WCDF Ratings Update?

[champion374]

I hope we come to final understanding soon to get titles or only title will be world champion which will be boring.

Bazkitcase i think the old time system was find.You dont have to play 100 tournaments to get a high rating

[Chexhero]

Hey everyone. I just put in the 3-move National tournament for masters. Majors and minors will be coming very soon. Here is the link: http://www.icheckers.net/ratings. Keep in mind that the tournament was calculated in a game by game format, as that is the only way the system can do it.

[megamau]

1. The compression syndrome.
The scoring by points per game instead of points per round greatly compresses the bandwith of the ratings.
The scoring by points per ballot (or round) in 3 Move Tournaments would result in a vastly improved rating list, and allow the Elo system to operate in the manner for which it was designed. (as in Chess)
Adherence to scoring by points per game in 3 Move Tournaments will, I believe, only prolong the current unsatisfactory situation.
When I pointed this out on a previous occasion, one commentator suggested that compression would be a good thing as it would lessen the difference between the lower and higher rated players !!

Scoring by game or by ballot will not change anything in the rating list. You are simply multiplying the rating differences by a constant, but the information content is the same.
If you have three players at 2200, 2100 and 2000 or if you have the same three players at 1100, 1050 and 1000 or at 1650, 1575 and 1500, what you get is the same.

2. The apples and oranges syndrome.
The practice, when calculating the ratings, of including such a wide range of strengths of tournaments from the strongest master tournaments right down to friendly or fun day events, all in one huge amorphous mass.

Actually, having vastly different strenghts across the data is is not a problem. ELO system handles this very well, take for example the EGD database, which has player from complete beginner to world champions.
The only issue is if "not serious" play is included, meaning the players are not really trying to win or the conditions are different (e.g. extreme blitz games or "just for fun" games in a pub, etc.)

3. Allocation.
One of the main purposes of the ratings is:
To assist in grading players in tournaments when these are classified as Master, Major, Minor etc. In short to make certain that a player cannot enter a class , below or above that which one belongs.

This is not a problem, correct ?

4. The frozen rating syndrome.
Enter one event and if you come away with a good score, do not enter any further events until the 3 year cycle is up.

This is indeed an issue. The ratings for "active" players should probably only include active players in the last year. People can check "all players" if they are interested.

I think you left out the main problem. The basis ratings from which the system was started (i.e. the 2008 values) were calculated with a different system than the current one.
For example Alex ratings was 300 point higher, and I don't think he has gotten significantly weaker. This strongly favors people who play few games, because their ratings is more affected by the huge "starting" bonus.

It is hidden because many players are now out of the "active" list, but if you look at the "inactive" it is even more clear.
I'm sure that Ed Bruch and Levitt Leo were not 200 points stronger than Michele even at their peak.

The correct way to handle the ratings would be to actually find the tournament data from before 2008 and recalculate with the new algorithm, or simply start the system from scratch in 2008 and discard previous ratings.

[MostFamousDane]

Scoring by game or by ballot will not change anything in the rating list. You are simply multiplying the rating differences by a constant, but the information content is the same.
If you have three players at 2200, 2100 and 2000 or if you have the same three players at 1100, 1050 and 1000 or at 1650, 1575 and 1500, what you get is the same.

That is not correct - lets say two players have played two games - each side of a ballot. One game was a draw and the other game was a win. Scoring by game gives 1 point to losing player and 3 point to winner. Scoring by ballot gives winning player 2 points and losing player 0 points - huge difference.

Scoring by ballot is more correct since a winning player gets 2 point whether he wins one or two games which is correct since the only information we have is winning player played better than losing player - the rest (whether one or two games was won) is mostly determined by which ballot is drawn. The only downside is that it is simply to brutal since potentially several of the weakest players might end a tournament on 0 points.

[liam stephens]

megamau wrote:

I'm sure that Ed Bruch and Levitt Leo were not 200 points stronger than Michele even at their peak.

You make my case. (The Compression Syndrome)

The ratings should reflect not only the ranking order of players but also the difference in their relative strength. If player A is 10 times better than player B their rating difference should reflect this. If memory serves me correctly, I believe that Alex Moiseyev has stated that one of his goals was to reach or exceed the rating of 2800+ achieved by Marion Tinsley.
Under the present rating system of points per game this is most unlikely to be achievable. In the critical 3 move ballots the weaker player will often be good enough to draw with the strong side of the opening, the effect of which is to compress the ratings.
[As an aside – I believe it is also the case that in match play the ratings depend only on the overall result of the match, (not on the individual games) and no credit it is given for winning by a large score.]

