An Accurate Stratego Rating System - What do YOU Think?

[psychicwarrior]

This is from another thread that got caught up in tangential issues, so I decided to make it its own topic here. Comments welcome.

The ranking system's recent changes, my opinion is - Its a step in the right direction but 4 more steps need to be taken if both accuracy in player strength and a more competitive site is the objective.

Those changes, I believe are as follows:

1) The Adequate- Non Adq ratio is a very positive step yet needs refinement.

2) Weight recent games more heavily

3) Adjust and weight favorably more frequent players.

4) A higher minimum number of games to simply qualify for a ranking.


Point #1)

The ratio 1:2 is even slightly too soft in my mind yet I also believe very high ranked players must have a little more leeway as the available players that qualify as adequate to them is very narrow.

Often there are only a few players on Gravon and thus the availability to play top players can be somewhat limited.

For top 5 players, they may rarely see an "Adequate" opponent and quickly be unranked.

Point #2)

What I believe a new rating system must also incorporate is a weighted recency formula. Like WSC but not as radical.

Consider the player with 1000 games.. may for example win 400 of his last 500 games. Quite an amazing feat. Yet if that's counter-balanced with a poor record of his 1st 500 then he looks like an average player when in reality he is a top player based on his last 500 games. More games than most players play in a year.

Compare this to another player who has only 50 games yet won his 1st 35 and lost his last 15 looks far superior and this is a gross inaccuracy.

So I strongly suggest a heavier " weight" for more recent games.


Point #3)

I strongly believe there needs to also be some favour for players who play many games.. so that a true and accurate ranking by player strength is achieved.

It's no secret that the more often you play, the more well known your style, setups and opponents general feel for your game is and thus you are far more beatable than a player who barely plays and little knowledge about his game is known.

Consider for ex. a frequent player has 1000 games on record for the year and then wins 10 in a row, he moves oh so slightly in the rankings with a mere 1% additional win %.

Compare that to a player who has only played 30 games, he jumps up dramatically with 10 wins yet they both accomplished the same feat ! The player with 1000 games is thus in effect severely penalized for his frequent play.

So volume of games/wins should be weighted to accurately assess player strength. This I feel is a must.



Point #4)

Minimum Game Volume to be ranked.

I would suggest an average of 15games/mth so by end of March for ex. only a player with 45 games qualifies and by year end they must have played 180 to be ranked.

That way a player who has played 30 games and displays a high win % is not falsely and inaccurately rated above a player who has 1000 games yet a slightly lower win % etc.

A clear example of this is our beloved friend dcannies who has but 21 games and is ranked ahead of players who have 300 -500 games + and almost as high a win %.

Clearly this is grossly unfair and more importantly very inaccurate to those who have maintained a win % over a much higher volume of games.

This is, in my humble opinion, as big a flaw in the rankings is players who cherry pick and and play non adequate players as has been noted here previously.


So ideally a rating system that weights:

1) Keeping the current Adq/Non Adq ration but expanding the Adq definition for top players.
2) Recency
3) Win/game volume
4) A higher minimum number of games

This will make a ranking system, in my opinion, far more accurate and the site much more competitive.

[gentleben]

Hi Ed, I am glad "you" decided to post this. Here is my response.

Point 1: I think the 1 out of 3 ratio is high enough. Though it is still possible for someone to get a higher ranking than they deserve they could never make the top 10-20 that way. So leave it alone.

Point 2: I think weighting recent games is unnecessary because a player's ability changes very slowly and winning and losing streaks will average themselves out over time (enough games played). Weighting recent games also leads to the wild fluctuations in rank that the WSC was prone to do.

Point 3: I would be in favor of giving a slight advantage to someone who plays more frequently but I do not have any idea of a fair way to do that.

Point 4: I REALLY like the idea of raising the minimum required games number from 3 games in 2 months to something higher per month. 15/month would eliminate 7 of the current top 20 and is not unreasonable to me. Though I am not sure whether this would cause those players to play more or get discouraged. But I do think it is not right to expect others to be as addicted to this game as I am.

[psychicwarrior]

Hi Ed, I am glad "you" decided to post this. Here is my response.

Point 1: I think the 1 out of 3 ratio is high enough. Though it is still possible for someone to get a higher ranking than they deserve they could never make the top 10-20 that way. So leave it alone.

Well.. I think 2 things on this.. certainly 1:2 is good step forward..but why not 1-1.5 andddd broaden the range for top players? That's a must as I noted originally.

1 of 3 still allows chumps to play low rankers 2/3 of the time. If they want to do that , they should not, I believe, qualify comparably for rankings with those players who constantly play opponents of equal or better strength.

