What have you done with my achievements and triumphs? I thank in advance any good Samaritan who can and wants to help me

[lkaufman]

While I was expecting that something would be done to discourage playing for draws, the solution implemented was rather drastic. I think many people felt that the Leaderboard looked ridiculous when someone with almost no wins at all was on top, and they do have a point. It has nothing to do with you personally or your lack of a FIDE title or rating, it's just that it's known to be relatively easy to make early repetition draws since the engine doesn't mind. The argument on your side is that if two players have for example one win and nine losses each, but one player has no draws while the other has ten, then the player with the ten draws has done better. There are some very smart people involved, and surely they can come up with rules that don't ignore draws while not encouraging one to play for them either unless you are actually much worse than when the game began. If I think of something that seems fair I'll propose it. A lot of the problem is that the rules were designed for the queen odds bot where many players scored over 50%, but at the smaller odds almost no one plays at a slow enough time control to score that well. The rules don't work well in such cases.

If it is easy to get draws by early repetition, it is a deficiency of the bot, and to discourage draws, you should improve the bot. Here the developers want to learn from people who can get easy draws and improve the bot, but not give credit to the people who found these deficiencies.

I don't agree that it is a deficiency of the bot. The bots were designed with the assumption that players would choose handicaps and time limits that gave them roughly equal chances to win, so that they would have no incentive to make an early draw unless things had gone wrong. This is more or less the case for strong players playing at queen odds, but for knight odds and rook odds almost no one is strong enough and willing to play slow enough games to have balanced chances, so they are usually unable to win games except rarely and likely to accept early repetitions. These early reps are not something we didn't know about or anticipate, it's just that our goal was to be able to win matches against very strong players, and in a match a draw benefits neither player, it's just like the game was never played. We could change the goal to something like maximizing the score against a player 300 elo weaker than the handicapped bot, and then it would never take draws unless the situation was much worse than at the start, but at least until we actually win matches from grandmasters at classical time controls we probably won't do that. This is of course a different issue than the leaderboard question.

[syzygy]

If it is easy to get draws by early repetition, it is a deficiency of the bot, and to discourage draws, you should improve the bot. Here the developers want to learn from people who can get easy draws and improve the bot, but not give credit to the people who found these deficiencies.

With queen odds, that is not a deficiency of the bot. It is a deficiency of the rules.

But I agree with Uri that it is not correct to change the rules retroactively. Players should know what the rules are when they start a game, so they can decide their strategy.

[Uri Blass]

While I was expecting that something would be done to discourage playing for draws, the solution implemented was rather drastic. I think many people felt that the Leaderboard looked ridiculous when someone with almost no wins at all was on top, and they do have a point. It has nothing to do with you personally or your lack of a FIDE title or rating, it's just that it's known to be relatively easy to make early repetition draws since the engine doesn't mind. The argument on your side is that if two players have for example one win and nine losses each, but one player has no draws while the other has ten, then the player with the ten draws has done better. There are some very smart people involved, and surely they can come up with rules that don't ignore draws while not encouraging one to play for them either unless you are actually much worse than when the game began. If I think of something that seems fair I'll propose it. A lot of the problem is that the rules were designed for the queen odds bot where many players scored over 50%, but at the smaller odds almost no one plays at a slow enough time control to score that well. The rules don't work well in such cases.

If it is easy to get draws by early repetition, it is a deficiency of the bot, and to discourage draws, you should improve the bot. Here the developers want to learn from people who can get easy draws and improve the bot, but not give credit to the people who found these deficiencies.

I don't agree that it is a deficiency of the bot. The bots were designed with the assumption that players would choose handicaps and time limits that gave them roughly equal chances to win, so that they would have no incentive to make an early draw unless things had gone wrong. This is more or less the case for strong players playing at queen odds, but for knight odds and rook odds almost no one is strong enough and willing to play slow enough games to have balanced chances, so they are usually unable to win games except rarely and likely to accept early repetitions. These early reps are not something we didn't know about or anticipate, it's just that our goal was to be able to win matches against very strong players, and in a match a draw benefits neither player, it's just like the game was never played. We could change the goal to something like maximizing the score against a player 300 elo weaker than the handicapped bot, and then it would never take draws unless the situation was much worse than at the start, but at least until we actually win matches from grandmasters at classical time controls we probably won't do that. This is of course a different issue than the leaderboard question.

