APCR (ASL Player Comparative Ratings)

[Aaron Cleavin]

I don't see any reason to change from AREA on the basis of remarks so far.
AREA can be revitalized and maintained.
At the very least, no matter what, AREA ratings must be updated to include the last two years of results, so we have a benchmark to compare those results with this new method based on chess, which, by the way, has no random element, so I'm not convinced that it is better by any means.

I am sure everyone would like to see what the new ratings would be after being transmogrified by this new method. Wouldn't it be interesting if Steve Pleva was all of a sudden ranked lower from #1?

No, we absolutely need AREA as a benchmark.

Chess has actually random elements in terms of Human performance

If you read the paper you will see it is not linked to Chess in any way , rather to statistics
AREA was based on ELO which is based in the way Chess ranking were done 30 years ago, GLICKO has a number of advantages over this, the paper is only 6 pages and is actually quite accessible.

 

[Honosbinda]

Chess has actually random elements in terms of Human performance

If you read the paper you will see it is not linked to Chess in any way , rather to statistics
AREA was based on ELO which is based in the way Chess ranking were done 30 years ago, GLICKO has a number of advantages over this, the paper is only 6 pages and is actually quite accessible.

Random by definition effectively means 'happening without method or conscious decision' and I would say that poor performance in chess is still linked to cognition. Although some aspects of move choice may be subconscious, I don't think chess players are playing at random, no matter how badly they play. In order to be random, one method would be to find all the allowable moves, assign each move an integer (or a range of) value, then use a randomization method (white noise) to pick the move they are next going to do.

So we disagree on that. But that's a quibble

I can certainly accept statistical means and trust your and others' judgement on the matter of GLICKO being superior. I'm only asserting that the AREA data must be accounted for and if there are transformations in rank afterwards, there would be questions, right?

Above, you've implied you've had a look at the raw AREA data, which is great. The hard part is doing what is necessary, that is, back-tracing the played scenarios and documented results that have missing data and determining the missing data! It must be done. Not sure why attacker or defender is important information in this case.

Also, if it's important to include draws in the actual play of ASL in tournaments, then that information needs to be included, not excluded so that it fits the model. Which shouldn't be a problem.

 

[Aaron Cleavin]

Random by definition effectively means 'happening without method or conscious decision' and I would say that poor performance in chess is still linked to cognition. Although some aspects of move choice may be subconscious, I don't think chess players are playing at random, no matter how badly they play. In order to be random, one method would be to find all the allowable moves, assign each move an integer (or a range of) value, then use a randomization method (white noise) to pick the move they are next going to do.

So we disagree on that. But that's a quibble

I can certainly accept statistical means and trust your and others' judgement on the matter of GLICKO being superior. I'm only asserting that the AREA data must be accounted for and if there are transformations in rank afterwards, there would be questions, right?

Above, you've implied you've had a look at the raw AREA data, which is great. The hard part is doing what is necessary, that is, back-tracing the played scenarios and documented results that have missing data and determining the missing data! It must be done. Not sure why attacker or defender is important information in this case.

Also, if it's important to include draws in the actual play of ASL in tournaments, then that information needs to be included, not excluded so that it fits the model. Which shouldn't be a problem.

Perhaps it be accurate to say Chess does not (for the most part) have any exogenous random variables involved in play.

APCR Needs a good Scenario code and who attacks and defend to track not only the players ranking , but the performance of a scenario, and of a player on attack/defense.
Games with such data would contribute to a messy environment, the decision could be made later as to whether to include or exclude stuff from AREA (I would think if it is 5% maybe exclude but if more like 15% inclusion might be better)

Draws should not be included in calculation if the game was incomplete rather than an actual draw as rankings will be affected which would be incorrect, few scenarios even allow draws, so for the most part we are talking of incomplete rather than drawn scenarios.
Including them in the db is very easy to do.

 

[Honosbinda]

Perhaps it be accurate to say Chess does not (for the most part) have any exogenous random variables involved in play.

APCR Needs a good Scenario code and who attacks and defend to track not only the players ranking , but the performance of a scenario, and of a player on attack/defense.
Games with such data would contribute to a messy environment, the decision could be made later as to whether to include or exclude stuff from AREA (I would think if it is 5% maybe exclude but if more like 15% inclusion might be better)

Draws should not be included in calculation if the game was incomplete rather than an actual draw as rankings will be affected which would be incorrect, few scenarios even allow draws, so for the most part we are talking of incomplete rather than drawn scenarios.
Including them in the db is very easy to do.

I suppose one could argue that the subconscious mind operates in seemingly random ways.

I hadn't realized you meant for APCR to operate as an adjunct to the ROAR database information (a rudimentary measure of scenario performance but the best we've got to pick 'balanced' scenarios that appear for tourneys). ROAR doesn't take into account relative player abilities, so we have to assume that a large number of scenario playings null these out.

