She only learns about things she has seen before, mainly positional, the network doesn't understand anything about tactics, maybe she will learn some shallow tactics by means of the history planes, but the majority of the tactics have to be resolved by the MCTS, and it looks like there is something amiss there.AdminX wrote:Well isn't the theory that she's supposed to teach herself. Otherwise what is the purpose of self play and reinforcement learning? The same might even be said of giving it Syzygy support.Joost Buijs wrote:Indeed, this is a simple 4 ply tactic, I would expect that the MCTS picks this up, maybe the network thinks Qxd2 is so bad that it never considers this move in the playouts. I have no clue about how the algorithm of LCZero exactly works, I never looked at the code, but I have the feeling that there are many things that need improvement before you can even start thinking about reaching the level of Stockfish.AlvaroBegue wrote:I have evidence that Leela is horrible at tactics at any time control. I posted a reproducible problem on the LCZero forum, but the responses I got are mostly from fan boys that don't want to see the problem.mar wrote:[...] So tactics seems to be the Achilles heel of Leela, at least at this fast TC.
https://groups.google.com/forum/#!topic ... q3lg9QV2XQ
Basically you can enter these UCI commands:[d]r3r1k1/pp3ppp/1npb4/8/3P3N/1PP2Q2/q2B1PPP/3RR1K1 w - - 1 20[/d]Code: Select all
position fen r4rk1/pp1b1ppp/1npb4/q5B1/3P3N/1BP5/P4PPP/R2QR1K1 w - - 9 16 moves g5d2 f8e8 d1f3 d7e6 a1d1 e6b3 a2b3 a5a2 c3c4 go infinite
This is a position where Qxd2 (a2d2) gains a bishop. The queen cannot be recaptured because of a bank-rank mate. This tactic is so simple that is obvious to me (Elo ~1500).
After 3 minutes of thinking time and over 200K playouts, LCZero had considered the correct move 0 times (!!!).
They need a better search algorithm but they are not looking for one yet.
LCZero: Progress and Scaling. Relation to CCRL Elo
Moderators: hgm, Rebel, chrisw
-
- Posts: 1563
- Joined: Thu Jul 16, 2009 10:47 am
- Location: Almere, The Netherlands
Re: LCZero: Progress and Scaling. Relation to CCRL Elo
-
- Posts: 291
- Joined: Wed May 08, 2013 6:49 am
Re: LCZero: Progress and Scaling. Relation to CCRL Elo
It's probably because tactics is a conflict in its learning because dropping a piece is usually learned as bad. For example out of 100 queen sacrifices 99 of it might be loosing and only one wins. As a result of the probability it learns to avoid giving up its queen.AdminX wrote:Sounds like she will need an additional tactical module added to her system.syzygy wrote:But there will necessarily be a limit on the tactics that its NN can resolve itself. If the NN incorrectly classifies a tactical move as bad with the result that the search never even looks at it, there is a problem.AdminX wrote:Well isn't the theory that she's supposed to teach herself. Otherwise what is the purpose of self play and reinforcement learning.
-
- Posts: 2554
- Joined: Fri Nov 26, 2010 2:00 pm
- Location: Czech Republic
- Full name: Martin Sedlak
Re: LCZero: Progress and Scaling. Relation to CCRL Elo
Fully agreed. This is THE thing they should focus on now.AlvaroBegue wrote:They need a better search algorithm but they are not looking for one yet.
Current SF dev should be as strong as A0.
I wonder if something simple and stupid (like dedicating 1 thread for say material-only alphabeta searcher to guide tree expansion) may work, since net evals are very expensive.
But shouldn't the search converge regardless given more time to think? Your position is 2 plies + qsearch, piece of cake for any AB searcher out there...
I can't imagine A0 beating SF when missing tactics such as this one.
EDIT: the tactics is slightly more complicated than I thought, not 2 plies but still very easy for AB searchers
-
- Posts: 6339
- Joined: Mon Mar 13, 2006 2:34 pm
- Location: Acworth, GA
Re: LCZero: Progress and Scaling. Relation to CCRL Elo
So she is unable to comprehend paradox situations in her NN. Odd given some of her exchange sacs.Jhoravi wrote:It's probably because tactics is a conflict in its learning because dropping a piece is usually learned as bad. For example out of 100 queen sacrifices 99 of it might be loosing and only one wins. As a result of the probability it learns to avoid giving up its queen.AdminX wrote:Sounds like she will need an additional tactical module added to her system.syzygy wrote:But there will necessarily be a limit on the tactics that its NN can resolve itself. If the NN incorrectly classifies a tactical move as bad with the result that the search never even looks at it, there is a problem.AdminX wrote:Well isn't the theory that she's supposed to teach herself. Otherwise what is the purpose of self play and reinforcement learning.