With regard to Allocation (mentioned above)
No problem with the rule, but in its application. I believe it was often the case that players could opt to enter whichever section they wanted.
In Draughts/Checkers it is often the case that rules are “More honor'd in the breach than the observance”. The most glaring example of this, being the failure to record games and hand in completed score sheets.

[J_D_C_]

Is it possible to have a tournament scored by game, but ratings calculated by ballot?

[Chexhero]

4. The frozen rating syndrome.
Enter one event and if you come away with a good score, do not enter any further events until the 3 year cycle is up.

This is indeed an issue. The ratings for "active" players should probably only include active players in the last year. People can check "all players" if they are interested.

I think you left out the main problem. The basis ratings from which the system was started (i.e. the 2008 values) were calculated with a different system than the current one.
For example Alex ratings was 300 point higher, and I don't think he has gotten significantly weaker. This strongly favors people who play few games, because their ratings is more affected by the huge "starting" bonus.

It is hidden because many players are now out of the "active" list, but if you look at the "inactive" it is even more clear.
I'm sure that Ed Bruch and Levitt Leo were not 200 points stronger than Michele even at their peak.

The correct way to handle the ratings would be to actually find the tournament data from before 2008 and recalculate with the new algorithm, or simply start the system from scratch in 2008 and discard previous ratings.

I do recognize this "frozen rating syndrome" as a problem in our current system. This was obviously not thought through from when this rating system was first made. There is a way we can remedy this problem without having to completely wipe out 7 to 8 years of rating data. For those players that are inactive, we can put them in provisional status when they return to tournament play. In this provisional status, they will not be able to affect "active players" ratings for X amount of games. Only their rating can be affected. This is arguably fair because who knows how much the inactive player has improved or declined over a 3 or more year period of tournament inactivity. Furthermore, we can calculate their individual performance using a different rating formula. This formula will change their rating more rapidly and allow it to get to where it should be more quickly. I do not know what the formula would be at this point, but I have people, including math experts, helping me figure something out. I also plan to use this method for new provisional players who start at a 1600 rating. Under the current system, they are giving a 20 game provisional period where their performance does not affect their opponents ratings, rather, only their rating changes. This is all good except it can take more than a 20 game provisional period in order for a player to reach their true rating. Currently, we have players at about 1800 to 1900 in the ratings who just came out of their provisional period of 20 games. Their true rating however, should be somewhere around 2000 to 2100. Therefore, it obviously becomes unfair when a higher rated player has to play them. However, if we can apply a formula that can get them to their true rating more quickly, then we can overcome this problem. We cannot do this under the current rating system, but we are in the midst of making a new one that will be more flexible.

Scoring by ballot is more correct since a winning player gets 2 point whether he wins one or two games which is correct since the only information we have is winning player played better than losing player - the rest (whether one or two games was won) is mostly determined by which ballot is drawn. The only downside is that it is simply to brutal since potentially several of the weakest players might end a tournament on 0 points.

How do you know that whether one or two games are won by the stronger player is mainly a result of which ballot is drawn? Is there scientific data that proves this to be the case? I can tell you that I have lost quite a few games from the strong side of a critical opening against stronger players. On a lot of the more critical openings, if you don't play the strongest moves, a good player on the weak side can easily turn it around into a even position, or even an advantage for themselves. Furthermore, many of the openings in the 3 move deck are fairly comparable in difficulty. I would say about 20 to 25 of the openings are "very critical," but at least that many or more are equal in strength. I also find many of the 3 move openings in the deck to be pretty comparable in overall strength. I think there is a good argument to be made that game scoring is actually the superior method since it actually considers every game.

The ratings should reflect not only the ranking order of players but also the difference in their relative strength. If player A is 10 times better than player B their rating difference should reflect this.

I would argue our current rating system does reflect the relative strength differences between players, despite the ratings being more compressed overall. Player A and B who were separated by 50 points in the older system are now separated by 30 points in the new system. I don't know how accurate that actually is, but the point is that the 30 point difference is really no different from the 50 point difference. It is just a more compressed system of measurement. For example, like inches versus centimeters. The difference between A and B may be 2.54 centimeters, but only 1 inch overall. I can understand though why many would prefer the older rating system that had less compressed ratings. The argument for the newer rating system is it allows for a more competitive system overall since players ratings are closer together. Higher level players may be less likely to like it, but it can better keep the spirits up for lower rated players.

