Unofficial spring '42 league

[The Dessert Fox]

Each person starts with a 100 rating

When they win they get their opponents rating - 50 + 100n (N=1 for a win, .5 for a draw and 0 for a lose) then we divide this by 10 and add it to their rating

i.e.

My rating = 100
Zhuk (My opponent) rating = 125
I win

My rating change

125-50 = 75 + (100*1) = 175
175/10 = 17.5
100+17.5=117.5

Zhuk’s rating change

100-50 = 50 + (100*0) = 50
50/10 = 5
125-5=120

If there is a gap greater than 40 points it is just 40 points higher or lower than your rating

I.E.
Me=100
Opp=150
I lose

My rating
140-50=90+0=90
90/10=9
100-9=91

Opp rating
110-50=60+100=160
160/10=16
150+16=166

This is to stop lower rated players losing lots of points when play higher rated players

[Granada]

Desertfox, many thanks for the effort. Is it the system used in the other ladders? Or where does the system come from? By doing some basic calculations, it leads me to following conclusions:

1. The average rating of all players is increasing, which might lead to the conclusion that the handicap of players joining the ladder later will increase.

2. The player with lower rating loses the more points the higher is the ranking of the player he plays.

I know there is the 40 points gap but still it seems to me to go a bit against common sense.  When I sometimes play on the Free Internet Chess Server, the system works the other way round. Your ranking would always change within the margin of 16 points. And the sum is always 0. You lose exactly the same number points your oponent gains. The weaker the player you play the more points you can lose, the less you can take and vice versa.

3. The proposed system is a bit difficult to carry out manually.

As things stand at this point I am unfortunately still the only player to take part in all four league games. So I tried to apply your system on the four games I have played.

1. game was Packersplus vs me
2. game MikeWatroba vs me
3. game Packersplus vs. me
4. game Packersplus vs me

By chance, I won all four games. It gave me really tough time to count all the changes in the Desert Fox rating.

But I belive they went like this. 1. Granada 115. Packersplus 95; 2. Granada 130. MikeWatroba 93,5; 3. Granada 144,5, Packersplus 86,5; 4. Granada 158,15, Packersplus 78,85.

I am sparing you the counting. And I do admit I might got something wrong. So soon my oponent might be deviding 108,15 by 2 getting really weird numbers at the end.

This really leads me to a question whether a less complicated system of counting without a need to use decimal numbers would not serve us better.

–-

And hence coming back to the inspiration of the FICS system based on the rating difference between players, let me propose an alternative system:

1. difference less then 40. Winner 8 / Loser 8, Draw 0.
2. difference 41-80. A. Higher ranked player wins: Winner +7 / Loser -7 B. Lower ranked player wins: Winner 9 / Loser -9.
3. difference 81-120. A. Winner 6 / Loser -6; B. Winner 10 / Loser -10.
4. difference 121-160. A. Winner 5 / Loser -5; B. Winner 11 / Loser -11.
5. difference 161-200. A. Winner 4 / Loser -4; B. Winner 12 / Loser -12.
6. difference 201-240. A. Winner 3 / Loser -3; B. Winner 13 / Loser -13.
7. difference 241-300. A. Winner 2 / Loser -2; B. Winner 14 / Loser -14.
8. difference 301 and more. A. Winner 1 / Loser -1; B. Winner 15 / Loser -15.

Or even better the rating zones should increase progressively, as follows, which reflects actually the real ratings change in stake:

1. difference less then 40. Winner 8 / Loser 8.
2. difference 41-100. A. Higher ranked player wins: Winner +7 / Loser -7 B. Lower ranked player wins: Winner 9 / Loser -9.
3. difference 101-180. A. Winner 6 / Loser -6; B. Winner 10 / Loser -10.  
4. difference 181-280. A. Winner 5 / Loser -5; B. Winner 11 / Loser -11.
5. difference 281-400. A. Winner 4 / Loser -4; B. Winner 12 / Loser -12.
6. difference 401-540. A. Winner 3 / Loser -3; B. Winner 13 / Loser -13.
7. difference 541-700. A. Winner 2 / Loser -2; B. Winner 14 / Loser -14.
8. difference 701 and more. A. Winner 1 / Loser -1; B. Winner 15 / Loser -15.

