Are people satisfied with the current rating system?

[Alex_Moiseyev]

Would a GM really beat a master 9 out of 10? Would it be more likely 3 wins and 7 draws

Tell me who is Master and who is Grandmaster, and I give you an answer

[Richard Pask]

Some years ago, Bill Robertie assigned levels of complexity to different games (I believe checkers was .

In order to be at level 3, for example, you needed to consistently score 75% against a level 2 player; to be at level 4, to consistently score 75% against a level 3 player etc...

I used a similar principle in Logical Checkers, Book 1, to conclude that there are 10 distinct levels in the game (actually Chinook and Tinsley registered was might be called 10+).

Level 1: Basic novice (90% of the population); Level 2: Advanced Novice (9% of the population); Levels 3-6: expert; Levels 7 and 8: master; levels 9 and 10: grandmaster.

The main point of contention in Robertie's thesis was, what constitutes the smallest 'unit of currency' in a given game? (Chess, checkers, backgammon ...) In 3-move checkers, the smallest possible unit is a 2-game heat. This would not render the 10 levels I describe. In my view, 4-game heats probably do this, 6-game heats almost certainly do and 8-game heats definitely do. (This is essentially just echoing the lively comments Willie Ryan made about the 1950 Us Nat Ty in Big League Checkers where he argued strongly for 8-game heats over 4-game heats.) To give a concrete example: I'm confident that a super-grandmaster like Alex (middle of level 10) would register 75% against a run-of-the-mill one like myself at my peak (middle of level 9) assuming 8-game heats were played. That is, he would likely win 2 heats and draw 2 heats. (In fact, he would probably do better than this, perhaps winning all 4 heats, and might only require 4 or 6-game heats).

Whether or not it is reasonable to define one unit to be an 8-game heat is a moot point: I think it is. However, given that ACF heats consist of 4 games, some might disagree.

Now, by extending one unit to represent a 16-game heat or even a 32-game heat, I believe the level of complexity could artificially be extended to 11 and 12 levels respectively. However, I think this is stretching a point and is not a reasonable model.

[chipschap]

There are two different things here, and it's important to keep them distinct. If we talk about levels (which is really granularity of resolution), then if n is the number of games, we can resolve 2n+1 levels. So 4 games resolves 9 levels and 6 games resolves 13.

However, such a minimal number of games doesn't provide particularly accurate, stable, or reliable results. As more and more games are played, granularity increases still further--- which is probably not terribly important--- but reliability and accuracy definitely go up significantly.

USCF chess has 13 levels (classes J-A, expert, national master, senior master, but the classes below 'D' often get lumped) and doesn't consider ratings more than provisional until 20 games have been played.

[Richard Pask]

It's possible that I am getting crossed wires here -possibly not. My main point is that whereas in chess 'a game' literally refers to one game played at classical time controls, in my view in checkers 'a game' - the smallest possible unit - needs to refer to a 4,6 or 8-game heat. Certainly, in order to receive any rating at all, a player would need to play a decent number of 'games': in chess this is perhaps 20 single games; in checkers it might also be 20, but 20 4-game heats.

On the basis of 4-game heats, we may be safe in asserting there are 10 distinct levels.
On the basis of 6-game heats there are almost certainly 10 distinct levels.
On the basis of 8-game heats, there are definitely 10 distinct levels.

Sorry if I'm merely repeating myself!

[nboatman]

Comparing chess ratings with checker ratings, I do wonder if expected score is influenced by the high percentage of draws (at least at upper levels). I know how the math works, and the numbers given above are what the formulas tell us. Would a GM really beat a master 9 out of 10? Would it be more likely 3 wins and 7 draws (made up numbers just to illustrate the point) for a score of 13, not 18?

The higher percentage of draws is a big factor, and we need to set up our rating formula to account for this. Based on the current rating formula and our desired range for GM, master, and expert, the formula would expect the GM to average 1 win and 1 draw per 2-game round against a master (the 9 out of 10 was GM vs. expert). You're very much asking the right question about what kind of score we should really expect.

For a GM versus a master (at similar levels within the GM and master ranks - e.g. not a top GM vs a lesser master), my guess would be that, on average, the GM would get a little less than 2.5 points out of every 4 possible. This would work out to something between 12 and 13 points per 10 games. Bob, you said you were just throwing out some numbers to illustrate, but is something like 13 points over 10 games what you'd expect for a GM versus a master?

Does this seem reasonable to everyone? Everyone please speak up.

