV> <DIV>We need answers .. and I think we can handle them LOL</DIV> <DIV>END OF RANT!!!</DIV> <DIV>BTW there are simialr problems with all random number driven events, such as trade skill events.  But lets stick to one example for now. </DIV> <HR> </BLOCKQUOTE>Where did you get that Idea? Before launch I read about lotto looting and understood it to be <a href="http://www.random.org/" target=_blank>random</a>. Never met anyone that didnt undedrstand that. Now there maybe an agrument about weather the lotto is truely random

[Ashlian]

<BR> <BLOCKQUOTE> <HR> CheezeBurgerAnimal wrote:<BR> <BLOCKQUOTE> <HR> </BLOCKQUOTE>Where did you get that Idea? Before launch I read about lotto looting and understood it to be <A href="http://www.random.org/" target=_blank>random</A>. Never met anyone that didnt undedrstand that. Now there maybe an agrument about weather the lotto is truely random <BR> <HR> </BLOCKQUOTE> <P>I think this is exactly an argument about whether or not the lotto is truly random. Though by the definition you give above, and by the definition I've heard used for any computer game, there is no such thing as a true random number. They all use pseudo random number generators, due to the difficulty inherent in providing a truly random seed for an algorithm. The true random number generators I've heard described use something much as they described in the webpage you cited, like a measurement taken of radioactive decay. Somehow I think they don't have a lump of uranium or radium over at the EQ2 server farm, so they're very likely using a pseudo random number generator, which relies on an algorithm to more or less approximate randomness. The questions being posed here have to do with how the seed is being generated, which can have immense effects on the results you'll get from the number generator. I think we all understand that it's SUPPOSED to be random. Or at least as random as you can get without devoting processing power to it that a real time video game can't afford to. I would love to hear more, this effects me even more as far as harvesting is concerned.....it's a guild joke that I harvest so many less rares than the couple of other people who devote the same amount of time to it.</P> <P>Ashlian Liadan, 32 Fury, 32 Tailor of Mistmoore</P> <P>PS - To anyone questioning SavinDwarf's qualifications as an associate professor.....England does not have exactly the same educational system as the US. You can indeed commonly find people with bachelor's degrees as professors in countries other than the US. Whether this is true in his case, heck if I know, but it's certainly possible. It even happens in the US if the field is esoteric enough.</P>

[Big Da]

<BR> <BLOCKQUOTE> <HR> Ashlian wrote: <P>PS - To anyone questioning SavinDwarf's qualifications as an associate professor.....England does not have exactly the same educational system as the US. You can indeed commonly find people with bachelor's degrees as professors in countries other than the US. Whether this is true in his case, heck if I know, but it's certainly possible. It even happens in the US if the field is esoteric enough.<BR></P> <HR> </BLOCKQUOTE><BR> <DIV>FYI Professors dont need a degree at all, it is a title or position not a qualification.</DIV>

[SavinDwa]

<P>hmm...  Terrorising has brought up an interesting angle I had not though of.  We know they are probably assigning us numbers and doing a roll after that.  What if the numbers are assigned based on the order you click on the loot button?  I'm a ranger, I'm constantly running all over the place to get behind mobs.  In addition, in a group of six players the mobs are dieng so fast and I'm always making sure that I engage the mob the main tank is on.  When things start to look bad I'm hovering over the evac button.  Our group kills non stop, we leave to the nuker and the second tank to get the loot. Main tank, Ranger and 2 healers are always on to the next group. So to cut a long story short, I almost never do the looting (unless its a chest and someone screams out chesty!!).  In addition, and this is what makes me different, I never have time to read what the loot is and just click OK get get the message off the screen.  I bet I'm the first to do that almost every time.  We have one of us who always examines the loot and decides whether to decline or not -- they are probably always last, or close to it. </P> <P>If they have the same bug in the lotto code as they hand in the /ran function and they assign the first person to click OK a 1 then it would explain why I get about half the number of loot items as the top 3-4 in the group (we normally have 5 in our group -- sometimes 6).</P> <P>Now if that is true .. I wonder what the exploit that Terrorrising is talking about?  Going last would be bad, unless there is another wrinkle?  To make matters worse, I've started to try and look at the lot more if its not a common drop and I've noticed that my share has gone up?  Of course they could have stealth fixed the problem sometime in the last 6 weeks.  the bad news is that if its based on the order of clicking OK then its goign to be very tough to go through the logs and get meaningful data.  The last time I got meaningful data I just got everyone to agree to click yes on anything for the whole night (we would pass the stuff around afterwards) and I got them to do this for about 6 sessions) -- but I bet I was still faster than most at clicking yes .. most people take the time to smell the roses and read what it is... my view was if I get it then I can look at it later and if I don't then why wprry LOL.</P> <P>It still would help if SOE would tell us how this works.  I really hope there is not an exploit?  But I can't see what they would be doing to allow one.  ??</P> <P><SPAN class=time_text>P.S. I'm ignoring all the conversations about probablility and statistics.  It was going down hill fast.  The thread was beginning to bring up a lot of formaula to do with statistcial sampling when boas might creep in.  Here we are dealing with a "known" formula -- i.e. we should each have an equal chance to get each item that we don't decline on.  If we get a sample of 1000 loots where they players all had open slots and no one declined on any lotto or was unable to get it becaise it was lore we should se a relatively even spread.  In a 5 player group that would be 200 each.  If the results come up with 250, 250, 250, 125, 125 for the 5 players there is something wrong.  The odds of that happening by chance are astronomical.</SPAN></P><p>Message Edited by SavinDwarf on <span class=date_text>05-24-2005</span> <span class=time_text>06:37 AM</span>