Point 2: I think weighting recent games is unnecessary because a player's ability changes very slowly and winning and losing streaks will average themselves out over time (enough games played). Weighting recent games also leads to the wild fluctuations in rank that the WSC was prone to do.

Oh I believe it is Ben. We all improve etc. You can't tell me a new player or even an experienced one, .. that if in a 1000 game example.. his 1st 100 games are as indicative of his strength as his last 100. They clearly are not. His last 100 is far more telling of what kind of player he is. Wouldn't you agree?

WSC is too wild, I concur and I noted that previously. In my mind to make games 1-100 equal weight as games 900-1000 ( in the 1000 game example) is simply inaccurate. It's an easier formula but these Germans are very good at formulas and algorithms :-0 and can do it easily, I am sure.

We see that all the time in sport. "Power Rankings judging who are the best teams judge the last few games far more heavily then games a few mths ago. Recency is very important because it's very indicative. Accuracy is what we want here.

Another example. Say early in the year, Player A beats Player B 8/10 times. Then later in the year. Player B beats Player A 8/10 times. With no recency parameter, you would say they are equal.

But clearly Player B is now better and an accurate rating system must account for that to be credible.

Point 3: I would be in favor of giving a slight advantage to someone who plays more frequently but I do not have any idea of a fair way to do that.

We agree in principle. That's reassuring. Of Course there are easy mathematical ways to weight heavier in proportion to games played . No need to get into specific formulas here. I am a Mathematician by education and will gladly help if called upon.

I hope they realize that player who has played 500 games can't be on equal keel to the guy who plays 50. It is simply a gross distortion of reality if it's kept that way. This is major shortcoming of the current Kleier System.

Point 4: I REALLY like the idea of raising the minimum required games number from 3 games in 2 months to something higher per month. 15/month would eliminate 7 of the current top 20 and is not unreasonable to me. Though I am not sure whether this would cause those players to play more or get discouraged. But I do think it is not right to expect others to be as addicted to this game as I am.

We again agree in principle. It's not a matter of encouragement or not. Rating systems are for accuracy of a players strength. Not an incentive to play more.

Any inaccuracy distorts reality and any favor to one set of players will in turn disfavour another. So all must be rated based on purely what is indicative of a players strength.

Clearly a player who has play 500 games will generally be much stronger than one who plays only 30 assuming of course they are close in the key ratings parameters.

I hope the Admins at Gravon make a solid move on this for the benefit of all and for the higher credibility and accuracy of the site's player ratings.

I think the key elements mentioned here will clearly progress Gravon to an even higher elite standard and the players will quickly realize the brilliance of doing so.

[unbiasbob]

Ed, you're going off on too many tangents for right now. What's important now for this new challenge rankings is to make sure that all players included in the yearly challenge ranking play a minimum of 1 in 3 adequate opponents. Stratego is a genius for devising this system. I am delighted that GB is on the same page. eddy, you have a 1600 game buffer in case you get the urge to focus on playing non adequate players (why do I think this aint gonna happen). High ranked chickens go beyond dozer, way beyond

[psychicwarrior]

Thanks Unbias.. but I am not worried about me.. I am focused on a optimal system which I believe all will want and benefit from.

Its one simple algorithm .. that shouldn't take decades or the Mayan calendar to measure its implementation. Right?

[stratego]

hi ed,

[gentleben]

A few more comments:

1. I think the currrent ratio is high enough. There is no reason to make the better players use a higher ratio. No one can get into the top 10 without playing good players.

2. If I read Stratego's comments correctly it looks like the games are already being rated for recency of play. And Ed, to respond to your question, "You can't tell me a new player or even an experienced one, .. that if in a 1000 game example.. his 1st 100 games are as indicative of his strength as his last 100. They clearly are not. His last 100 is far more telling of what kind of player he is. Wouldn't you agree?" I am not sure you would want what you are asking. Using your numbers as an example, this year your first 100 games had a 64% winning percentage. Your last 100 was at 57%. Your current overall percentage is 60.5%. It looks like you are getting worse the more you play.

3. I agree with Stratego and Hermann. A player who wins enough against the best players deserves a good rank regardless of how many games they play so there is no need to weight a player for the number of games they play. But I still think the more a player plays, the more accurate is his ranking.

4. It is good to see Stratego thinks the minimum number of games is too low. Two games in 3 months is too low. I think this number needs to be higher and I think it would make players play more. I don't care how busy someone is, if they cannot play at least 1 game a week they should not complain if they are not in the rankings. But I think the minimum number of games should at least be raised to 2 games a week. Anyone seriously interested in getting a rank should not be upset with that.

