Czuch: It is calculated at what the BKR's are at the end of the game.

I never liked this since if someone starts to time out on a lot of games, the first people who they start to time out with will get better BKR results... while those who time out in the later games once their opponent BKR is like 500 will get less.

But I understand why it is like this - since in most cases (not tons of time outs - but normal game playing), a persons BKR at the end of the game will more likely reflect their true rating. For example in the Knight Fight game, I started out really good with over 2000+ BKR after my first 4-15 games or so. But now my BKR is more reflective of my true talent in the game (1600 or something) so it makes more since to use the current rating to calculate the BKR change then my inflated 2000+ that I had by getting a lot of good early wins.)

coan.net: Okay, thanks, I can see it both ways now, i guess....


I think it doesnt make sense in many ways because so many games last too long on a turn based site. It just doesnt make any sense in a game that can last over a year to base the new BKR on where we are at the end of the game instead of the beginning, does it?

Czuch:  Theoretically the players' ratings at the end of a match should be more indicative of their "true" ability than their ratings at the beginning, since being later in time they should reflect a larger statistical sample (i.e. more games played).  Remember your own rating will also change during the term of a match.  This is just a feature of a turn based site you can't do anything about.

One thing I have noticed that might be fixed is in setting the BKR range for a tournament.  Eligibility is based on a player's rating at the time he or she joins the tournament.  At the time the tournament starts, his or her rating may be significantly lower than the eligibility threshold.   This  diminishes the utility of being able to set a minimum BKR for entry.  This would not be particularly important if the BKR system for backgammon worked properly, but it does not so this is a major problem.
 

alanback: Well, using the example of a tournament, if I want to play in a tournament with all people who have a similar BKR as myself, maybe because the game has a high luck value, so I make it for only people with a BKR over 2200, then the game goes on for a year and some guy beats me but they now have a 1800 BKR, does that make sense to use their 1800 BKR instead of the one we both started with?


Okay, how about with ponds then? I sign up for a large pond with 200 people and my BKR is 1900. My computer goes down and I miss the first round and come in last place. In the mean time my BKR rises to 2400 by the time this other game finally ends, and then I am killed because now I have lost to everyone else, most of whom have a lower BKR now than mine!

Shouldnt at least in ponds the rating be calculated at the time you actually lose, rather than when the thing finally ends?

Czuch: Why would using the BKR at game start make more sense than the BKR at game end? As for a counter example, take a game with a 30 day move limit. At the start, my opponent has a rating of 2200. 4 weeks later, when he makes his first move, his rating has dropped to 1800. After his first move, the game is finished in a few hours. Wouldn't 1800 make more sense than using 2200 in this particular case?

The problem is that whatever BKR you use, there will always be an example that can be constructed that shows the choice was bad. Ratings will chance over time, and games here take a long time. There's no way around it. The only answer is to not take ratings too seriously.

AbigailII: Wouldn't 1800 make more sense than using 2200 in this particular case?


Not really, since the only way someone would have such a bkr drop is if they had timed out on a bunch of games or something strange like that. I would think it would be more fair to take their "true" ranking, not the one they got from a bunch of timer outs?


Anyway, any comments on doing a pond bkr straight away after a person has actually fallen and therefore technically already lost the game? Why wait for the whole pond to finish, especially like the one I mentioned, where I could have finished another 50 or 100 ponds since I fell in that one and the time it actually finished?

"Splash, you fell in the pond. Old BKR 2315 new BKR 2245."

Instead of... ""your pond is finished 1 year after you fell, your BKR when you splashed was 2315, your BKR now is 2600, your new BKR 1950"!

Czuch:  I think you have overlooked the fact that the purpose of the ratings system is to measure your ability, not to allow you to manage your rating.

alanback: yeah, it doesnt really matter much anyway, except maybe you can tell me how, when I time out in a pond in the first round, and then a year later I have my rating calculated and I go from 2400 to 1900, that really is a measure of my ability?

Obviously if you want to measure my true ability in pond games, then I am a lot closer to 2400 than to 1900....

Dont misunderstand me, I fully realize that I messed up in a big pond, and my rating will reflect that, but why does it have to wait until the whole pond is finished? I should have gotten that 1900 rating a year ago and moved on from it, instead, I spend a year building my BKR in other smaller games, only for waste. It just doesnt make any sense to me

Czuch:  I have to confess I know nothing about ponds or how the BKR system works there.  I know that in backgammon large ratings swings are possible, though 500 points would be extreme.