Simple logic: The vast majority of what you term 'real' team sports are played by teams where the player membership is long-term and generally static from game to game. Allegiance, on the other hand, is generally played as pickup games where the teams are of arbitrary composition (the obvious exception being SGs). You're quite right in assuming that Trueskill would perform poorly when applied to 'real' team sports, since the sytem has no way of differentiating between players who always play on the same team in the same games. The assumption does not hold true for pickup games.Fragtzack wrote:QUOTE (Fragtzack @ Dec 14 2007, 09:23 AM) I just don't get how such smart people like you guys doing this development are overlooking such very simple logic.
Spidey's Command List
Granary Sergeant Baker - Special Bread Service (Wurf - 13th Oct 2011)
Seriously, we've gone to considerable lengths to approach the ranking problem from the correct angle given the specifics of our circumstances. Whilst I appreciate the ideas being put forth in this thread, it's not as if there have been any major revelations. Spending eighteen months on a project affords one the time to consider all aspects of the available solutions. As certain people have noted, performance stats have a definite place within the broader framework of any eventual Alleg stats system. The problems with performance stats arise when attempting to balance games in a mathematically valid fashion based on said data. To put it colloquially "How many probe drops are worth one eject?" (I believe this was already mentioned elsewhere)
AllegSkill can be considered to have two distinct phases of development:
AS Mk I deals solely with ranking issues and game balance. To be quite honest the ranks generated are only useful when an appropriate balance system is applied to games, whether that be autobalance or spontaneous community balance. Your rank contains no information regarding your specific talents in Allegiance, just a rating of your tendency to win games based on some unobserved in-game behaviour.
AS Mk II is intended to provide a plethora of performance-related stats. I for one love whoring miners, appearing at the top of the in-game leader board etc etc. I'm absolutely certain that I'm not alone. Whilst AS Mk I allows games to be balanced based on players' 'overall' skill, AS Mk II would rate specific skills. As stated, balancing games based on MkII stats is a mathematical minefield, and as such is unlikely to be attempted by the stats group. This does not, however, detract from the validity of performance stats when they're considered in the correct context. Components of AS specific to Mk II will definately not use Trueskill.
(and yes, nanning/probing/flying in circles will count /smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" />)
AllegSkill can be considered to have two distinct phases of development:
AS Mk I deals solely with ranking issues and game balance. To be quite honest the ranks generated are only useful when an appropriate balance system is applied to games, whether that be autobalance or spontaneous community balance. Your rank contains no information regarding your specific talents in Allegiance, just a rating of your tendency to win games based on some unobserved in-game behaviour.
AS Mk II is intended to provide a plethora of performance-related stats. I for one love whoring miners, appearing at the top of the in-game leader board etc etc. I'm absolutely certain that I'm not alone. Whilst AS Mk I allows games to be balanced based on players' 'overall' skill, AS Mk II would rate specific skills. As stated, balancing games based on MkII stats is a mathematical minefield, and as such is unlikely to be attempted by the stats group. This does not, however, detract from the validity of performance stats when they're considered in the correct context. Components of AS specific to Mk II will definately not use Trueskill.
(and yes, nanning/probing/flying in circles will count /smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" />)
Last edited by sgt_baker on Fri Dec 14, 2007 11:58 am, edited 1 time in total.
Granary Sergeant Baker - Special Bread Service (Wurf - 13th Oct 2011)
Professional sport teams individuals vary a lot from season to season, especially in the NFL. Pretty sure that only one, maybe two players were with Steve Largeant from his starting year to his ending year with the Seahawks. Statements about the real sports teams being static for long term doesn't hold much truth according to the record book of team rosters.
We are not talking about ranks in Allegiance over just a one month period or just one year. We are not talking about ranks in Pro sports over just one season. Most ranking systems are based on all time performance. Based on all time performance, there was huge differential between Steve Largeant's starting team players and his ending season players. Allegiance systems are based on all time performance, my analogy to Steve Largeant is based on his all time performance. Limiting the scope of team differential to a specific time period such as one season and then applying that time period stats to an all time ranking doesn't work in either system. My analogy to Steve Largeant all time ranking and Trueskill is very valid. Pro teams rosters change, a lot. So does high school and college rosters.