Unless there is a specific condition for a draw set in the rules, there is no point in deciding what should be the ratings change when a draw occurs assuming there would be none.

With this system it would be much easier to count manually. Coming back to the four games with everybody starting at 1000 points, it would be: 1. Granada 1008, Packersplus 992; 2. Granada 1016 MikeWatroba 992; 3. Granada 1024 Packersplus 984. 4. Granada 1032 Packersplus 976.

So next time I would play Packers in case he wins, he would get 9 points and I would lose 9, while my win would bring me only 7 points and his loss would be 7 as well. If I happen to win my rating would get to 1041 so then anybody entering the ladder would play me for 9 points while having at stake only 7, oddly enough even Hobbes (unless we all agree that he, Zhuk and some other heavyweights should enter the ladder at 1500  :-D).

This system is much easier to count, isnt it? It also keeps the avearge rating of all the players at 1000. The only problem is we really would need a database to keep a track of all the changes. But a way to bypass this would be to start a thread either here or at the TripleA Warclub where you would actually have to update the ladder standings after each and every game. This of course is possible, though not very practical.

What do you think?

[KaLeu]

I hope you don’t mind me interfering - it’s really none of my business since I don’t intend to play, but I do have some experience with chess rating systems.

I see a few problems with the system proposed:

(a) The explanation and examples suggest that it’s actually better to lose against a lower rated opponent than against a higher rated one. For instance, if your rating in the first example would have been 110, then we’d see (changes in bold):

My rating = 110
Zhuk (My opponent) rating = 125
I win

My rating change

125-50 = 75 + (100*1) = 175
175/10 = 17.5
110+17.5=127.5

Zhuk’s rating change

110-50 = 60 + (100*0) = 60
60/10 = 6
125-6=119

So Zhuk lost an extra point because he played a stronger opponent than in the original example. This effect becomes even stranger at lower ratings, because from what I understand of the calculation, it’s actually possible to gain points by losing to someone rated at 40.

(b) The method adds points to the rating pool as a whole. In the above example, you gained more points than Zhuk lost. When that happens, all ratings tend to drift upwards over time - it’s happened with international chess ratings as well. The effect seems to be rather strong here, and that implies that players who play a lot tend to get overrated as compared to those who play fewer games. Some say that it’s good to “reward” active players by doing that, but it’s contrary to the principle of rating lists.

The basic thought behind a rating system is that it’s not a competition, but a statistical tool to compare relative playing strengths. That implies that the system should be able to compute an expected average outcome for a game between two players on the list. In other words: if everyone keeps playing at the same level as when they started, that expected outcome will be met, and the ratings will hardly change at all. Conversely, a player who improves and achieves better results relative to the others in the rating pool, will move up.

The FIDE rating system used in chess works like that, but it’s a bit of a pain to do the calculations. If you’re interested, I can suggest a manageable alternative that works in approximately the same way but is much easier to understand and calculate than the FIDE ratings. It would be a simplified version of a system proposed by Jeff Sonas, an expert in the field of chess rating.

[KaLeu]

Ah, Granada made the same points but didn’t need all my wordiness.  :-)

Anyway, like I said, I have an idea on how to do this if you’re interested.

[The Dessert Fox]

actually you can only gain points if you have rating 80 or lower from losing to a 40 rated player

[newpaintbrush]

1.  The point changes count be calculated by computer instead of manually.

2.  I think each “match” should consist of two games, with a player playing BOTH Axis and Allies.  A win and a loss would count as a draw; two wins a win, two losses a loss.  This is to eliminate any bias from Axis or Allies being more powerful, and would help determine if there is a bias.  (If Axis win 75% of games in a large sample of first out of two games played, Axis might have an advantage) & so on.