"The bots were designed with the assumption that players would choose handicaps and time limits that gave them roughly equal chances to win"
The problem is that this assumption is not correct because the bot wins most of the games(even if we talk about LeelaQueenOdds).
You can be careful to start with this assumption but as soon as you get a positive score against some human at some time control than it is better if you do not use this assumption and assume the human is weaker.

Another problem is that it seems the bot does not use the time of the human for the decision if to accept a draw by repetition or not and humans may draw by repetition when they are in time trouble.

[lkaufman]

While I was expecting that something would be done to discourage playing for draws, the solution implemented was rather drastic. I think many people felt that the Leaderboard looked ridiculous when someone with almost no wins at all was on top, and they do have a point. It has nothing to do with you personally or your lack of a FIDE title or rating, it's just that it's known to be relatively easy to make early repetition draws since the engine doesn't mind. The argument on your side is that if two players have for example one win and nine losses each, but one player has no draws while the other has ten, then the player with the ten draws has done better. There are some very smart people involved, and surely they can come up with rules that don't ignore draws while not encouraging one to play for them either unless you are actually much worse than when the game began. If I think of something that seems fair I'll propose it. A lot of the problem is that the rules were designed for the queen odds bot where many players scored over 50%, but at the smaller odds almost no one plays at a slow enough time control to score that well. The rules don't work well in such cases.

If it is easy to get draws by early repetition, it is a deficiency of the bot, and to discourage draws, you should improve the bot. Here the developers want to learn from people who can get easy draws and improve the bot, but not give credit to the people who found these deficiencies.

I don't agree that it is a deficiency of the bot. The bots were designed with the assumption that players would choose handicaps and time limits that gave them roughly equal chances to win, so that they would have no incentive to make an early draw unless things had gone wrong. This is more or less the case for strong players playing at queen odds, but for knight odds and rook odds almost no one is strong enough and willing to play slow enough games to have balanced chances, so they are usually unable to win games except rarely and likely to accept early repetitions. These early reps are not something we didn't know about or anticipate, it's just that our goal was to be able to win matches against very strong players, and in a match a draw benefits neither player, it's just like the game was never played. We could change the goal to something like maximizing the score against a player 300 elo weaker than the handicapped bot, and then it would never take draws unless the situation was much worse than at the start, but at least until we actually win matches from grandmasters at classical time controls we probably won't do that. This is of course a different issue than the leaderboard question.

"The bots were designed with the assumption that players would choose handicaps and time limits that gave them roughly equal chances to win"
The problem is that this assumption is not correct because the bot wins most of the games(even if we talk about LeelaQueenOdds).
You can be careful to start with this assumption but as soon as you get a positive score against some human at some time control than it is better if you do not use this assumption and assume the human is weaker.

Another problem is that it seems the bot does not use the time of the human for the decision if to accept a draw by repetition or not and humans may draw by repetition when they are in time trouble.

Yes, currently the bots do not consider any information about the opponent's identity, rating, past record, or clock time. Perhaps in the future this might change, but currently the priority is simply to improve the play of the bots regardless of the opponent. Given that we are more interested in blitz and Rapid than in bullet chess, I think this makes sense. The knight and rook odds bots are already virtually unbeatable in bullet and even in 3'0" blitz, so it would primarily just be relevant for queen odds.
Meanwhile I note that Catecan has resumed play against LQO, and is playing at 1'3" so he has some chance to win a game, which indeed he did today! Congratulations, Catecan!
I'll also note that there is discussion about just what the draw rules should be going forward. One proposal, which seems rather popular, is to resume rating draws regardless of how many occur, but to ignore (not rate) draws that occur until half the chessmen have disappeared from the board. This is similar in spirit to the widely used tournament rule forbidding draws by agreement before move 30 or 40, which is another option being discussed (for non-rating). Personally I favor such rules against rating early draws rather than one that counts the number of draws since the last win or counts draws at half or quarter rate. If you make it to the endgame (or even anything close to an endgame) and can still force a draw, I think you have earned it! But the choice may depend on how easy such rules are to implement.