Define "Best Move" - Chess Forums - Page 3

Ahptoemiz · 2018-10-18T14:41:35-07:00

Assuming it’s true that for any given position there is an objectively “best” move (or alternatively a co-equal set of best moves) then how do we define this in all instances? In some instances defining best is trivial. If only one move delivers mate in one and all other moves draw then, obviously,

kindaspongey

Oct 20, 2018

0

#41

Ahptoemiz wrote:

… If I thought one side in a position might be able to force a win but I just couldn’t calculate it out to the end would I rate the position as "="? Probably not. I'd probably rate the side I suspected of having a win as having an advantage. ...

Again, I think that is a departure from the usual usage. "+-" and "-+" have been used for a "decisive" advantage without any regard (as far as I can tell) to whether or not calculation-verification is possible. Other notations are used for advantages that are not perceived as "decisive".

kindaspongey

Oct 20, 2018

0

#42

Ahptoemiz wrote:

… Someone might says equal means “equal chances” not drawn, but I’d personally fault an annotator for subjectivity who would assess a position like that. It's essentially saying “equal” based on the human equation and likely sub-optimal play by one side resulting in equal practical chances. …

We do not have access to anything other than potentially sub-optimal play in the vast majority of positions. Consequently, it strikes me as completely appropriate (and common practice) for annotators to comment with potentially sub-optimal play in mind.

kindaspongey

Oct 21, 2018

0

#43

Ahptoemiz wrote:

… It's essentially saying “equal” based on the human equation and likely sub-optimal play by one side resulting in equal practical chances. If we go down that road we’re going to get ridiculous pretty quick. We’re going to have to consider every evaluation based on practical considerations, like the strength of the players or time period in history. Eg: “oh the Steinitz Gambit is -/+ now at the top level but was = in Steinitz day and is +/= at the club level these days except on a Tuesday…”

We have been going down the road of usage for +-, etc. for decades. If you want to argue that that road has been a mistake, you might want to think about the degree to which the chess world has shown signs of perceiving a ridiculous result. I don't think there is any question that evaluations have changed over time (and might be different for club players), and, as far as I can tell, the chess world is okay with that. For the most part, we do not have access to absolute truth and usual communication is not about that.

Daniel1115

Oct 21, 2018

0

#44

stiggling wrote:

It's not circular reasoning so... I don't know what to tell you.

Well it is. The evaluation is derived based on the best moves. You claim that the best moves is the move that best preserves the evaluation. You cant determine the evaluation without the best moves, so you are using circular logic.

WSama

Oct 21, 2018

0

#45

In my opinion there are two main factors regarding the 'best move':

First of all the evaluation of 'best move' should be constrained to a 'measurable' line of moves/play, because the number of variations to tread on can blow up exponentially if we don't watch it.
Second, and most close to the first, time. The first point holds an immediate effect on the time used on any evaluation, and vice versa. Any complex position should be allocated a limited time allowance or else we could spend years trying to find the best move. For example, it can be argued that one could begin looking for the winning move from the very first move in a game, but this is crazy, so we segment the game.

So, only with these two points (and perhaps more) is it feasible, or realistic, to calculate the 'best move'. In this case the best move itself being one that evaluates to the best advantage under the constraints of a certain amount of moves and time.

We use our understanding of the game, of chess, to segment the game more accordingly, so that we may apply the best methods of evaluation. The key word here being 'method'. These change over time.

What if GM's knew the perfect games but just play along, ahahaha...

stiggling

Oct 21, 2018

0

#46

Daniel1115 wrote:

stiggling wrote:

It's not circular reasoning so... I don't know what to tell you.

Well it is. The evaluation is derived based on the best moves. You claim that the best moves is the move that best preserves the evaluation. You cant determine the evaluation without the best moves, so you are using circular logic.

When I say "true" evaluation I mean e.g. endgame tablebases. That's just a collection of every possible position. The "best" move is just selecting a branch that contains at least one draw for each of your future moves (or one win if you were winning). A mistake is selecting a branch that contains no draws (or no wins if you were winning) for your moves.

Anyway, even if we use a regular computer algorithm, it's not circular reasoning. The best move is the one that gives the best evaluation. The best evaluation is based on things like material, space, pawn structure, and blah blah blah. Yes the best moves result in the best evaluation, but as long as we're not defining them that way it's not circular reasoning.

Imagine I said the best way to save money is to buy products at the lowest price, and you say that's circular because there's no way to know what's best without knowing which costs less (??)

Ahptoemiz

Oct 21, 2018

0

#47

najdorf96 wrote:

With a Rubik's cube, when trying to solve it for the first time (yes, this is my "big" analogy) it seems impossible: any which way you turn it to solve one side jumbles up the other sides; you match 4 sides leaves you in a bad position to finally solve the last two sides. Obviously, there was a book written long ago to solve it however it was mixed up. It involved looking for a certain pattern on each turn. The "best" move. If you didn't follow that path, you'd likely come up short. In conclusion, yeah we're still chasing the "best" move in a game, but in my opinion, they are a plethora of known positions (ie the two bishop sac, rook exchange sac on Nc3, the infamous Bxh6 etc) that guides us often times to it. Nuances in the pawn structure, playing against an isolated pawn, overloading a certain point, weaknesses in color complexes, prophylaxis blah blah are all there. We just have to develop the ability to recognize them. Human subjectivity. Indeed.