[samuel]

But do you mean 2 games every week or an average GB? Someone could easily play 10 games in a day then not play for 5 weeks if its just an average. But anyone might have a week where they don't play at all, perhaps if they're on holiday or just busy, and then should they drop straight out of the rankings? That wouldn't sit well with the likes of Rave or Skilgannon who play games like they're going out of fashion!

[unbiasbob]

Yeah GB but I think, because I like round numbers, how bout 2.5 games per week or 10 games per month

Oh and Samuel, I think you need to start a new topic...Ace is Back?


LOL. all in good fun. Ace made it back by playing some games vs GB and edie. He suffered some losses but did it kill him? Didn't think so. That was one major rampage over well....nothing

[psychicwarrior]

Well Ben,

My statements were not based on my individual statistics but purely as a measure of strength, recency counts.

I think 4 games/mth is still pathetically low to be included and compared with people who play hundreds. I suggested 15 games a mth which is but a few hours - given the average game is 10-15 min in length. If one wants to be included on the ranking, I dont think that is too much to require.

Stratego's comment about dcAnnies is ridiculous. Clearly he is showing favoritism because he knows him. The fact the he won an outside tournament is completely irrelevant. No one, in my mind should be any exception.

As for simply using % wins regardless of whether you play 1200 or 25 games is also very inaccurate and misleading for the reasons already mentioned. Clearly that concept is too high for Dieter and Hermann to realize. Some of their little buddies would be lower ranked as a result so they wont do that.

Logic isn't a priority here.

[gentleben]

Samuel, I intended 2 per week to be an average. That would mean by the end of June a player would have had to play at least 52 games to be ranked. But I am in favor of a higher number. Bob's 10 per month (60 games by July 1st) or even Ed's 15 per month (90 games by July 1st) are fine by me.

And Bob, you are correct. Ace is back (like we all knew he would be) and there was a lot of fussing over nothing. And isn't it interesting that when forced to play harder players a person's winning percentage (and thus their rating) goes down.

Even Dozer has suffered losses lately though his current rating is still undeserving. He has only won 6 of his last 11. He slipped up and played two good players and lost (treibich and Bart), treibich being his 1st game against a 1500+ player since his loss to jvg. Bart was above 1500 when they played but has found it harder to play here than he thought it would be. Dozer played what he thought was a newbie and though Bart is new here he is not a new player. Dozer got a surprise.

[stratego]

hi ed,

just because i didnt agree with your comments in your pm there is no need
to call my statements ridiculous.

i just faced the situation of dcannies to show that there are several reasons
why a player doesnt play more games.
beeing the currrent wc champ he is the same maniac like most here, but for
privat reasons he cannot be around for the moment.
this has nothing to do to with exceptions. you should read my postings more carefully.

also i didnt like your statement about hermann`s or my buddies. these alligations
just have the effect, that i stop reading your postings. whe need a solution
(i posted one with 2 questions) and not these kind of "kindergarten".
by the way, dozer plays from germany - i do not see any "buddy behaviour".

what we have to talk about:

1. the rules come into effect for the challange ranking only - not the classic
2. how many games must a player play during the year?

i prefer a simple solution like:
25 games per quarter => 100 per year.
this quarantee a stable ranking for a player and within the ranked players
we get a good mix of various opponents.

3. if we change this, all have to notice, that the no. of u.g. will raise and
starting a new ranking in 2011 will show less players for a long time.

4. if a player doesnt meet the requirements:
- a player must play 25 games or he will not be listed
- if he was in list, but plays less then the 25/quarter he stays in list, but his
name is grey.
- of course he must also fit with the current challenge rules.

25/quarter is a high number of games. just because some of us are able to
play several days a week it should not give the impression that all can do it.

any other thoughts?

stratego

[psychicwarrior]

My Comments Below Dieter;



just because i didnt agree with your comments in your pm there is no need
to call my statements ridiculous.

THis wasnt the reason. Your reasoning is based on personal preferences and not an impartial system at rating the strength of players.

i just faced the situation of dcannies to show that there are several reasons
why a player doesnt play more games.
beeing the currrent wc champ he is the same maniac like most here, but for
privat reasons he cannot be around for the moment.
this has nothing to do to with exceptions. you should read my postings more carefully.

It shouldn't matter what his reasoning is. I don't think you want to hear everyone's personal circumstance and what they should be given preference for. That is grossly unfair to the rest of the players here.

I have nothing against him or anyone else You explaining why he he gets special favour and is included in the ranking , to the disfavor of all who play here , is simply partial treatment and personal sentiment, which in my opinion, should have no place in a objective ratings system.