In addition to the GM vs. master question (which we're using as a proxy for 200 rating point differences), there are other questions to answer. Here's one that I'm particularly interested in:

What should we expect when a GM plays an expert (I'm using this question to proxy a 400 rating point difference, based on the ranges Clayton outlined)? Certainly the expert shouldn't do as well as the master player, but I don't think he would do terribly. Draws are pretty common, and I would expect the expert to manage some draws... I think he might manage about 4 or 5 draws out of 10 games. Overall, my guess is that the GM would average somewhere between 15 and 16 points per 10 games. What does everyone else think here?

Tell me who is Master and who is Grandmaster, and I give you an answer

Unfortunately, we need the answer to help us figure out whom we'll call a grandmaster and whom we'll call a master, since ratings will figure into these titles. So, you need to think about what grandmaster and master level should represent. If there are players that would think of a typical grandmasters and typical masters (no need to tell us whom you're imagining), how would you expect these players to do against each other over 10 or 20 games?

[Alex_Moiseyev]

Titles shall not not rely entirely on rating only. It's combination of performance and rating and actually performance part is most important.

To earn ACF Grandmaster title, you have to get two grandmaster norms, at least one in ACF event, at least one in ACF National where you have to finish among top 3. Also in this event you should have at least 2 grandmasters and final score of of player is somewhat around 75% points.

… plus rating

I really hate if someday Grandmaster title will depend on rating only. This will lead to title inflation and checkers de-gradation.

For me personally rating is just a nice childish toy and shall not be taken seriously. It is OK if it exists, and it's not a tragedy if it's not there. if people like it and enjoy - why not ? It's a good promotional and motivation Instrument.

AM

[nboatman]

But what is a grandmaster "norm"? As far as I can tell, the by-laws don't specify. However, doesn't it involve performing sufficiently well against a field that is sufficiently strong? I'm not sure how "sufficiently well" is defined, but isn't the rating of one's opponents used to determine whether the field is sufficiently strong?

[Alex_Moiseyev]

But what is a grandmaster "norm"? As far as I can tell, the by-laws don't specify.

Hi, Nick!

ACF BY-LAWS have two important parts - permanent Constitutional part which require approval of full Board. And this is one which you can see on ACF website now. Second part have detail definition of norms, rating, dates etc ... and this part require approval of only ACF Executive Committee and can be (and suppose to) change more frequently.

We will post this part of laws on ACF website soon along with lists of ACF masters, grandmaster and experts.

but isn't the rating of one's opponents used to determine whether the field is sufficiently strong?

Not really. If you are good enough to finish among top 3 in ACF National Master Division, you well deserve to achieve GM norm, if this event had at least 2 GM's, you played with both of them at least 14 games and scored at least 75% points … something like this.

So … my advice to you - don't be too crazy with rating (it's just a number) and concentrate on study and practice

AM

[nboatman]

I’m glad to hear that there are additional details about title norms that are in writing somewhere. The information currently on the website is too vague.

I think this discussion about ratings has gotten a bit off track. I think I may have inadvertently given the impression that titles should be awarded on the basis of rating alone. This was never my intention. I did suspect that the norms might be defined in terms of opponents’ ratings, but it sounds like the board may have devised norms that don’t reference ratings at all. I don’t think this should stop us from trying to make the ratings more meaningful.

don't be too crazy with rating (it's just a number)

Rating is literally a number, but it’s a number that can convey meaningful information about playing strength. Plus, ratings are a fun feature, not to mention that newcomers to our game will tend to evaluate us in relation to chess and will expect there to be meaningful ratings.

So, to summarize what I’m looking for... I’d like to get some consensus about the outcomes people would expect from long matches between various levels of players. I chose GM, master, and expert because I think people have some sort of rating-independent idea what those mean. The idea here is just to get reasonable outcome expectations (that take into account the high frequency of draws in checkers) that can help give shape to our ratings.

[Alex_Moiseyev]

Nick, actually rating is part of formula for titles. It will be true to say that titles are based on PERFORMANCE + RATING. After you achieve 2 GM norms, you still should have 2,200+ rating to be awarded by GM title.

Same thing for Masters => 2 norms based on performance and rating. For Experts title I believe is based on rating only. Last time me and Joe Moore spoke in details about this several months ago. We even created list of holders, but it needs to be approve by Execs

AM

[champion374]

Clint Also spoke about having different ratings ex,Gayp 3 move 11 man.What you think Alex?

[Alex_Moiseyev]

Clint Also spoke about having different ratings ex,Gayp 3 move 11 man.What you think Alex?

I totally agree with Clint. With our GM criteria we have such outstanding GAYP player as Jack Francis who achieved both ACF GM norms. However in 3-moves restrictions Jack is maintaining level of strong master, not GM. There are some other examples.

Also every win in GAYP against anyone - GM, master or experts is something very special and doesn't happen very often.

Regarding 11 man ballot - very small number of events. Not sufficient statistics.

So ... I agree with Clint.

AM

[champion374]

i await the 3 move ratings.