[Granada]

@Bunnies:

1.  The point changes count be calculated by computer instead of manually.

In the perfect world, surely. The best would be of course if it is directly included in the TripleA software. But before we get there I don´t think there is a way to get this done in any different way, then just count it and publish a new complete ladder after every single game.

@Bunnies:

2.  I think each “match” should consist of two games, with a player playing BOTH Axis and Allies.  A win and a loss would count as a draw; two wins a win, two losses a loss.  This is to eliminate any bias from Axis or Allies being more powerful, and would help determine if there is a bias.  (If Axis win 75% of games in a large sample of first out of two games played, Axis might have an advantage) & so on.

Bunny, please, dont make it more complex then it is necessary. People can always agree to play a revange game with altered sides.  That is what we did with Packers. Moreover, in the rules suggested by Zhuk, there is the bid option. I would say Allies and Axis is a bit like black and white pieces in the chess.

But what do you people think of the 16 points a game counting system I suggested based on the experience with the FICS (Free Internet Chess Server)? Would not this be the easiest way for us to go before there is an official ladder?

[newpaintbrush]

Regardless, the game winner should report whether Axis or Allies were used by the winner, if a bid, to whom and for how much, and whether Low Luck or dice - that, and date.  Tracking these statistics by player matchup is the only thing that’s going to support claims that Axis or Allies have an advantage, and we may as well keep good records considering it isn’t much more trouble.

@Granada:  Re:  “publishing a new complete ladder” - what’s the rush, Gran?  Sounds like you really wanna get moving on this!  Also, what was that about a “perfect world”?  I think you meant to reprimand me for something, but I’m afraid I missed your meaning.

As far as FICS or whatnot - I’m TOTALLY AGAINST IT.  Weak players will be very popular as stronger players try to boost their ratings, and stronger players will have a harder time finding opponents.  Any rating system that always rewards a winner with points risks this happening.

'Far as I know, XBox did some serious statistical ratings research - they put out a paper that I read a month or so ago for fun, can’t remember too much of the specifics.  IMO FIDE stuff is in the right direction, or at least a system that doesn’t ALWAYS reward the winner with points.  But I will say that I heard there are better system than FIDE’s (wouldn’t know personally, as I don’t track rating systems, so maybe someone more familiar with that sort of thing can comment), and I have heard that chess players game FIDE’s system.  (Although with chess politics the way they are, I’m not sure if I’d blame the rating system or politics or both.)

[KaLeu]

Improvements to the chess rating system are indeed being discussed. I’m going to live dangerously and post a link, hoping that it won’t be considered spamming: http://www.chessbase.com/newsdetail.asp?newsid=6687

It’s totally possible to set up a system that will avoid undue boosting of ratings by playing against weak opponents.

[Granada]

@Bunnies:

Regardless, the game winner should report whether Axis or Allies were used by the winner, if a bid, to whom and for how much, and whether Low Luck or dice - that, and date.  Tracking these statistics by player matchup is the only thing that’s going to support claims that Axis or Allies have an advantage, and we may as well keep good records considering it isn’t much more trouble.

I do agree on this.

@Bunnies:

@Granada:  Re:  “publishing a new complete ladder” - what’s the rush, Gran?   Sounds like you really wanna get moving on this!  Also, what was that about a “perfect world”?  I think you meant to reprimand me for something, but I’m afraid I missed your meaning.

What I meant was, that unless you have computer doing it for you, you have to do it yourself. So lets say we would start a thread here called Bunny and Granny Unofficial Ladder. As the start you would publish the starting ladder after lets say first 50 league games are included and counted. After then, when any two players would play a game, they would count the points change themselves and include it into the the new ladder.

So lets say you would be 3rd with 1250 points, I would be 14th with 1050 points, you would win so you would get 6 points plus and I would get 6 points less. We would post the outcome of the game, who played what, whether ther was a bid, LL etc… and we would include it into the new edition of ladder: Bunny ovecomes Zhuk with 1256 and moves to 2nd , while Granada gets down with 1044 to the 16th. Does it make sense?

