Details about Leela Chalange in Jerusalem

[lkaufman]

1)IM Yannay Ben Ari fide rating 2457(national rating 2489) age 21
2)IM Roy Vagman 2377(national rating 2357) age 17
3)Ilay Zakin 2311(2329) fide master
4)Maxim Milavsky 2284(2307) means fide rating 2284 national rating 2307(not active recently except one blitz tournament)
A national master
5)Eliran Shilon Rahav 2270(2306) third place in U14 europe championship and has expected fide rating above 2300. national master and also fide master after the next update.
https://s1.chess-results.com/tnr1282645 ... 0&SNode=S0
6)Guy Gov 2224(2254) national master
7)Itay Sitbon 2169(2211) national master
8)Pavel Chernomordik 2167(2237) national and fide master
9)Neria Gurevich 2131(2187) 12 years old player senior candidate to national master
10)Yosef Gurevich 2081(2201) senior candidate to national master
11)Itay Tsafrir 2075(2085) candidate to national master
12)Alon Cohen 2057(2025) candidate to national master
13)Igor Berenboym 2043(2056) senior candidate to national master
14)Lior Zakin 1939(1992)
15)Hillel Gurevich 1905(2085) candidate to national master
16)Oded Ross 0(2082) candidate to national master but unactive)
17)Slavik Altman 1716(1756) 2 ranks below candidate master.
18)Eyal Cohen 1671(1832)

I decided to allow players with national elo below 2000 to play with NN odds in the first game when they win only 1/3 of the prize money
for the same result against R odds(these players are 14,17,18 but I am not sure if they choose this option).

18 will have no choice if some strong player decide to join in the last moment.

I note that FM Ilay Zakin (above) played 20 games with LeelaRookOdds in the last day at 5'3", scoring two draws, 18 losses. Will classical time limits be enough to raise his score from 5% to equal chances? Perhaps so, we shall see. If possible let me know tonite who will play with two knights odds (I can make a last minute change if necessary). I will probably have to shut down LQO during the event unless someone else on the team can host it for those hours.

[Uri Blass]

Pavel Chernomordik decided not to play so again there are only 18 players.

Based on my knowledge all prefer to play against the rook odds in the first game and not against the NN bot.
I will tell them that they cannot decide in the last second about it.

[lkaufman]

Pavel Chernomordik decided not to play so again there are only 18 players.

Based on my knowledge all prefer to play against the rook odds in the first game and not against the NN bot.
I will tell them that they cannot decide in the last second about it.

If there are 18 players for rook odds, I think we should have the players below 2000 play on "LeelaPieceOdds" which runs on the 4090. It is not currently set up for rook odds but I can fix that, the only issue is that to start a game, the player or operator will need to copy the board from "LeelaRookOdds" and paste it in "LeelaPieceOdds". The reason I want to do this is that although calculations tell me that 18 games on the 5090 should work, I could be wrong, better to be safe. Also if Leela is on move at the same time in too many games, bad things may happen, so this would minimize that risk. The moves may take an extra second or so on the 4090, that shouldn't make a noticeable difference, and play should be very slightly faster on the 5090. Is this okay?

[Uri Blass]

Pavel Chernomordik decided not to play so again there are only 18 players.

Based on my knowledge all prefer to play against the rook odds in the first game and not against the NN bot.
I will tell them that they cannot decide in the last second about it.

If there are 18 players for rook odds, I think we should have the players below 2000 play on "LeelaPieceOdds" which runs on the 4090. It is not currently set up for rook odds but I can fix that, the only issue is that to start a game, the player or operator will need to copy the board from "LeelaRookOdds" and paste it in "LeelaPieceOdds". The reason I want to do this is that although calculations tell me that 18 games on the 5090 should work, I could be wrong, better to be safe. Also if Leela is on move at the same time in too many games, bad things may happen, so this would minimize that risk. The moves may take an extra second or so on the 4090, that shouldn't make a noticeable difference, and play should be very slightly faster on the 5090. Is this okay?

I assume no problem.

There are 3 players with national elo below 2000(other may have fide rating below 2000 but at least have the rank of israeli candidate master.

[Uri Blass]

I can add that in the beginning I do not expect 18 games at the same time because 4 players are going to come late.

Problem of 18 games at the same time may be when they join when part of the players are going to be in the second game and part of the players are going to be in their first game.

[lkaufman]

I can add that in the beginning I do not expect 18 games at the same time because 4 players are going to come late.

Problem of 18 games at the same time may be when they join when part of the players are going to be in the second game and part of the players are going to be in their first game.

I modified LeelaPieceOdds to accept Rook (a1) odds with normal settings. Perhaps you can try a game or two yourself just to make sure it works properly. Just copy the fen from LeelaRookOdds and paste in LeelaPieceOdds.

[lkaufman]

I can add that in the beginning I do not expect 18 games at the same time because 4 players are going to come late.

Problem of 18 games at the same time may be when they join when part of the players are going to be in the second game and part of the players are going to be in their first game.

Will they start later than some of the first group should finish? If so maybe I can just run all rook odds games on 5090, and any knight odds games on 4090, unless no one wins at rook odds, then maybe I should use LeelaPieceOdds to be safe.

[Uri Blass]

I can add that in the beginning I do not expect 18 games at the same time because 4 players are going to come late.

Problem of 18 games at the same time may be when they join when part of the players are going to be in the second game and part of the players are going to be in their first game.

Will they start later than some of the first group should finish? If so maybe I can just run all rook odds games on 5090, and any knight odds games on 4090, unless no one wins at rook odds, then maybe I should use LeelaPieceOdds to be safe.

I guess they will start 2 hours later or 2.5 hours later(not sure about the exact time they arrive).

I guess at least part of the games are going to take less than 2 hours.

[lkaufman]

So I think what I'll do is to start all the games of those present at the starting time (or by the end of the staggered start) on the 5090, they just need to challenge LeelaRookOdds at the proper TC with Leela playing White. I'll set the max number of games to that number, 16 or whatever it might be but not more than 17. Anyone starting a game after that at rook odds, whether the first or second game, can do the same unless the max number is in progress, in which case a challenge will be declined. In that case, you can either ask the player to wait for some game to finish (if it looks like a short wait), or he can play by challenging LeelaPieceOdds and pasting in the position from LRO. If a player is going to play the second game at knight odds, just challenge LeelaKnightOdds (with Leela White) and it will automatically play on the 4090. That should avoid any significant waiting and also avoid any significant risk of a crash due to overloading.

[lkaufman]

Pavel Chernomordik decided not to play so again there are only 18 players.!

Based on my knowledge all prefer to play against the rook odds in the first game and not against the NN bot.
I will tell them that they cannot decide in the last second about it.

I think I know why he decided not to play. He played 26 blitz games (3'2") with LeelaRookOdds, and lost every one of them! I guess he concluded it was "mission impossible". Eight of the other players have played multiple games with it in the last few days, most players choosing either Rapid or 5'3", some 5'2" or 3'2" (one even played at 90'30"). I count the total non-bullet score (including Pavel) as 1 win, 4 draws, and 184 losses! The only winner was the top rated IM, who scored 1 win and 1 draw in six games at 5'3". So perhaps he is favored in his match, but probably every one else (except perhaps the other IM, who didn't play yet) will be just hoping to get a draw.