Having APCR determine scenario performance accounting for relative abilities of players is very valuable.

Player on attack/defense -- so does this mean that in chess, players are also ranked as to black/white gameplay performance? Because the way to distinguish this in ASL can be who moves first, as the VC might not readily clarify attacker/defender criteria. I think only the sides are identified in a scenario, and one would have to read the scenario card and assess which side is the attacker/defender.

There are also those cases where the initial move is determined at random. How important is this variable?

In any event, it would seem possible to back-trace a lot of the missing data UNLESS the scenario itself is not specified. In that case, the data is no good and would need to be thrown out.

 

[Aaron Cleavin]

I suppose one could argue that the subconscious mind operates in seemingly random ways.

I hadn't realized you meant for APCR to operate as an adjunct to the ROAR database information (a rudimentary measure of scenario performance but the best we've got to pick 'balanced' scenarios that appear for tourneys). ROAR doesn't take into account relative player abilities, so we have to assume that a large number of scenario playings null these out.

Having APCR determine scenario performance accounting for relative abilities of players is very valuable.

Player on attack/defense -- so does this mean that in chess, players are also ranked as to black/white gameplay performance? Because the way to distinguish this in ASL can be who moves first, as the VC might not readily clarify attacker/defender criteria. I think only the sides are identified in a scenario, and one would have to read the scenario card and assess which side is the attacker/defender.

There are also those cases where the initial move is determined at random. How important is this variable?

In any event, it would seem possible to back-trace a lot of the missing data UNLESS the scenario itself is not specified. In that case, the data is no good and would need to be thrown out.

Some % of ROAR Records have a blank (or poorly formed) Scenario Code, I can guess on the much of the Garbage but not on the blanks.
The ASL Scenario Archive detriines A/D for evey published scenario and I am leveraging this.
Most (86-90%) of AREA record specify who is attacking defending.

2-3% of scenarios have no scenario code and oddly these are all tie games, there are also tie games with a scenario code recorded

 

[Aaron Cleavin]

After Due consideration Draws will be recorded but NOT included in a rating computation as I believe for the most part they reflect incomplete rather than drawn games.

Designing not in a vacuum but with community input is vital

 

[ecz]

Aaron while I have no further remarks about your decision to not count draws, I still have a doubt about the necessity/utility to indicate who is the attacker. I believe that once you have the names of the players, the scenario ID and the sides they have chosen, it's redundant indicate also who was the attacking side. Besides, as a TD director that recorded for AREA 2000+ games over the years, in events where players are free to decide their scenarios, like ASLOK or VASLeague, it's simply impossible for the TD check individually the scenarios to fill this camp. Also note ROAR does not show nothing so, unless you already know the scenario or the players kindly specify this info, you should physically search and check the scenario. So it's really necessary you do not make this information as a prerequisite to be included in APCR. At least for the huge mass of scenarios already recorded in AREA where this info lacks.

also I think that scenarios where the default convention axis/Communist vs not Nazi-Fascist/"bad guy" first is NA because it's impossible to define who is the "bad guy" and who is the "good guy" are probably less than 0,3% so this is a negligible problem

 

[ecz]

I agree with Zov .. eg with Chinese Civil War scenarios ie Red Chinese vs GMT - you might start a new war trying to label whichever side as "bad". I am sure there are other "situations" ..

The default convention I use for VASLeague includes as "first player named" Axis, Fascists, Nazi, NK, and any force fighting VS Allied, US sides or their allied (i.e. the side the average American would call the "good guys"). So in case of Chinese Civil War the Red Chinese player should be named first.

Other "situations" are possible, but let me say that they are not common and we can convive with a few of debatable cases on more than 5000 scenarios.

 

[ecz]

I disagree, I think player one should be whatever player one is on the scenario card.

This info is not useful to see at glance who plays what or who is the attacker unless you have the card in front of you.

 

[ecz]

Player on attack/defense -- so does this mean that in chess, players are also ranked as to black/white gameplay performance? Because the way to distinguish this in ASL can be who moves first, as the VC might not readily clarify attacker/defender criteria. I think only the sides are identified in a scenario, and one would have to read the scenario card and assess which side is the attacker/defender.

There are also those cases where the initial move is determined at random. How important is this variable?

I agree on this, it's the reason why I never asked to VASLeague players to specify attacker/defender but just sides.

 

[hongkongwargamer]

The default convention I use for VASLeague includes as "first player named" Axis, Fascists, Nazi, NK, and any force fighting VS Allied, US sides or their allied (i.e. the side the average American would call the "good guys"). So in case of Chinese Civil War the Red Chinese player should be named first.

Other "situations" are possible, but let me say that they are not common and we can convive with a few of debatable cases on more than 5000 scenarios.