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
__________________________________________________________________
Ted Summers
-
- Posts: 408
- Joined: Sat Mar 06, 2010 9:28 am
Re: LCZero: Progress and Scaling. Relation to CCRL Elo
On the other hand it means that LCZero is already positionally extremely strong if it can reach 3000 Elo level while being tactically thus weak. I have the opposite feeling that an update for better tactical search shouldn‘t be that hard to find and therefore SF will be bettered soon.Joost Buijs wrote:Indeed, this is a simple 4 ply tactic, I would expect that the MCTS picks this up, maybe the network thinks Qxd2 is so bad that it never considers this move in the playouts. I have no clue about how the algorithm of LCZero exactly works, I never looked at the code, but I have the feeling that there are many things that need improvement before you can even start thinking about reaching the level of Stockfish.AlvaroBegue wrote:I have evidence that Leela is horrible at tactics at any time control. I posted a reproducible problem on the LCZero forum, but the responses I got are mostly from fan boys that don't want to see the problem.mar wrote:[...] So tactics seems to be the Achilles heel of Leela, at least at this fast TC.
https://groups.google.com/forum/#!topic ... q3lg9QV2XQ
Basically you can enter these UCI commands:[d]r3r1k1/pp3ppp/1npb4/8/3P3N/1PP2Q2/q2B1PPP/3RR1K1 w - - 1 20[/d]Code: Select all
position fen r4rk1/pp1b1ppp/1npb4/q5B1/3P3N/1BP5/P4PPP/R2QR1K1 w - - 9 16 moves g5d2 f8e8 d1f3 d7e6 a1d1 e6b3 a2b3 a5a2 c3c4 go infinite
This is a position where Qxd2 (a2d2) gains a bishop. The queen cannot be recaptured because of a bank-rank mate. This tactic is so simple that is obvious to me (Elo ~1500).
After 3 minutes of thinking time and over 200K playouts, LCZero had considered the correct move 0 times (!!!).
They need a better search algorithm but they are not looking for one yet.
-
- Posts: 6339
- Joined: Mon Mar 13, 2006 2:34 pm
- Location: Acworth, GA
Re: LCZero: Progress and Scaling. Relation to CCRL Elo
While totally destroying the relationship between tactics and strategy in chess in terms of value as we knew it. Who would have thought that something so tactically weak could slaughter a Grandmaster.Ralph Stoesser wrote:On the other hand it means that LCZero is already positionally extremely strong if it can reach 3000 Elo level while being tactically thus weak. I have the opposite feeling that an update for better tactical search shouldn‘t be that hard to find and therefore SF will be bettered soon.Joost Buijs wrote:Indeed, this is a simple 4 ply tactic, I would expect that the MCTS picks this up, maybe the network thinks Qxd2 is so bad that it never considers this move in the playouts. I have no clue about how the algorithm of LCZero exactly works, I never looked at the code, but I have the feeling that there are many things that need improvement before you can even start thinking about reaching the level of Stockfish.AlvaroBegue wrote:I have evidence that Leela is horrible at tactics at any time control. I posted a reproducible problem on the LCZero forum, but the responses I got are mostly from fan boys that don't want to see the problem.mar wrote:[...] So tactics seems to be the Achilles heel of Leela, at least at this fast TC.
https://groups.google.com/forum/#!topic ... q3lg9QV2XQ
Basically you can enter these UCI commands:[d]r3r1k1/pp3ppp/1npb4/8/3P3N/1PP2Q2/q2B1PPP/3RR1K1 w - - 1 20[/d]Code: Select all
position fen r4rk1/pp1b1ppp/1npb4/q5B1/3P3N/1BP5/P4PPP/R2QR1K1 w - - 9 16 moves g5d2 f8e8 d1f3 d7e6 a1d1 e6b3 a2b3 a5a2 c3c4 go infinite
This is a position where Qxd2 (a2d2) gains a bishop. The queen cannot be recaptured because of a bank-rank mate. This tactic is so simple that is obvious to me (Elo ~1500).
