LCZero Accomplishments and Goals Thus Far

[JJJ]

Lczero is a very recent engines. But it is already more stronger than many engines who needed years to be that strong.

So be patient, and one day you will be happy to see him crush Stockfish.

[Albert Silver]

It still blunders hugely after the +300 elo.

It does, no question about it. Even with the new cuDNN speedups (which need testing to see how that translates into results). It is a fascinating engine, but there is a ways to go yet.

[Laskos]

It still blunders hugely after the +300 elo.

It does, no question about it. Even with the new cuDNN speedups (which need testing to see how that translates into results). It is a fascinating engine, but there is a ways to go yet.

With this sort of speedu-up due to cuDNN, at LTC, say 60'+ 15'', your LC0 should be about the level of Houdini 1.5a on 4 threads. It would be a fascinating match, H1.5 was a very tactically astute engine. In 10 games, there will be some LC0 wins, and I would be curious how LC0 missing mates in 3, and generally being very weak in tactics, especially in tactical puzzles, can beat H1.5, mostly positionally and on long term "plans". Seeing this cuDNN speed-up, I will get a GTX 1060 in less than 2 weeks.

[IQ]

Because it addresses me I don't feel it would be correct to act against it myself, so I will leave the decision to another moderator.[/b][/color]

I am amazed that whenever you think this forum can't sink to new lows it actually does. Also I think maybe its a good idea to let Milos post just stand and not delete it, as it reveals a lot about him - and informs new users about his social skills and trollish attitudes.

[JJJ]

It still blunders hugely after the +300 elo.

It does, no question about it. Even with the new cuDNN speedups (which need testing to see how that translates into results). It is a fascinating engine, but there is a ways to go yet.

With this sort of speedu-up due to cuDNN, at LTC, say 60'+ 15'', your LC0 should be about the level of Houdini 1.5a on 4 threads. It would be a fascinating match, H1.5 was a very tactically astute engine. In 10 games, there will be some LC0 wins, and I would be curious how LC0 missing mates in 3, and generally being very weak in tactics, especially in tactical puzzles, can beat H1.5, mostly positionally and on long term "plans". Seeing this cuDNN speed-up, I will get a GTX 1060 in less than 2 weeks.

I will run 10 games with Leela ID 127 against Houdini 1.5 at 3min+2 sec, dont have time to higher time control. Still might be interesting. I m not sure Leela is as strong as Houdini 1.5a, which at the time I started chess was the strongest engine.

[Laskos]

It still blunders hugely after the +300 elo.

It does, no question about it. Even with the new cuDNN speedups (which need testing to see how that translates into results). It is a fascinating engine, but there is a ways to go yet.

With this sort of speedu-up due to cuDNN, at LTC, say 60'+ 15'', your LC0 should be about the level of Houdini 1.5a on 4 threads. It would be a fascinating match, H1.5 was a very tactically astute engine. In 10 games, there will be some LC0 wins, and I would be curious how LC0 missing mates in 3, and generally being very weak in tactics, especially in tactical puzzles, can beat H1.5, mostly positionally and on long term "plans". Seeing this cuDNN speed-up, I will get a GTX 1060 in less than 2 weeks.

I will run 10 games with Leela ID 127 against Houdini 1.5 at 3min+2 sec, dont have time to higher time control. Still might be interesting. I m not sure Leela is as strong as Houdini 1.5a, which at the time I started chess was the strongest engine.

LTC and that 4-8x boost in NPS due to cuDNN are vital. Also, why ID127? Take some of the latest IDs.

[JJJ]

It still blunders hugely after the +300 elo.

It does, no question about it. Even with the new cuDNN speedups (which need testing to see how that translates into results). It is a fascinating engine, but there is a ways to go yet.

With this sort of speedu-up due to cuDNN, at LTC, say 60'+ 15'', your LC0 should be about the level of Houdini 1.5a on 4 threads. It would be a fascinating match, H1.5 was a very tactically astute engine. In 10 games, there will be some LC0 wins, and I would be curious how LC0 missing mates in 3, and generally being very weak in tactics, especially in tactical puzzles, can beat H1.5, mostly positionally and on long term "plans". Seeing this cuDNN speed-up, I will get a GTX 1060 in less than 2 weeks.

I will run 10 games with Leela ID 127 against Houdini 1.5 at 3min+2 sec, dont have time to higher time control. Still might be interesting. I m not sure Leela is as strong as Houdini 1.5a, which at the time I started chess was the strongest engine.

LTC and that 4-8x boost in NPS due to cuDNN are vital. Also, why ID127? Take some of the latest IDs.

About speed, there isn't much than I can do. About ID 127, it is almost the same than the lastest, perhaps 10 elo below, not even sure. But yeah, I guess I can use the last. I might do that later and increase time control to rapid.

[Damir]

It still blunders hugely after the +300 elo.

It does, no question about it. Even with the new cuDNN speedups (which need testing to see how that translates into results). It is a fascinating engine, but there is a ways to go yet.

With this sort of speedu-up due to cuDNN, at LTC, say 60'+ 15'', your LC0 should be about the level of Houdini 1.5a on 4 threads. It would be a fascinating match, H1.5 was a very tactically astute engine. In 10 games, there will be some LC0 wins, and I would be curious how LC0 missing mates in 3, and generally being very weak in tactics, especially in tactical puzzles, can beat H1.5, mostly positionally and on long term "plans". Seeing this cuDNN speed-up, I will get a GTX 1060 in less than 2 weeks.

Maybe you should try this baby.

ASUS Strix GeForce GTX 1080 Ti OC, 11 GB

[frankp]

https://blog.slavv.com/picking-a-gpu-fo ... 4795c273b9

Might be relevant.

[CMCanavessi]

It still blunders hugely after the +300 elo.

It does, no question about it. Even with the new cuDNN speedups (which need testing to see how that translates into results). It is a fascinating engine, but there is a ways to go yet.

With this sort of speedu-up due to cuDNN, at LTC, say 60'+ 15'', your LC0 should be about the level of Houdini 1.5a on 4 threads. It would be a fascinating match, H1.5 was a very tactically astute engine. In 10 games, there will be some LC0 wins, and I would be curious how LC0 missing mates in 3, and generally being very weak in tactics, especially in tactical puzzles, can beat H1.5, mostly positionally and on long term "plans". Seeing this cuDNN speed-up, I will get a GTX 1060 in less than 2 weeks.

I will run 10 games with Leela ID 127 against Houdini 1.5 at 3min+2 sec, dont have time to higher time control. Still might be interesting. I m not sure Leela is as strong as Houdini 1.5a, which at the time I started chess was the strongest engine.

LTC and that 4-8x boost in NPS due to cuDNN are vital. Also, why ID127? Take some of the latest IDs.

About speed, there isn't much than I can do. About ID 127, it is almost the same than the lastest, perhaps 10 elo below, not even sure. But yeah, I guess I can use the last. I might do that later and increase time control to rapid.

You probably meant 227. ID 232 is MUCH stronger.