As for the other thing I did not want to “reprimand” you. I just say that it needlessly complicates things to squeeze two games into one result.

@Bunnies:

As far as FICS or whatnot - I’m TOTALLY AGAINST IT.  Weak players will be very popular as stronger players try to boost their ratings, and stronger players will have a harder time finding opponents.  Any rating system that always rewards a winner with points risks this happening.

This is correct. It can be resolved easily by including a ninth category, that in fact is included in the FICS counting – when the difference between the players is so large that if you win as the stronger player you get 0 points, while when you lose, you lose 16; with your opponent’s rating changing respectively.

So after this change the proposed system would look like this:

1. difference less then 40. Winner 8 / Loser 8.
2. difference 41-100. A. Higher ranked player wins: Winner +7 / Loser -7 B. Lower ranked player wins: Winner 9 / Loser -9.
3. difference 101-180. A. Winner 6 / Loser -6; B. Winner 10 / Loser -10.  
4. difference 181-280. A. Winner 5 / Loser -5; B. Winner 11 / Loser -11.
5. difference 281-400. A. Winner 4 / Loser -4; B. Winner 12 / Loser -12.
6. difference 401-540. A. Winner 3 / Loser -3; B. Winner 13 / Loser -13.
7. difference 541-700. A. Winner 2 / Loser -2; B. Winner 14 / Loser -14.
8. difference 701-880 A. Winner 1 / Loser -1; B. Winner 15 / Loser -15.
9. difference 881 and more. A. Winner 0 / Loser 0 B. Winner 16 / Loser - 16.

Where the option A. describes the situation when the player with higher rating wins, while option B. describes the situation when the player with lower rating wins.

I think it goes in the direction of the Elo-like chess rating system, but experts like Herr KaLeun must know much much more about this then I do. Would the system I propose make a sense to you?

[The Dessert Fox]

I think we would then need to make sure that players couldn’t challenge so high above them that there is no point in the othe person playing. Because if a 100 challenged a 1000 then there would be no point for the 1000 playing because they can’t get any points. I believe that if we use the system suggested by Granda we should make a 500 challenge limit rule. I.E. you can only challenge 500 points above you but you can challenge as far down as you like

[Zhukov44]

This is all very interesting…I’m not terribly intuitive when it comes to math so I’ll have to try come back with a pot of coffee and digest all this.

To people who use or might use the Spring 42 league…would you prefer if you just use a numerical ranking system for competition instead of W/L records?

Naturally the nice thing about an automated ladder is it does the math for us.  But if it’s not TOO much maintenance then maybe we can do the math for ourselves until somebody comes along and programs an automated ladder system for us.

This also reminds me of a post Bung made recently in the TripleA War Club where he was interested in created a global A&A rankings system that went beyond just TripleA, but would incorporate the data from all the clubs.

[Granada]

@Zhukov44:

This is all very interesting…I’m not terribly intuitive when it comes to math so I’ll have to try come back with a pot of coffee and digest all this.

To people who use or might use the Spring 42 league…would you prefer if you just use a numerical ranking system for competition instead of W/L records?

Naturally the nice thing about an automated ladder is it does the math for us.  But if it’s not TOO much maintenance then maybe we can do the math for ourselves until somebody comes along and programs an automated ladder system for us.

This also reminds me of a post Bung made recently in the TripleA War Club where he was interested in created a global A&A rankings system that went beyond just TripleA, but would incorporate the data from all the clubs.

I am definitely willing to prepare the illustration of how the rating might work with the manual counting after lets say first 20-30 games of the league are played and at least ten players are involved.

[DarthMaximus]

Stickied thread for you guys.

[Zhukov44]

Thanks for the sticky.

The link to the league forums on the TripleA War Club is here.  Come get your war on!

[Zhukov44]

Thanks for the sticky.

The link to the league forums on the TripleA War Club is here.  Come get your war on!

http://www.tripleawarclub.org/modules/newbb/viewforum.php?forum=16

[Panther]

Outdated so unstickied today.