Dylan Sharp Vs. Harvey Williamson (G4)

[Harvey Williamson]

Still Stockfish self-play. Still hoping for something "novel".

There are only a couple of possible moves in the positions so far. Hopefully I will unleash the novelty you are waiting for soon.........

[Zenmastur]

Hey Max

Items A, B, and C either are conclusions or contain conclusions that are false.
E.G. in the case of A it could be that the PV didn't have a blunder in it and the “true” value of the root position is actually 0.00.

E.g. in the case of B it could be that the Pv doesn't have a blunder in it and the “true” value of the root position is -2.xx or less. In fact, if you are analyzing for black, one of the main reasons why you might conduct your analysis using compound searches etc. is to find positions that "prove" that this, is in fact, the case.

Have you ever seen that happen?

Yes I have. That's why I brought it up.

With this much material on the board at move 6 Stockfish giving some 1.50 scores at high depth and when you force the next 15 moves it goes down to 0.00 or more than 2.00 and when you try to find alternative moves you can't find any?

I guess it's a thing that could happen. It could also happen that when you put a random mover against Stockfish 10, the random mover wins on its first try, but you wouldn't use such unlikely cases to defend a stance. Stockfish's Depth 60 becomes increasingly more crappy as you check moves from its PV, and the eval is nothing more than what you'd get if you played the entire PV and played a DEPTH 1 move at the tail of the variation.

There is a reason that happens. I suggest you spend some time and figure out why it happens and then find a way to fix it. I think you'll find that after you have done that your CC game results will improve. Most programs ( at least that I have used have this problem) and there is a way to mitigate it's effects. To actually fix it would require coding changes in the program and this would likely take considerable time to find the “right” fix. Your machine is so old there's probably not much you can do about it until you upgrade it. I'll give you one hint: It has nothing to do with a lack of processing power.

Depth 60 showing the mainline of the game yet showing a wrong score is pure fantasy, that only happens in positions near the end of the game (and when I say end, I don't mean endgames, I mean when a side can force checkmate or force a draw from the root position.)

Hmmm....

You must live in a different universe than I do. Depth 60 searches can certainly return scores that aren't representative of the “true” value of a position regardless of the phase of the game the root position comes from. They might be more rare than the same coming from a 40-ply search but they do exist.

Regards,

Zenmastur

[Ras]

As for Stockfish says, Stockfish says: there have been quite some games against LC0 where Stockfish thought it was completely fine - only that it wasn't and found out too late. The interesting question is whether this game will lead to positions that Stockfish simply doesn't understand - and obviously, this cannot be answered with Stockfish analysis.

[zullil]

As for Stockfish says, Stockfish says: there have been quite some games against LC0 where Stockfish thought it was completely fine - only that it wasn't and found out too late. The interesting question is whether this game will lead to positions that Stockfish simply doesn't understand - and obviously, this cannot be answered with Stockfish analysis.

Can you provide one example at a time control like this one? What you say is certainly true, but is much more common at very short time controls.

[Harvey Williamson]

As for Stockfish says, Stockfish says: there have been quite some games against LC0 where Stockfish thought it was completely fine - only that it wasn't and found out too late. The interesting question is whether this game will lead to positions that Stockfish simply doesn't understand - and obviously, this cannot be answered with Stockfish analysis.

Can you provide one example at a time control like this one? What you say is certainly true, but is much more common at very short time controls.

I do use LC0 for analysis with 2 good graphics cards but I find that after 30 minutes or so the hash is full. Therefore at correspondence time controls I need to use other engines.

[zullil]

As for Stockfish says, Stockfish says: there have been quite some games against LC0 where Stockfish thought it was completely fine - only that it wasn't and found out too late. The interesting question is whether this game will lead to positions that Stockfish simply doesn't understand - and obviously, this cannot be answered with Stockfish analysis.

Can you provide one example at a time control like this one? What you say is certainly true, but is much more common at very short time controls.

I do use LC0 for analysis with 2 good graphics cards but I find that after 30 minutes or so the hash is full. Therefore at correspondence time controls I need to use other engines.

I have just one GPU but 128 GB RAM. So I can run longer than 30 minutes, but your point is a good one.

[Ras]

Can you provide one example at a time control like this one? What you say is certainly true, but is much more common at very short time controls.

Inn correspondence chess, players who follow Stockfish blindly are not too hard opponents for experienced correspondence players. At least one, maybe even both of the players have experience in correspondence chess. LC0 was just an example that some structural issues do exist even with Stockfish, as incredibly strong as it is.

[Dann Corbit]

A fast reply today as I expected this.
[pgn]1. g4 d5 2. g5 e5 3. d4 exd4 4. Nf3 c5 5. Bg2 Ne7 6. c3 dxc3 7. Nxc3 Nbc6 8. O-O d4[/pgn] and another conditional if 9. Ne4 Ng6

Still Stockfish self-play. Still hoping for something "novel".

I guess that anyone who has been analyzing for a long time used Stockfish. First, it has been available for a long time, and it is the strongest engine that is alpha-beta. LC0 may be stronger on the right hardware, but it is newer and I also guess that automated database improvement systems do not know how to deal with LC0 evals.

So I will be quite surprised if there is any variation from Stockfish lines, and if it does happen, I guess it will be an LC0 move.

[zullil]

Can you provide one example at a time control like this one? What you say is certainly true, but is much more common at very short time controls.

Inn correspondence chess, players who follow Stockfish blindly are not too hard opponents for experienced correspondence players. At least one, maybe even both of the players have experience in correspondence chess. LC0 was just an example that some structural issues do exist even with Stockfish, as incredibly strong as it is.

This is why I'm still following the game. I'm looking forward to seeing something better than simply looking at Stockfish-dev at depth 60. Especially something from White, who has frequently made what I consider "extraordinary" claims, none of which he has been able or willing to support.

[Harvey Williamson]

Can you provide one example at a time control like this one? What you say is certainly true, but is much more common at very short time controls.

Inn correspondence chess, players who follow Stockfish blindly are not too hard opponents for experienced correspondence players. At least one, maybe even both of the players have experience in correspondence chess. LC0 was just an example that some structural issues do exist even with Stockfish, as incredibly strong as it is.

This is why I'm still following the game. I'm looking forward to seeing something better than simply looking at Stockfish-dev at depth 60. Especially something from White, who has frequently made what I consider "extraordinary" claims, none of which he has been able or willing to support.

I have no idea yet if whites claims are wrong. My gut feeling is that they are. If I can prove that I don’t know. I have maintained a 2500+ iccf rating for several years and you can not do this by following one engine blindly.