[Ashlian]

<BR> <BLOCKQUOTE> <HR> Big Dave wrote:<BR> <BR> <BLOCKQUOTE> <HR> Ashlian wrote: <P>PS - To anyone questioning SavinDwarf's qualifications as an associate professor.....England does not have exactly the same educational system as the US. You can indeed commonly find people with bachelor's degrees as professors in countries other than the US. Whether this is true in his case, heck if I know, but it's certainly possible. It even happens in the US if the field is esoteric enough.<BR></P> <HR> </BLOCKQUOTE> <DIV>FYI Professors dont need a degree at all, it is a title or position not a qualification.</DIV><BR> <HR> </BLOCKQUOTE> <P>What I meant with the comment quoted was that you can be considered to have met the requirements many accredited schools have for professorships with a bachelor's degree, but you are more likely to see this situation in countries other than the US. I also believe that, while it's certainly possible to create your own University of Everquest and call yourself Professor Spindlecog, that accredited schools generally have certain requirements for professorships, based on the requirements for accreditation to teach a certain topic (which certainly varies by country). Accreditation is a very big deal.....an entire graduating class at a law school in Florida was required to take their entire curriculum over while I was there, because the school was accredited after they'd been accepted into the program and they wouldn't make it retroactive. Poor lawyers!</P> <P>On the topic at hand, I hope that you statistical gurus can determine if this has been fixed or not. I have a tendency, for space's sake, to turn down things like goblin meat (yuck). This is inspiring me to attempt to log a few nights of everyone in my guild group lottoing....there seemed to be some difference in my "luck" after we reformed, but it wasn't a large enough sample by far. Savin, thanks for all your hard work, I'll try to put some effort in myself.</P> <P>Ashlian Liadan, 32 Fury, 32 Tailor of Mistmoore<BR></P>

[ogie]

<P>Hi,</P> <P>As someone with over 20 years of software development experience, I find that this issue is one of the most irksome, because it has such a significant  impact (causation not correlation) upon one's rewards in this game.</P> <P>I haven't read through every post in this thread, however I didn't notice that anyone highlighted one specific sub-issue with this design flaw, and that is chests that have multiple objects inside. When one is opened (had 3 last night in Sol's Eye) each group member's loot window shows several items, with the option to lotto for 'Selected', 'All', or 'Cancel'. </P> <P>In what is clearly a flawed implementation of the concept, the distribution of the loot from such a chest on a regular basis is that all items go to a single person. It seems highly unlikely that each item in the chest is being given a separate 'roll' and then those members subscribing to each item for that roll are being given an equal chance based upon the distribution to win the item.</P> <P>Has anyone else seen evidence of this issue?</P> <P>Thanks,</P> <P>Twilshar</P> <P> </P> <p>Message Edited by ogiers on <span class=date_text>05-25-2005</span> <span class=time_text>12:55 PM</span>

[SavinDwa]