The original ranking system of Allegiance was fairly accurate with only one glaring problem, the lack of accurate recording of scout actions and ranking those scouting actions. Why the big need to go to a radically different ranking system from the accepted norm that has been used for hundreds of years? I am fairly certain my previous post answered this question, but I am hoping some one can show a logical good for the community reason instead of a personal subjective reason. There is technical reasons why it would be difficult to reward scout whores with the AZ system. However, this coding issue that can be over come with less effort and less pain to the community then a new system.
The biggest reason until proven otherwise was a few scout whores complained about lack of recognition. This number of scout whores is very, very small in comparison to the masses that were on the AZ and the masses that wanted the AZ system to be brought back a few years ago. Vote I tell ya, you will see.
Everyone in the top 20 ranks of the AZ belonged there, or at least everyone in the top 20 of the AZ were ranked accurately enough to be considered being in the top 40 or so.
Individual ranking system for team based for real sports = fairly accurate. Yes, polls are the best way...but people base their votes off stats.
Individual ranking system for Allegiance using the AZ system = fairly accurate. Yes, scout whores got the shaft...but scout whores was the exception problem and not the standard.
hmmm, two fairly accurate systems based on the same theory.
Elo+Helo. Two fairly inaccurate systems based on the same theory. Now the community is going to be forced to endure another 2 years of a new experiment based on this same theory?
We have two proven systems that generate fairly accurate results with real team sports and the Allegiance AZ. Neither is totally accurate, no system will be. However, the experiment the community has been forced to endure for the last 2 years is very,very inaccurate and Trueskill is based on the same theory.
You do know that the AZ ranking system rewarded people for winning? hmmmm...
Trueskill does start to sound really good and accurate when teams are autobalanced but I still don't see a new unproven system is better for this community then the proven AZ system or the proven system real team sports are based on. I am fairly certain MS Research whole theory of trueskill is based on autobalanced assumption. We can not have autobalance in Allegiance because of a opinions of the minority though. I highly disagree that the masses of Allegiance do not want autobalance. I contend the naysayers are the minority. Once again, a vote can solve this issue.
Ask the community about autobalance, there is a great voting infrastructure in place. Instead, a few will decide for the community what the community really wants? Hmm...Wonder why the "D" word is bantered about?
I am highly against the current ranking system and a Trueskill system until the hard questions are asked from the community in the form of a vote. We have been forced to endure an experiment for the last 2 years that has produced very inaccurate results. These inaccurate results are producing huge negatives for this community in the form of stacking. People leave this community forever every week because of stacking. I have never heard of anyone leaving due to a lack of rankings or lack of stats.
In closing,
Pushing forth the theories of a few people without accurately assessing what the community wants causes a lot of discontent and people lose faith in the ones forcing their radical ideas upon others. Your development work is good..but your numbers are very small compared to the community at large. My personal opinion is no rankings is best. However, if the community votes for a ranking system and votes for a ranking system based on Elo theory I will support the masses opinion. I do not support the theories of a few being forced as an experiment on the many.
We are not talking about ranks in Allegiance over just a one month period or just one year. We are not talking about ranks in Pro sports over just one season. Most ranking systems are based on all time performance. Based on all time performance, there was huge differential between Steve Largeant's starting team players and his ending season players. Allegiance systems are based on all time performance, my analogy to Steve Largeant is based on his all time performance. Limiting the scope of team differential to a specific time period such as one season and then applying that time period stats to an all time ranking doesn't work in either system. My analogy to Steve Largeant all time ranking and Trueskill is very valid. Pro teams rosters change, a lot. So does high school and college rosters.
The original ranking system of Allegiance was fairly accurate with only one glaring problem, the lack of accurate recording of scout actions and ranking those scouting actions. Why the big need to go to a radically different ranking system from the accepted norm that has been used for hundreds of years? I am fairly certain my previous post answered this question, but I am hoping some one can show a logical good for the community reason instead of a personal subjective reason. There is technical reasons why it would be difficult to reward scout whores with the AZ system. However, this coding issue that can be over come with less effort and less pain to the community then a new system.
The biggest reason until proven otherwise was a few scout whores complained about lack of recognition. This number of scout whores is very, very small in comparison to the masses that were on the AZ and the masses that wanted the AZ system to be brought back a few years ago. Vote I tell ya, you will see.
Everyone in the top 20 ranks of the AZ belonged there, or at least everyone in the top 20 of the AZ were ranked accurately enough to be considered being in the top 40 or so.
Individual ranking system for team based for real sports = fairly accurate. Yes, polls are the best way...but people base their votes off stats.