Is it possible to have a tournament scored by game, but ratings calculated by ballot?

I would first ask if you are going to calculate the tournament by ballot, then why not just score it that way? It definitely is possible, but arguably not the most plausible. For example, say player A and player B have the same rating before the tournament. They both played the same opponents since the tournament was round robin. Now say player A wins the tournament by having a greater difference of wins as compared to loses. However, player B who had a lesser difference of wins to loses still tied player A in the amount of rounds won and the amount of rounds lost. Player A still did better overall, but will now not get credit for it in the ratings. That is just one easy example of why ballot scoring fails. Of course, like others you can argue how many wins or loses in a round mainly depends on the opening drawn, but I already presented arguments against that case above.

[megamau]

Scoring by game or by ballot will not change anything in the rating list. You are simply multiplying the rating differences by a constant, but the information content is the same.
If you have three players at 2200, 2100 and 2000 or if you have the same three players at 1100, 1050 and 1000 or at 1650, 1575 and 1500, what you get is the same.

That is not correct - lets say two players have played two games - each side of a ballot. One game was a draw and the other game was a win. Scoring by game gives 1 point to losing player and 3 point to winner. Scoring by ballot gives winning player 2 points and losing player 0 points - huge difference. Scoring by ballot is more correct since a winning player gets 2 point whether he wins one or two games which is correct since the only information we have is winning player played better than losing player - the rest (whether one or two games was won) is mostly determined by which ballot is drawn. The only downside is that it is simply to brutal since potentially several of the weakest players might end a tournament on 0 points.

For scoring purposes in tournaments, you can use ballots scoring. But for ELO ratings, you are throwing away information by considering 3-1 and 4-0 the same result.
I will link an excel file that explains it.

Using game scoring only compresses the elo range, but it does not change the ordering or the relative distances. If I am 200 elo point weaker than Michele and 100 elo points stronger than a beginner in ballot scoring, I will be 100 elo point weaker than Michele and 50 elo points stronger than the same beginner in game scoring, but nothing has changed.
What counts in Elo is only the "relative difference", nothing else. Italian checkers Elo use an higher K, and Michele has 5000+ points while I have 2900+; does this make any difference as long as the relative distances are the same ?

In summary: it makes more sense to use game scoring for Elo, because it preserves more information and because can be applied also to tournament without ballots.

[MostFamousDane]

Scoring by game or by ballot will not change anything in the rating list. You are simply multiplying the rating differences by a constant, but the information content is the same.
If you have three players at 2200, 2100 and 2000 or if you have the same three players at 1100, 1050 and 1000 or at 1650, 1575 and 1500, what you get is the same.

That is not correct - lets say two players have played two games - each side of a ballot. One game was a draw and the other game was a win. Scoring by game gives 1 point to losing player and 3 point to winner. Scoring by ballot gives winning player 2 points and losing player 0 points - huge difference. Scoring by ballot is more correct since a winning player gets 2 point whether he wins one or two games which is correct since the only information we have is winning player played better than losing player - the rest (whether one or two games was won) is mostly determined by which ballot is drawn. The only downside is that it is simply to brutal since potentially several of the weakest players might end a tournament on 0 points.

For scoring purposes in tournaments, you can use ballots scoring. But for ELO ratings, you are throwing away information by considering 3-1 and 4-0 the same result.

My argument is that you are adding something that is partly random and therefore introduce noise - see my previous reply.

[champion374]

So i guess we will get titles on the next ratings list then Alex?

[Alex_Moiseyev]

So i guess we will get titles on the next ratings list then Alex?

You mean "ACF titles" ?

[Alex_Moiseyev]

People constantly make the same mistake - there is no game in 3-moves or 11-man ballot events. There is ballot - minimatch.

If I play World title match, 40 games, it doesn't matter I won 21 games, or 1 game and 39 draws. In both cases I won match.

We can start using rating for any serious purposes - titles/pairing/endorsement etc, only when we have it accurate 100%, not 79%.

AM

[champion374]

I ment Wcdf

[Alex_Moiseyev]

I ment Wcdf

WCDF already have titles. I think last GM title was given to Matteo Bernini (Italy) in fall 2015. Not sure about IM.

[champion374]

I asked because i did not see them on that list.

Im far from a title but with hard work will get there soon