<P>Ogiers,</P> <P> </P> <P>I noticed that last night. What makes it a bit strange is I think I've seen cases where two people got stuff.....but I can't be sure...I wonder if the code is working like this:</P> <P>1) roll at the chest level.  FInd winner.  Give them all the items you can or that they selected.  </P> <P>2) If there are still items in the chest roll again and repeat (I think it may even ask again .. I swear I've seen the same chest come up more than once and with less items the second time??)</P> <P> </P> <P>I really wish a dev would jump in here and tell us how it works ... there are lots of quirks.</P> <P>Last night I started making a point of being very slow to click on the button when the loot window pops .... my percentage of srops went through the roof.  Of course it was a bit confusing since I know some people were declining as well ....</P>

[molinaro]

<DIV>SavinDwarf,</DIV> <DIV> </DIV> <DIV>Some numbers for you:</DIV> <DIV> </DIV> <DIV>50 drops, chance at 5 or less wins with 5 people is 4.8%</DIV> <DIV>50 drops, chance at 15 or more wins with 5 people is 6.7%</DIV> <DIV> </DIV> <DIV>So, how many times a day do you think a group of 5 people lotto's on 50 drops, accross all EQ2 servers?  I would say the number is in the 1000s, meaning that your 'highly improbable' situation better be hapening to dozens of groups every day.  If it wasn't hapening, it would be a sure sign of a problem.</DIV> <DIV> </DIV> <DIV> </DIV> <DIV> </DIV> <DIV>"character in that group, which is a Ranger, has NEVER ended up the night with anything but the worst or equal worse loot count.  We adjust it all at the end of the night anyway so I don't care, but its a problem.  In total we are talking about 2000+ loot drops.  Our group varies from 5-6 players most of the time, we may have had one night with 4 players?  Even if we assume that there were 6 of us I should have gotten about 300 loot drops.  I have had more like 130-150."</DIV> <DIV> </DIV> <DIV>You do understand that off the cuff talk like that is not equal to a sample?  If you have logs of each hunting session, showing who lotto'd for what, and who won what.. well that's a sample.</DIV> <DIV> </DIV> <DIV> </DIV> <DIV>"The problem is that there is almost no known proven random source and hence all information dervived from such a source may in the end begin to show a pattern. "</DIV> <DIV> </DIV> <DIV>Random number generators are not truly random.. that's true.  The way in which they are not random is that after a fixed length of a string of random numbers, they repeat the exact same sequence, if not reseeded.</DIV> <DIV> </DIV> <DIV>With an MMORPG that is irrelevant.  They spit far too many before repeating for it to be of any concern.  And most importantly, you are sharing the RNG's output with every other player, and every other thing going on in the game!  You make it sound as if the random number generator is sitting there doing nothing while it waits for you client to ask for another lol.</DIV> <DIV> </DIV> <DIV>You may get the 456th, 612th, 12345th, 13008th, etc.. number that it spits out over the past hour.</DIV> <DIV> </DIV> <DIV>You client isn't dealing with the RNGs output.  It is dealing with A RANDOM SAMPLE of it's output.</DIV> <DIV> </DIV> <DIV> </DIV> <DIV>----------------</DIV> <DIV> </DIV> <DIV>I want to make one thing clear.  Everything I just said has nothing to do with saying that there is a bug, or there is no bug.   I have no opinion, and no facts on that question.  </DIV> <DIV> </DIV> <DIV>All I am trying to point out is the difference between 'evidence' and false assumptions.</DIV>

[molinaro]

<P>Ogiers,</P> <P>To put it simply, NO.</P> <P>That is completely false.  I hate the way that line keeps getting repeated without anyone actualy checking if it's true.  I've checked.  I decided to pay carefull attention since the first time I saw that comment months ago.</P> <P>I've seen 1 person get all the loot, 4 out of 35 chests with multiple items, so far.</P> <P> </P>

[SavinDwa]