After 3 minutes of thinking time and over 200K playouts, LCZero had considered the correct move 0 times (!!!).
They need a better search algorithm but they are not looking for one yet.
Last edited by AdminX on Sat May 05, 2018 5:57 pm, edited 1 time in total.
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
__________________________________________________________________
Ted Summers
-
- Posts: 143
- Joined: Wed Jan 17, 2018 1:26 pm
Re: LCZero: Progress and Scaling. Relation to CCRL Elo
A sufficiently deep neural net with enough training can increasingly resolve tactics without search. You can already see that even current Leela on 1 node (i.e. easy mode on play.lczero.org) can at least see simple tactics such as which pieces are protected and which aren't. If a neural net were incapable of resolving any tactics, this would not be possible. It often misses when tactics get more complicated, yet this tactical ability is already a lot better than it was a few weeks ago, or the pure net could not defeat Stockfish at Skill level 5-6 (which it can now).mar wrote:Fully agreed. This is THE thing they should focus on now.AlvaroBegue wrote:They need a better search algorithm but they are not looking for one yet.
Current SF dev should be as strong as A0.
I wonder if something simple and stupid (like dedicating 1 thread for say material-only alphabeta searcher to guide tree expansion) may work, since net evals are very expensive.
But shouldn't the search converge regardless given more time to think? Your position is 2 plies + qsearch, piece of cake for any AB searcher out there...
I can't imagine A0 beating SF when missing tactics such as this one.
EDIT: the tactics is slightly more complicated than I thought, not 2 plies but still very easy for AB searchers
AlphaZero used larger neural net (256x20) and ran a lot more training games on it before the match against Stockfish, so it makes sense that their network had better tactical understanding than current Leela.
-
- Posts: 3019
- Joined: Wed Mar 08, 2006 9:57 pm
- Location: Rio de Janeiro, Brazil
Re: LCZero: Progress and Scaling. Relation to CCRL Elo
80,000 NPS didn't hurt either...jkiliani wrote:A sufficiently deep neural net with enough training can increasingly resolve tactics without search. You can already see that even current Leela on 1 node (i.e. easy mode on play.lczero.org) can at least see simple tactics such as which pieces are protected and which aren't. If a neural net were incapable of resolving any tactics, this would not be possible. It often misses when tactics get more complicated, yet this tactical ability is already a lot better than it was a few weeks ago, or the pure net could not defeat Stockfish at Skill level 5-6 (which it can now).mar wrote:Fully agreed. This is THE thing they should focus on now.AlvaroBegue wrote:They need a better search algorithm but they are not looking for one yet.
Current SF dev should be as strong as A0.
I wonder if something simple and stupid (like dedicating 1 thread for say material-only alphabeta searcher to guide tree expansion) may work, since net evals are very expensive.
But shouldn't the search converge regardless given more time to think? Your position is 2 plies + qsearch, piece of cake for any AB searcher out there...
I can't imagine A0 beating SF when missing tactics such as this one.
EDIT: the tactics is slightly more complicated than I thought, not 2 plies but still very easy for AB searchers
AlphaZero used larger neural net (256x20) and ran a lot more training games on it before the match against Stockfish, so it makes sense that their network had better tactical understanding than current Leela.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
-
- Posts: 5557
- Joined: Tue Feb 28, 2012 11:56 pm
Re: LCZero: Progress and Scaling. Relation to CCRL Elo
But they don't help if the search simply never looks at the key move.Albert Silver wrote:80,000 NPS didn't hurt either...
-
- Posts: 3019
- Joined: Wed Mar 08, 2006 9:57 pm
- Location: Rio de Janeiro, Brazil
Re: LCZero: Progress and Scaling. Relation to CCRL Elo
Assuredly, but you also have to give the network time since as it rains it will refine its concepts. All I have to do is look at how the engine has improved over the last month even in just tactics. In any case, one can widen the move selection or narrow it directly in the UCI engine settings. It warrants testing and that's exactly what I'm doing. I'll report when I have some data.syzygy wrote:But they don't help if the search simply never looks at the key move.Albert Silver wrote:80,000 NPS didn't hurt either...
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."