Why don't you just come out and then say its a rating system for you and your friends. Then we will all understand what the real truth is.. and stop wasting our time trying to improve things.

If you do this, then its a subjective rating system which is what it seems to be. Thus the sound logic I proposed was ignored and you get upset when I raise the volume.

I have stated numerous times which you always fail to address and instead tell me how smart and theoretical Hermann is , which is very irrelevant. You tell me to read his papers..again irrelevant. I don't care if he has 4000 books on it. The system as it is now stinks and is very innacurate! for the reasons already state.


His system, as is the ELO, is based on players playing the same approximate number of games. It completely breaks down and falls into the category of "terribly inadequate" in the online environment you have here where some players play 2000 games and some play 50.

His system is simply incompetent in that arena. It's based on % wins only which is very short sighted and not applicable of the environment.

I have stated this many times, but you choose to ignore it !! Why don't you and the Genius Mr. Kleier answer it intelligently. We have asked many times and still get no answer ????!!

and then wonder why me and others here get a bit inflammatory.

Address that please..without telling me how smart Hermann is or that I should read his papers...as that is again irrelevant.

Thank you.




also i didnt like your statement about hermann`s or my buddies. these alligations
just have the effect, that i stop reading your postings. whe need a solution
(i posted one with 2 questions) and not these kind of "kindergarten".
by the way, dozer plays from germany - i do not see any "buddy behaviour".

Well, then clearly you didn't interpret your comments about dcannies the way I and many others did here.

I will state again, no player playing only 21 games should be allowed! I don't care of its Buddha, Boris Becker or anyone else ! That is sheer preferential treatment and a joke of a rating system regardless of reasoning.

what we have to talk about:

1. the rules come into effect for the challange ranking only - not the classic



2. how many games must a player play during the year?

i prefer a simple solution like:
25 games per quarter => 100 per year.
this quarantee a stable ranking for a player and within the ranked players
we get a good mix of various opponents.

3. if we change this, all have to notice, that the no. of u.g. will raise and
starting a new ranking in 2011 will show less players for a long time.

4. if a player doesnt meet the requirements:
- a player must play 25 games or he will not be listed
- if he was in list, but plays less then the 25/quarter he stays in list, but his
name is grey.
- of course he must also fit with the current challenge rules.

25/quarter is a high number of games. just because some of us are able to
play several days a week it should not give the impression that all can do it.

any other thoughts?

stratego

What do you mean by Challenge ranking ?.. ALL I EVER SEE IS the Kleier system for classic.

25 games/qt still seems way low to be included. That means they play but a few hrs/mth. Hardly worthy of being competitive and in fact they have an advantage over players who play often for reasons already mentioned.

How can you rate a player who has played 25 games over a player who has played 250 ?

Please answer that?

How can you even put them into the same comparative standard?

I could come on under an alias..and I guaranteed you I can be in the top 5 in my 1st 25 games.. because no one knows me etc..

Is that what you are trying to encourage? regardless, its what you are doing by allowing so few games.

Trust me, I am thankful for such a fine site and many volunteers as yourself making a brilliant game of it.

You asked for feedback on a rating system and I gave you many fine details all of which were ignored and basically you told me - Hermann is very smart.

I have told you what I know here and I strongly believe if you implement these things, which are not so hard, you will have a masterful , competitive rating system assessing the strength of each player.

Something tells me Dieter, with all due respect, that is not what you are seeking.

[gentleben]

Ed, you are insulting and insults are unproductive, and I agree with you Stratego, it would make me want to ignore him too. Ed, you have not made your case good enough. You say you have been ignored in the past and your system is logical. I have been here 4 years or so and do not remember any posts by you suggesting these kinds of changes. And I am not seeing specific enough logic from you.

On your first point, there is no need to raise the adq/non-adq ratio. This would not make the system better. The most recent changes have fixed the problem with Dozer and raising it higher will not make it any more masterful or competetive. It will only make it harder to find adequate players.

On your second point, the system is already rated for recency of play.

On your fourth point, we are now down to arguing over your number of 180 games a year verses his of 100 games a year. That difference is slight. I like your number better but 80 games will not make the ratings that much more accurate.

The third point is the most questionable and the hardest to implement. A rating system has to be based on percentages, number of wins versus strength of player. And though I agree a rating will be more accurate the more someone plays I do not agree that the ranking of a player who has played fewer games will always change much by playing more. Percentages are inescapable. The Kleier system is not "terribly incompetent."

