pgt: ~~ I am definitely setting up 2200 as minimum. But there is always one (and only one) player who manages to accept the game with less than 2100 rating! It must be that the site isn't working. Since it first happened I have been VERY careful to check, and it has happened twice more.

Yes, I thought it was too much to hope that it would be as simple as "not plugged in". Thank God there are Reshers in the world, eh? ;o)

~~ As for points - it always seems to me that in a simple game, you get the same points whether you just win, or win with a gammon, or win with a backgammon. That's rubbish.

That's how it works with proper backgammon, ie. the version of the Elo formula that FIBS, DailyGammon and various other sites use. It's getting to the match length and the win that matters./ How you do it doesn't count. Same as in tennis or other sports. You may get a pat on the back for winning by a country mile but you don't get any more of a win.

~~ As for "match", in super-rugby, the losing team can still get bonus points for scoring 4 tries, or for losing by less than 7 points. It seems to me that if you lose an 11 point match 11-0, then you should not be rewarded, but if you lost 11-10, you should still lose, but perhaps lose 2 points instead of 8. (But it's only a game)

Ah, that's interesting. Sounds like we need super-backgammon! :-D

I had been going to say that rating points aren't in any way a reward, even if it's nice to get them, but that they contribute to a rating which is intended to be predictive - that is two ratings will give you each player's winning chances in a match between them.

That's still true (for the proper formula but not for the BrainKing version) but I was then going to say that it's match wins that go into the formula and give you those percentages and the games don't matter. However, I had second thoughts. Why, indeed, shouldn't 11-10 count as less of a win than 11-0? After all, it's reasonable to at least suspect that the player who scored 10 knows a bit more about the game than the one who scored nothing.

I think the answer is that the backgammon formula, based on matches, is nice and concise as it is but could become rather complicated if it tried to take into account the games as well. I'm no mathematician so I don't know what would be required but it would certainly be more complicated with both scores involved rather than just the binary won/lost. It's not even certain that it would be more predictive. Mind you, once you had such a game-based formula (and a large number of matches), it would be straightforward to test.

It's certainly an interesting idea.

.

Getting back to that win/gammon/backgammon. It really depends on whether you're playing a money game (where the final score actually matters because it multiplies the stake; so a backgammon win in a $10 game would therefore be worth $30) or whether you're playing a 1-point match, in which case the extra 2 points are superfluous.

It matters because if you win a gammon or backgammon in a 1-pointer then you've possibly played badly. Gammons and backgammons are usually achieved through taking risks that reduce your winning chances. As such they get flagged as errors by analysis. If you only need 1 point then you should play to win 1 point and not more. (If it happens by chance that's fine but no extra risks should be taken). I therefore wouldn't give credit for exceeding the match length, whether it's a 1-pointer or a 25-pointer. But a money game is arguably worth more. That rating formula that takes games into account might also be able to treat money games differently rather than as 1-pointers.