Strange Lc0 TCEC performance

[Leo]

I have had major doubts about Lc0 from the beginning. It reminds me of quantum computing. It has all this potential but is still years away from emerging. " 1 bird in the hand is better than 2 in the bush's."

[Dann Corbit]

I guess that they will do better next year.

Lc0 was not going to win this year, anyway. It's a long way below the best engines.

There seem to be many who disagree with you regarding the magnitude---and, in some cases, the sign---of the gap. Based on my visits to TCEC recently, Lc0 has quite a "cult" following. More so than any other engine, in my experience.

The amount of fanboyism there is unprecedented. Today was a very sad day for most of the crowd there. Most would be pretty depressive by tomorrow and TCEC viewers would drop to 1/4 of what was so far.

I think that people being excited by something is almost always a good thing.
Enthusiasm make it more fun for everyone.
The neural approach (as far as a successful one) is pretty new, so there is no telling where that technology will go.
Alpha beta searchers have been around for a very long time, and there have been thousands of steps in refinements.
By comparison, Matthew Lai's approach is a new copper penny.

[frankp]

I have had major doubts about Lc0 from the beginning. It reminds me of quantum computing. It has all this potential but is still years away from emerging. " 1 bird in the hand is better than 2 in the bush's."

I think A0 showed the potential in Go, chess and shogi.
The only issue seemed to be the special hardware requirement - several powerful TPU. Or we could all be running A0 if they had release it.

As for Lc0, as I understand it, it aims to reproduce the methods of A0. And seems to have come a long way in a very short time. Compared for example to the many decades of hard work by the pioneers of min/max, alpha-beta, bit-boards, egtb etc. Lc0 already plays competent and interesting chess, at worst and will probably only get better. Those that follow will, I imagine, be better still by building on the basics or adopting new hybrid approaches. The sticking point for me at the moment is the hardware requirement. But that will change too. I find it a fascinating project. The most interesting thing that has happened in chess for several decades.

What surprises me, well actually it does not, is the hostility (not suggesting you) some exhibit towards the project. Apparently for even daring to exist.

[fersbery]

I have had major doubts about Lc0 from the beginning. It reminds me of quantum computing. It has all this potential but is still years away from emerging. " 1 bird in the hand is better than 2 in the bush's."

I think A0 showed the potential in Go, chess and shogi.
The only issue seemed to be the special hardware requirement - several powerful TPU. Or we could all be running A0 if they had release it.

As for Lc0, as I understand it, it aims to reproduce the methods of A0. And seems to have come a long way in a very short time. Compared for example to the many decades of hard work by the pioneers of min/max, alpha-beta, bit-boards, egtb etc. Lc0 already plays competent and interesting chess, at worst and will probably only get better. Those that follow will, I imagine, be better still by building on the basics or adopting new hybrid approaches. The sticking point for me at the moment is the hardware requirement. But that will change too. I find it a fascinating project. The most interesting thing that has happened in chess for several decades.

What surprises me, well actually it does not, is the hostility (not suggesting you) some exhibit towards the project. Apparently for even daring to exist.

Could not agree more.

[AndrewGrant]

Without crippled hardware, Leela would be on first place.
Deux maybe second place!

99% will go offline without NN engines at TCEC.

Shame to see people leave, but will return the chat to actual talk of chess development and play.

I would hope NN-fans would be happy that TCEC did all this work to get their own server.

[jkiliani]

Without crippled hardware, Leela would be on first place.
Deux maybe second place!

99% will go offline without NN engines at TCEC.

Shame to see people leave, but will return the chat to actual talk of chess development and play.

I would hope NN-fans would be happy that TCEC did all this work to get their own server.

It's pretty clear by now that actual chess engine development in the next few years is going to be about how to use neural networks most effectively, since Leela has demonstrated the AlphaZero concept rather conclusively. Alpha-Beta searchers may have a clear edge with tactics and the use of tablebases, but they can't compete with a strong NN engine in quiet positional play. In addition to improvements on the tree search and neural net architecture, the next few years will see people trying to fuse NN and AB engines to combine the strengths of traditional engines (tactics, endgame when using TB) with the strengths of neural networks (opening, quiet midgame). As AB-engines go, Ethereal is a very fine piece of engineering, but a successful fusion of its search with Leela's would make it world-class.

There's a reason NN has so many fans, and it's not (mainly) the playing strength: It has revived computer chess by producing actually exciting games that break with preconceptions about how you're supposed to play, and yet manage to win somehow. Chess.com has realised this, which is why they are going all in for GPU infrastructure. I'll give TCEC points for effort, but they'll have to up their game in Season 14 to get viewers back.

[AndrewGrant]

Without crippled hardware, Leela would be on first place.
Deux maybe second place!