<P>Molinaro,</P> <P>My actuals test on this were not extensive.  I turned logging on for only six sessions.  I turned logging off as soon as someone saif their inventory was full.  I gnored any lore items because they would cause statistical problems. I treated all other loot equally.  I asked everyone to not decline on anything.</P> <P>On 4 occassions we had a group of 5, one occassiona a group of 6 and once a group of four.  In all cases we had two players that had about the equal lowest number of lotton wins, the rest were about the same and about twice the number of wins of the lowest two.  The samples ranged from 45 drops to 70 drops, the total was about 350 drops.  I was one of the low loot wins on all 6 logs.  The other low person varied.  But two always were in the high group.  But I could never find a connection there, or why I was the low.  I'm pretty sure it had nothing to do with the order of the group being filled, I tried all sorts of stuff to break that.</P> <P>i don't have the time to check the math so lets just asusme you are correct.  4.8% of the time there will be one person with 5 drops out of the 50.  What are the odds of getting one person with 5 drops out of 50 in 6 consecutive test of 50?  0.48 ^^ 5 .. pretty slim.  What are the odds of getting two people with 8 or lower and three with 11 or higher? in 5 consecutive batches?  I never had a single player get within 10% of what they would normally have expected to get?  For instance in a group of 5 we should see 10 on average and 9-11 should really be the mean.  I never saw one 8, 9 or 10 and only one 11 in the whole sample.  (real numbers are different since the loot total changed from night to night).</P> <P> </P> <P>I have not repeated these tests in the last 12 weeks.  These were all done late january to early feb.</P> <P>Conclusive evidence of a problem? No .. worthy of a look at the code .. yes?  the odds of getting those results were astronomical.  Might they have stealth fixed it?  maybe, they stealth fixed the random number bug.  How likely is it there is/was a bug?  High, the pattern follows the exact same problem the random number bug had.  Has it been fixed?  maybe.. but it requires testing to find out and abnormal behavior in the group to get stats .. That is a pain since we dont; really care and adjust anyway, decline all over the pace, and give stuff to who needs it.</P> <P>By the way, the mathematics of calculating the chance of getting two people with 6 or lower results in 5 when picking 50 items is fun and not as easy as it seems.  </P>

[Byzanth]

<DIV><FONT color=#99ff00>I would assume that the RND function is at the core of the script for "fizzling".</FONT></DIV> <DIV><FONT color=#99ff00></FONT> </DIV> <DIV><FONT color=#99ff00>Just see for yourself how streaky "Fizzling" can be.</FONT></DIV> <DIV><FONT color=#99ff00></FONT> </DIV> <DIV><FONT color=#99ff00>Cast a Spell and then Jump to interupt it.... Do this over and over again.</FONT></DIV> <DIV><FONT color=#99ff00></FONT> </DIV> <DIV><FONT color=#99ff00>You will see something very similar to this. </FONT></DIV> <DIV><FONT color=#99ff00></FONT> </DIV> <DIV><FONT color=#99ff00>Fizzle</FONT></DIV> <DIV><FONT color=#99ff00>Fizzle</FONT></DIV> <DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV><FONT color=#99ff00>Fizzle</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV><FONT color=#99ff00>Fizzle</FONT></DIV> <DIV><FONT color=#99ff00>Fizzle</FONT></DIV> <DIV><FONT color=#99ff00>Fizzle</FONT></DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV><FONT color=#99ff00>Fizzle</FONT></DIV> <DIV><FONT color=#99ff00>Fizzle</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV> <DIV><FONT color=#99ff00>Interupted</FONT></DIV> <DIV><FONT color=#99ff00></FONT> </DIV> <DIV><FONT color=#99ff00> The Random numbers kind of "Stick" so to speak.  That's why Player 2 sometimes gets the complete contents of that master chest <img src="/smilies/8a80c6485cd926be453217d59a84a888.gif" border="0" alt="SMILEY" /></FONT></DIV> <DIV><FONT color=#99ff00></FONT> </DIV> <DIV><FONT color=#99ff00>Why do you think that the national lottery still uses pingbong balls with numbers on them?   Answer: Because people don't trust computers.</FONT></DIV> <DIV><FONT color=#99ff00><EM></EM></FONT> </DIV> <DIV><FONT color=#99ff00><EM>(Last night I walked away with 3 out of 48 items in a 5 person group.)</EM></FONT></DIV> <DIV><EM><FONT color=#99ff00></FONT></EM> </DIV> <DIV><EM><FONT color=#99ff00></FONT></EM> </DIV> <DIV><FONT color=#99ff00>Yeah... I'm for "Automated Round Robin" too. <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" /> <EM> </EM></FONT></DIV> <DIV><SPAN class=time_text></SPAN> </DIV></DIV></DIV></DIV></DIV></DIV></DIV></DIV></DIV></DIV></DIV></DIV></DIV></DIV></DIV> <P><SPAN class=time_text></SPAN> </P><p>Message Edited by Byzanthei on <span class=date_text>08-19-2005</span> <span class=time_text>03:09 PM</span>