Individual ranking system for Allegiance using the AZ system = fairly accurate. Yes, scout whores got the shaft...but scout whores was the exception problem and not the standard.
hmmm, two fairly accurate systems based on the same theory.
Elo+Helo. Two fairly inaccurate systems based on the same theory. Now the community is going to be forced to endure another 2 years of a new experiment based on this same theory?
We have two proven systems that generate fairly accurate results with real team sports and the Allegiance AZ. Neither is totally accurate, no system will be. However, the experiment the community has been forced to endure for the last 2 years is very,very inaccurate and Trueskill is based on the same theory.
You do know that the AZ ranking system rewarded people for winning? hmmmm...
Trueskill does start to sound really good and accurate when teams are autobalanced but I still don't see a new unproven system is better for this community then the proven AZ system or the proven system real team sports are based on. I am fairly certain MS Research whole theory of trueskill is based on autobalanced assumption. We can not have autobalance in Allegiance because of a opinions of the minority though. I highly disagree that the masses of Allegiance do not want autobalance. I contend the naysayers are the minority. Once again, a vote can solve this issue.
Ask the community about autobalance, there is a great voting infrastructure in place. Instead, a few will decide for the community what the community really wants? Hmm...Wonder why the "D" word is bantered about?
I am highly against the current ranking system and a Trueskill system until the hard questions are asked from the community in the form of a vote. We have been forced to endure an experiment for the last 2 years that has produced very inaccurate results. These inaccurate results are producing huge negatives for this community in the form of stacking. People leave this community forever every week because of stacking. I have never heard of anyone leaving due to a lack of rankings or lack of stats.
In closing,
Pushing forth the theories of a few people without accurately assessing what the community wants causes a lot of discontent and people lose faith in the ones forcing their radical ideas upon others. Your development work is good..but your numbers are very small compared to the community at large. My personal opinion is no rankings is best. However, if the community votes for a ranking system and votes for a ranking system based on Elo theory I will support the masses opinion. I do not support the theories of a few being forced as an experiment on the many.
Do you plan on doing anything with the 'Squad Game' thing in the game screen thing? As you said, calculating ranks would be different for SGs, so are you going to do that?sgt_baker wrote:QUOTE (sgt_baker @ Dec 14 2007, 05:04 AM) Allegiance, on the other hand, is generally played as pickup games where the teams are of arbitrary composition (the obvious exception being SGs).
I'd assert that Elo is accurate when applied to head-to-head games. It was designed for this purpose and should be constrained to this application. Applying Elo to team-based games in a haphazard manner (which is exactly what Helo does) causes countless problems.
The Elo/Helo effort wasn't really based on any particular theory. The 'experiment' was a kludge of the highest order. Associating Truskill with said experiment is lunacy.
This is growing tiresome. I've never stated that AllegSkill is going to be implemented. It's not up to me. The reason we've been working on the project for 18 months IS TO AVOID ANOTHER ELO/HELO CLUSTER$#@! AND INVESTIGATE THE VIABILITY OF ANY RANKING SYSTEM IN ALLEGIANCE.
(for some reason it wouldn't let me include quotes...)
The Elo/Helo effort wasn't really based on any particular theory. The 'experiment' was a kludge of the highest order. Associating Truskill with said experiment is lunacy.
This is growing tiresome. I've never stated that AllegSkill is going to be implemented. It's not up to me. The reason we've been working on the project for 18 months IS TO AVOID ANOTHER ELO/HELO CLUSTER$#@! AND INVESTIGATE THE VIABILITY OF ANY RANKING SYSTEM IN ALLEGIANCE.
(for some reason it wouldn't let me include quotes...)
Granary Sergeant Baker - Special Bread Service (Wurf - 13th Oct 2011)
How people can decide whether a system works or not without understanding the mathematical fundaments of the system or having seen any of the generated results is beyond me. Oh, I remember, this is the internet... everyone is a $#@!ing genius.
Granary Sergeant Baker - Special Bread Service (Wurf - 13th Oct 2011)
No. AllegSkill is designed specifically to balance pickup games. SGs form such a small number of the total set of games played that they don't warrant an additional layer of abstraction.Andon wrote:QUOTE (Andon @ Dec 14 2007, 12:54 PM) Do you plan on doing anything with the 'Squad Game' thing in the game screen thing? As you said, calculating ranks would be different for SGs, so are you going to do that?
