Everyone on mersenneforum vs. Stockfish

Critique requested.

I'm considering the following strategem: starting from a desired endgame structure where Stockfish is mated,
working in reverse, I want to create specific intermediate game structures that are separated by a plie depth
just beyond Stockfish's capability relative to the time it has to `think.`
My intent is to bait Stockfish with moves leading to these intermediate structures.

To do: Quantify Stockfish's move thresholds, that is, discern the attribute(s) sought and acquired by Stockfish
as measures of control/influence.
Test understanding of Stockfish's protocols by making (hopefully non-suicidal) moves and predicting Stockfish's
response relative any numerical patterns observed.

Stockfish must measure my strength/weakness vs its strength/weakness at every plie. I would like to lure it into
a succession of weakened states, patiently, leading inevitably to a mate. To do this, I believe, requires signalling my
(feigned) weakness at the end of Stockfish's search tree capacity.


Does the above sound reasonable or unreasonable? It's bare bones and I only have heuristics and gut instinct to work with but I would like to intelligently assess what I am looking at when analyzing a game.

[gd_barnes]

Originally Posted by jwaltos

Critique requested.

I'm considering the following strategem: starting from a desired endgame structure where Stockfish is mated,
working in reverse, I want to create specific intermediate game structures that are separated by a plie depth
just beyond Stockfish's capability relative to the time it has to `think.`
My intent is to bait Stockfish with moves leading to these intermediate structures.

Suggestion: To look for its weaknesses start with a google search:
https://www.google.com/?gws_rd=ssl#q...ish+weaknesses

Does the above sound reasonable or unreasonable? It's bare bones and I only have heuristics and gut instinct to work with but I would like to intelligently assess what I am looking at when analyzing a game.

It is very unreasonable. You cannot use bare bone heristics and gut instinct to reverse engineer a program with such complexity that sees so far ahead in any reasonable amount of time. Attempting to do what you are suggesting reminds me of a guy in another thread that said he had "formulas" that could find 1 billion digit primes.

You obviously have a lot of programming and math background but...can we ask...what is your chess playing background?

Suggestion: Download Stockfish and its sourcecode. Here:
https://stockfishchess.org/download/

Note that this is only an engine so you'll also need to download a GUI interface. I personally like Arena since it's free and its engine is strong in its own right. There's also Fritz (you'd have to buy it) and several others that accept multiple engines.

I would suggest taking a look at its source and see if you can find weaknesses that we can exploit. (Good luck!)

Two things to keep in mind:
1. Stockfish is open source. It is continually being improved by the programming community at large.
2. Because of #1 Stockfish is one of the strongest chess engines in the world and it is much better than the best human grandmasters. (Yes even better than world champion Magnus Carlsen.)

When MooMoo started this effort, I downloaded Stockfish and played around with it. Its thinking speed is remarkable. On many middle-game moves it is able to look 25-35 plies ahead with just a few minutes of thinking. In the late middle-game and end game it is frequently looking 40+ plies ahead.

If there is a main weakness of Chess software, it is long-term strategy and very closed positions. These are extraordinarily difficult concepts for programs to overcome. I think you'll see this in some of the "Stockfish weaknesses" links.

[Brian-E]

I think we need to discuss the position. Every move. And when we come to vote we should take account of what people have said before marking our choice in the poll.

I think that has already gone wrong actually. The first move for White can arguably be categorised in one of three sets: {1.e4}, {1.d4, 1.c4, 1.Nf3, 1.g3, and possibly 1.b3}, {The other 14 or 15 legal moves}. The moves in the second set very often transpose into the same openings where the move order doesn't matter. On that basis the second set of moves won, but because the first set of White openings are characterised by the single opening move 1.e4 that move was selected. If we had discussed the move first, those in favour of the second set could have agreed on one of the moves within it. This problem will frequently occur in later positions which could lead to a tempting but inferior move being selected simply because there are multiple good moves available and the good players divide their votes amongst those good moves. (I'm not saying that 1.e4 is "tempting but inferior", of course, but I hope everyone can see how this problem can occur later!)

Therefore we need to discuss the moves and try to come to an agreement even though our subsequent voting will be free. See it as a sort of "political lobbying" as if amongst parliamentary representatives. :-)

So let's talk about move 2. The most frequent move against the French 1.e4 e6 is 2.d4. We attempt to set up a strong pawn centre. There are other good moves, though. 2.d3, 2.g3, 2.c4, 2.Nf3, even 2.b3, all more modest systems, the first two going for the "King's Indian Attack" with fianchetto of the f1 bishop.

What do people advocate?

[LaurV]

Originally Posted by Brian-E

I think we need to discuss the position. Every move. And when we come to vote we should take account of what people have said before marking our choice in the poll.

True. And fully agree with the rest of your post. From 365chess.com, both the 2.d4 and 2.d3 are good continuation. The rest are either giving too much statistic advantage to black, or were played too less for the numbers to be significant. I will look around the web this evening, to see what other chess sites say.

[axn]

Originally Posted by Brian-E

I think we need to discuss the position. Every move. And when we come to vote we should take account of what people have said before marking our choice in the poll.

Yes. 3-4 days of discussions. Then put up the viable candidates for 2-3 day poll. Although... not really sure whether that much preparation is necessary for opening phase.

EDIT:- Perhaps, create a new threads for discussing new moves?

Thank you GD. I will follow through on your suggestions. I provided some of my chess history but I essentially play using native intelligence - no concerted study of the game. There are few people available for me to play. Gnuchess and Fritz I like - Fritz provides an ELO rating so I have a benchmark to work from. My math and programming skills, the same. I study what I like, test my comprehension and every so often step into a cow pie.
And thank you Moo for initiating this thread.
Am in complete agreement with Brian that cooperation is imperative otherwise we are working inefficiently and at cross-purposes.

[Brian-E]

Well, I agree with LaurV that 2.d4 and 2.d3 are both good continuations. But my preference goes to 2.d4 with a view to playing the Tarrasch variation (1.e4 e6 2.d4 d5 3.Nd2). I believe this opening guarantees White a small but persistent edge which should hopefully deny the computer the dangerous counterplay which Black can drum up in some of the sharper variations of the French.

[science_man_88]

Originally Posted by Brian-E

Well, I agree with LaurV that 2.d4 and 2.d3 are both good continuations. But my preference goes to 2.d4 with a view to playing the Tarrasch variation (1.e4 e6 2.d4 d5 3.Nd2). I believe this opening guarantees White a small but persistent edge which should hopefully deny the computer the dangerous counterplay which Black can drum up in some of the sharper variations of the French.

looking at those two I'd also go with d4 simply because of central development and it keeps the king side bishop with a line of play

[gd_barnes]

d4 definitely.

[WMHalsdorf]

d4 since the other moves tend to favor the computer's strengths.

d4 - (bandwagon jumper).