In the Chinese Civil War .. it wasn't entirely clear as to which side the US was on. That's the point (without getting into details that will get me in trouble).

 

[dlazov]

So it should then be displayed/recorded as Attack/Defender.

It's stupid to say Good Guys/Bad Guys or Black/White. it's ASL.

 

[Jacometti]

I only feel I play the bad guys if I roll crap or make silly mistakes.

Mostly I feel I play the poor bastards under my command.

 

[Michael R]

The Canadian ASL Open uses a "draw" option when a match must be adjudicated, but the adjudicators all feel that they cannot pick a winner. Each player receives half of the ten points normally awarded for a win. This coincides with Aaron's view that the match was actually incomplete. I think that I will propose that we use the Steve Pleva method of adjudication in the future. At the New York State Championship, they adjudicate by deciding what is the percentage of chance that a player (both players?; I need to ask) will win. If the adjudicitor decide 80% for example, the player must make a DR that has a cumulative 80% probability. The player would need to roll 9 or less to "win" the match.

 

[von Marwitz]

To clarify around Draws, In Chess a stronger player will lose some ranking when drawing with a weaker player, as he might have some expectation of winning.
If a draw in fact represents and incomplete game then no exchange of ratings should occur.
This is a key design decision which if why I have been reluctant to include draws for now.

I can see Enrico's predicament though, soo -- If a game was started (A scenario and sides were chosen) then we could include it as an "I" incomplete game, recorded in system, but not used in any rating calculation.

This is what happened in the tournaments I attended that allowed draws.

I would support Enrico's suggestion for including the possibilities for draws.
A draw might be an incomplete game. In (some) tournaments this happens when time runs out. Usually people agree to call it a draw if the situation on the board warrants it.

I don't see, why there should no possibility to have this affect one's rating: Provided, I'd manage a "draw" against Pleva, I'd call that quite a feat, while the multiple GROFAZ himself would probably not count it among his best performances.

von Marwitz

 

[Aaron Cleavin]

But as almost no ASL scenario (

This is what happened in the tournaments I attended that allowed draws.

I would support Enrico's suggestion for including the possibilities for draws.
A draw might be an incomplete game. In (some) tournaments this happens when time runs out. Usually people agree to call it a draw if the situation on the board warrants it.

I don't see, why there should no possibility to have this affect one's rating: Provided, I'd manage a "draw" against Pleva, I'd call that quite a feat, while the multiple GROFAZ himself would probably not count it among his best performances.

von Marwitz

There are only a handful of scenarios (《10 out of 7-8000) which have actual draw VC, and these shouldn't be chosen for tournie play. Thus a draw in tournie play almost always represents an incomplete game. Penalizing a player for a case where his opponent ran out the clock is not good practice.

 

[ecz]

So it should then be displayed/recorded as Attack/Defender.

It's stupid to say Good Guys/Bad Guys or Black/White. it's ASL.

Good/Bad guys is obviously an abstraction to get the general idea of what I mean to have an easy recognizable way to discover by the mere order of the players the sides decided. It worked pretty well for 12 years of VASLeague identifying instantly who played the Axis faction. I hope that who will decide how manage this new thing is able to understand what I mean with no silly misunderstanding .

using the order of the card you get absolutely no info about sides or attacker/defender role...

for example a record like this:

john vs roger 0-1 BFP 73 Preliminary Move says nothing apart that roger won unless you do not take physically the card.
Taking the card you'll see that john was the defending Russian. Discover scenario by scenario who played who it's something I cannot do and I suspect most TDs will do neither.

Better to skip the attacker/defender role entry and concentrate only on:

1. result
2. scenario name and
3. sides.

and sides can be easily identified using a default convention like mine or even a better one if exists.

not counting that in many scenarios who is the attacker/defender is not possible to define .
Anyway I have said my opinion, I'm good with this, It's clear enough (IMO) and the ball is the camp of who will decide.

 

[dlazov]

using the order of the card you get absolutely no info about sides or attacker/defender role...

You’re incorrect, there always is a player one and a player two. Player one is always the first firer/mover so player one is always the attacker.

Been that way since ‘85

 

[jrv]

You’re incorrect, there always is a player one and a player two. Player one is always the first firer/mover so player one is always the attacker.

Been that way since ‘85

In Directive Number Three [AP7] the side that moves first is determined after setup and after an initial simultaneous MPh by a dr.

JR

 

[hongkongwargamer]

In Directive Number Three [AP7] the side that moves first is determined after setup and after an initial simultaneous MPh by a dr.

JR

Like ecz said .. this is negligible. Although I can't pull out a magic number like "less than 0.3%" .. he took the time to do the numbers and and sieve thru all the scenarios that might be questionable! I simply can't beat that kind of dedication.