Granary Sergeant Baker - Special Bread Service (Wurf - 13th Oct 2011)
Frag
Apples and Oranges on the NFL vs Allegiance and the difference is why it works. Steve played for the same team his entire career not a different team made up of essentially different people evey week for his career.
Next the orgional ranking system was not a ranking system but rather a points based system. A ranking system should use statisical evidence to achieve a rank. Points based are time driven, think exactly WoW. It also causes player behavior changes based on how the points are distributed. Remember when no on would fly a bomber in game until the devs changes points and people would base sit waiting for the chance to fly one? Or how KevDude spent countless hours scumming newbie kills simplely to be top dog. Or good ol Sabertooth who would drop rather then suffer a pk. And not to shock you but I got into the top 40 or nearly so IIRC [ might be suffering from memory ego creep ] at one moment during AZ and you didnt even know I was an AZer /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" /> I freely admit mine was time based since well I played a $#@! of a lot.
Finally I understand you scepticism and all with regards to this ranking system given the past but I urge you to try to keep an open mind. Also please form a group make it a large one since your sure to suffer massive casualities once everyone understands what they are tackling. Gather your in game stats, figure out how to make them meaniful in a valid staticsical based model, rigorously check the results, figure a way to keep checking the validity of your assumptions, and then bring it forward. Until then remember saying something is relatively easy is what got me in this mess in the first place /laugh.gif" style="vertical-align:middle" emoid=":lol:" border="0" alt="laugh.gif" />
edit: had maybe instead of made up and bunch more
Apples and Oranges on the NFL vs Allegiance and the difference is why it works. Steve played for the same team his entire career not a different team made up of essentially different people evey week for his career.
Next the orgional ranking system was not a ranking system but rather a points based system. A ranking system should use statisical evidence to achieve a rank. Points based are time driven, think exactly WoW. It also causes player behavior changes based on how the points are distributed. Remember when no on would fly a bomber in game until the devs changes points and people would base sit waiting for the chance to fly one? Or how KevDude spent countless hours scumming newbie kills simplely to be top dog. Or good ol Sabertooth who would drop rather then suffer a pk. And not to shock you but I got into the top 40 or nearly so IIRC [ might be suffering from memory ego creep ] at one moment during AZ and you didnt even know I was an AZer /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" /> I freely admit mine was time based since well I played a $#@! of a lot.
Finally I understand you scepticism and all with regards to this ranking system given the past but I urge you to try to keep an open mind. Also please form a group make it a large one since your sure to suffer massive casualities once everyone understands what they are tackling. Gather your in game stats, figure out how to make them meaniful in a valid staticsical based model, rigorously check the results, figure a way to keep checking the validity of your assumptions, and then bring it forward. Until then remember saying something is relatively easy is what got me in this mess in the first place /laugh.gif" style="vertical-align:middle" emoid=":lol:" border="0" alt="laugh.gif" />
edit: had maybe instead of made up and bunch more
Last edited by MrChaos on Fri Dec 14, 2007 2:03 pm, edited 1 time in total.
Ssssh
-
takingarms1
- Posts: 3052
- Joined: Fri Feb 20, 2004 8:00 am
It is clear to me from reading these posts that the people working on AllegSkill are very bright and have put quite a bit of time and effort into the undertaking. Before we jump to conclusions or criticize something we have not yet even seen, why don't we wait for the AllegSkill team to give us a proper explanation and some of their testing results?
"You give my regards to St. Peter. Or, whoever has his job, but in hell!"
- - - -
- - - -
Raveen wrote:QUOTE (Raveen @ Dec 14 2007, 02:02 AM) Oh and as for the perceived slights by RT, well provocation goes a long way doesn't it Bax? Particularly amusing is you taking the high ground towards a squad that you jumped from before you were kicked for being a jerk.
Aren't you full of @#(!. /laugh.gif" style="vertical-align:middle" emoid=":lol:" border="0" alt="laugh.gif" />
You were among one of many that tried to not get me to leave. Dumbass. /laugh.gif" style="vertical-align:middle" emoid=":lol:" border="0" alt="laugh.gif" />
EDIT: Hard to say the disgust I feel towards your statement Rav. /rolleyes.gif" style="vertical-align:middle" emoid=":roll:" border="0" alt="rolleyes.gif" />
Last edited by jgbaxter on Fri Dec 14, 2007 4:05 pm, edited 1 time in total.
n.b. I may not see a forum post replied to me or a pm sent to me for weeks and weeks...