99% will go offline without NN engines at TCEC.

Shame to see people leave, but will return the chat to actual talk of chess development and play.

I would hope NN-fans would be happy that TCEC did all this work to get their own server.

It's pretty clear by now that actual chess engine development in the next few years is going to be about how to use neural networks most effectively, since Leela has demonstrated the AlphaZero concept rather conclusively. Alpha-Beta searchers may have a clear edge with tactics and the use of tablebases, but they can't compete with a strong NN engine in quiet positional play. In addition to improvements on the tree search and neural net architecture, the next few years will see people trying to fuse NN and AB engines to combine the strengths of traditional engines (tactics, endgame when using TB) with the strengths of neural networks (opening, quiet midgame). As AB-engines go, Ethereal is a very fine piece of engineering, but a successful fusion of its search with Leela's would make it world-class.

There's a reason NN has so many fans, and it's not (mainly) the playing strength: It has revived computer chess by producing actually exciting games that break with preconceptions about how you're supposed to play, and yet manage to win somehow. Chess.com has realised this, which is why they are going all in for GPU infrastructure. I'll give TCEC points for effort, but they'll have to up their game in Season 14 to get viewers back.

Clear to who? Until you can write an NN that runs as well as AB on a CPU, then the only thing you are proving is that massive specialized hardware trumps a general purpose processor. Thats it. Cut and dry. No further interest.

Merging engines is something people with no clue what they are talking about say. Its condescending to even suggest it to an author.

No offense to Erik or to chess.com, but they have not realized jack sh*t. They are playing with absurd settings, no opening book, trying to force hyper threading with pondering. They even made their release without telling authors before hand. They don't have the slightest clue about ANYTHING that goes on in this forum. They have $$$, and they see NN fans, and business does the rest.

[jp]

...
Well the way i see it is that the server they rented for the GPUs is HORRIBLE. They even lowered the GPUs speed to 65% and they are still overheating!
So the way i see it, is with the hardware they have chosen they crippled the NNS a LOT! So to ensure a sensible tournament, better late than never, they "fixed"(by lowering GPUs speed to 65%/85% in each GPU, but at least preventing the complete crippling of NNs that was happening) that even perhaps against their rules as you say, 9 games before the end.
I find it fair not to be a complete unfairness to NNs.

Imagine what people would say if this had happened in AB engines in the 43 core system and that to be overheating or something and the engines had to run with 70% speed for example. They would corrected it as fast as possible of course. Even breaking their rules.
And this is the hypocrisy of many. That are screaming now for all these against the 2 NNs(mainly Leela) but in the opposing situation they would scream that things should be fixed in the hardware.

TCEC guys did not know there would be problems. The new start-up they rented from did not know there would be problems. TCEC had to go out & get hardware they didn't have before.

Then instead of being grateful people make it sound like they deliberately "crippled" the NNs.

Right at the start, before Div 4, TCEC said the next higher GPU option was MUCH MUCH MUCH more expensive. What do people want them to do? If people DONATE a TPU to TCEC, I'm sure they'll use it.

[jkiliani]

Clear to who? Until you can write an NN that runs as well as AB on a CPU, then the only thing you are proving is that massive specialized hardware trumps a general purpose processor. Thats it. Cut and dry. No further interest.

Merging engines is something people with no clue what they are talking about say. Its condescending to even suggest it to an author.

No offense to Erik or to chess.com, but they have not realized jack sh*t. They are playing with absurd settings, no opening book, trying to force hyper threading with pondering. They even made their release without telling authors before hand. They don't have the slightest clue about ANYTHING that goes on in this forum. They have $$$, and they see NN fans, and business does the rest.

The large majority of chess players think AlphaZero (and by extension, Leela) games are more entertaining to watch than AB engine games. Chess.com are simply providing what the audience wants to see, which is what any business is doing. You can disagree with their settings all you want, after all you're not forced to watch cccc. Even better, if the Chess.com tournament is a success, you'll be left alone by NN fans at TCEC, wouldn't that be great?

About merging engines, it's certainly your right to keep Ethereal "pure", but don't be too surprised if others pick up the challenge, and the result turns out better than regular AB.

[AndrewGrant]

Even better, if the Chess.com tournament is a success, you'll be left alone by NN fans at TCEC, wouldn't that be great?

I hope and pray.

About merging engines, it's certainly your right to keep Ethereal "pure", but don't be too surprised if others pick up the challenge, and the result turns out better than regular AB.

Not on a CPU you won't

On the topic of playstyle -- I'm yet to see anyone which actual chess knowledge show that Leela games are different in any capacity.

Watch a game of TCEC -- Leela best moves will match the opponents predictions for large chunks of the game. This "humanlike" marketing gimmick will die at some point.