You ask, "How can you rate a player who has played 25 games over a player who has played 250 ?" I would ask, how could you fairly rank the 250 player over the 25 player if the 25 player had won 70% of their games and the 250 player had only won 60% (assuming of course that they played equally challenging players)?

You want more details from Deiter but I think you need to provide a more detailed suggestion. And quit insulting people when it is not obvious at all the current system is a buddy-buddy system. I see no favoritism being shown here to anyone.

[psychicwarrior]

Ed, you are insulting and insults are unproductive, and I agree with you Stratego, it would make me want to ignore him too. Ed, you have not made your case good enough.


Sorry Ben, I thought you knew how to read. I posted an detailed and thorough explanation of what is needed to make a rankings system accurately reflect the strength of a player.

You say you have been ignored in the past and your system is logical. I have been here 4 years or so and do not remember any posts by you suggesting these kinds of changes. And I am not seeing specific enough logic from you.

On your first point, there is no need to raise the adq/non-adq ratio. This would not make the system better. The most recent changes have fixed the problem with Dozer and raising it higher will not make it any more masterful or competetive. It will only make it harder to find adequate players.

Clearly you didn't read the 1st point. If you did, you wouldn't be saying this.



On your second point, the system is already rated for recency of play.

I contest this fully. If within a year..the 1st games are weighted less than the last, it is so slight to be unnoticeable and certainly not significant enough. They have not posted the formula so I can only go by experience and see that that is certainly the case.


On your fourth point, we are now down to arguing over your number of 180 games a year verses his of 100 games a year. That difference is slight. I like your number better but 80 games will not make the ratings that much more accurate.

That's like saying in baseball, some teams play 162 games and some play 12 and we will just go by % wins.. Totally ridiculous ! The same here.

If your going to rank a player and allow that he be considered a top player he has to prove that more than playing an hour or two a month. That's not serious enough, in my opinion, nor fair to the other players.


The third point is the most questionable and the hardest to implement. A rating system has to be based on percentages, number of wins versus strength of player. And though I agree a rating will be more accurate the more someone plays I do not agree that the ranking of a player who has played fewer games will always change much by playing more. Percentages are inescapable. The Kleier system is not "terribly incompetent."


Difficulty is relevant Ben. It wouldn't be hard for anyone with a bit of mathematical intelligence. Percentages only work when the players play a relatively even # of games. I have stated this in detail . Perhaps too complex for simple minds.. how ever it is stated. I suggest you read it several times and ask someone smarter to explain it to you if you still don't get it.

In my opinion, The Kleier system is terribly incompetent. It doesn't fit the online environment. But since I see you didn't get many other details , I won't elaborate and repeat that discussion here.

I have substantial , credible and logical reasons as stated previously and still yet for this that go unanswered by Dieter except for him to say " Herman is really smart, read his papers!?".




You ask, "How can you rate a player who has played 25 games over a player who has played 250 ?" I would ask, how could you fairly rank the 250 player over the 25 player if the 25 player had won 70% of their games and the 250 player had only won 60% (assuming of course that they played equally challenging players)?


Ridiculous again. I am surprised an experienced player like you would be saying this. To restate my point as clearly it went unread the first time, I can come on as an alias, unknown to all.. and play 25 games and be in the top 5. Surprise and predictability are key in this game as you well know.

The more you play , the more well known and beatable you become. That is an incontrovertible statement that any intelligent rating system should account for.

That's exactly the reason they don't allow aliases in the 1st Place.

In online chess, and I used to be a Master Candidate in chess, its simple. FOr each win/loss you add/subtract so many points..depending on the strength of the opponent. Simple and effective. Too effective and too simple for Hermann KLeier I guess.

Here with the grossly incompetent Kleier system. I can beat NoChance twice, as I did recently, he being rated #1, and my rating/point total did not move !!!

Ridiculous and incompetent! Period .

However this is not an argument to discuss my particulars, but to make the point that the system breaks down when you have players playing a widely disparate number of games. Seems that has to be stated with ferocity and gently stating clearly had no effect.



You want more details from Deiter but I think you need to provide a more detailed suggestion. And quit insulting people when it is not obvious at all the current system is a buddy-buddy system. I see no favoritism being shown here to anyone.[/quote]

I guess you didn't read his defense of dcannies.

Ben, I suggest you go take a reading course please. I offered a detailed suggestion and offered to come up with, or assist on a formula if they asked.

If you are going to argue intelligently to the contrary at least have your facts straight. That would be appreciated and you would more credible instead of sounding like your just trying to defend Dieter without knowing what you are talking about.

Now Let Dieter answer for himself - He is old enough I believe. Thank you.