Thanks for the insight.

Ahptoemiz

Oct 21, 2018

0

#48

kindaspongey wrote:

Ahptoemiz wrote:

… I'd say the key distinction between "drawn" on the one hand and "even" or "equal" on the other is that the former reflects an assessment based (usually) on calculation and the latter two an assessment based on intuition. But otherwise they refer to the same thing. ...

I think that is a huge departure from the usual practice. One often sees the advice to be content to play the opening seeking nothing more than a playable middlegame - the whole idea being to be content to achieve an equal position where there is still a very real potential for one side to go wrong. A position seems to me to be routinely referred to as drawn if there is perceived to be little chance for an obscure forced win and little chance for one side to go wrong while playing with moderate competence. For example, one would say that the vast majority of king and two knight against king positions are drawn. The vast majority of evaluations (of any kind) seem to me to be based on intuition. For most of chess history, hasn't something as simple as queen and king against queen and king been beyond the powers of calculation?

I don't think what I'm saying is really incompatible with "usual practice". I'm not saying "equal" and "drawn" mean the same thing in all respects, just that they mean the same thing with respect to the evaluation of a position in terms of the expected outcome with precise play on both sides.

And I'd have to disagree with you about N+N+KvK or Q+KvK. I don't think, even going a long ways back, theory assessed these positions based primarily on intuition. I think analysts started with an end mating position and then tried to figure out if they could engineer getting there by force. If they could, it was considered a win. If they couldn't, it was considered a draw. I'm no historian of endgame theory (that's for sure) but endgames have that special advantage that retrograde analysis is viable in a way not typical of middle games or openings

Ahptoemiz

Oct 21, 2018

0

#49

kindaspongey wrote:

Ahptoemiz wrote:

… Someone might says equal means “equal chances” not drawn, but I’d personally fault an annotator for subjectivity who would assess a position like that. It's essentially saying “equal” based on the human equation and likely sub-optimal play by one side resulting in equal practical chances. …

We do not have access to anything other than potentially sub-optimal play in the vast majority of positions. Consequently, it strikes me as completely appropriate (and common practice) for annotators to comment with potentially sub-optimal play in mind.

We don't have access to anything other than sub-optimal play, but I think what (most) annotators try to do is evaluate the position based on their best guess of what the result would be from optimal play. That's the intuition part.

I mean if I annotate a position on move five as =, I mean by that it's my intuition that the 32 piece tablebase with show "drawn in 102", or something similar

Ahptoemiz

Oct 21, 2018

0

#50

RedGirlZ wrote:

You're having a misconception about this. The "best" move isn't objective. It's subjective, even with the computer. Computers haven't solved chess, so it's impossible for them to know what the OBJECTIVE, best move is. They can only go off how they've been programmed, which move least damages their numerical advantage over the opponent, which move best improves their advantage and position. In terms of mate in 1 and 3, Mate in 1 is considered better simply because it completes the game quicker, which is what the engine has been programmed to do, "FINISH THE GAME" "WIN". This actually happened to me some games ago where I mated someone in 3 moves, but I missed a mate in one, and the engine said I still did a good move, but the best move (In green), was a different one. In terms of the engines analysis, a mate in 1 is better than a mate in 3. Subjectively, to us it makes no difference because in both scenarios a win is achieved.

Similarly in drawn positions, the engine defines the best move by a move that doesn't worsen your position, improves it etc. It's very rarely an OBJECTIVELY best move, so rare in fact the OBJECTIVELY best move in many positions is probably rarely played, unless the moves are very obvious.

Change it to subjective, and your answer becomes clear. You can't talk about objectively best moves per position until the game has been solved.

But doesn't your answer actually assume what you're trying to disprove? You say computers haven't solved chess therefore moves are subjective. But you're not saying chess can't be solved (to some degree you seem to be implying it can). And if it can be solved then isnt there is a best move (set of moves) for any position? What's lacking is not the existence of best moves, it's our knowledge. Saying otherwise is akin to claiming that whether the world is flat or round was factually subjective until the Greeks proved which.

Ahptoemiz

Oct 21, 2018

0

#51

@RedGirlZ... My reading of what you said was since we can't know what the objective best move is then the best move is subjective. I'm just calling that out and saying it's not true. The best move is the objective best move regardless of whether we know what it is or not. Remember, you specifically said "The best move isn't objective, It's subjective".

kindaspongey

Oct 21, 2018

0

#52

Ahptoemiz wrote:

kindaspongey wrote:

Ahptoemiz wrote:

… I'd say the key distinction between "drawn" on the one hand and "even" or "equal" on the other is that the former reflects an assessment based (usually) on calculation and the latter two an assessment based on intuition. But otherwise they refer to the same thing. ...

I think that is a huge departure from the usual practice. One often sees the advice to be content to play the opening seeking nothing more than a playable middlegame - the whole idea being to be content to achieve an equal position where there is still a very real potential for one side to go wrong. A position seems to me to be routinely referred to as drawn if there is perceived to be little chance for an obscure forced win and little chance for one side to go wrong while playing with moderate competence. For example, one would say that the vast majority of king and two knight against king positions are drawn. The vast majority of evaluations (of any kind) seem to me to be based on intuition. For most of chess history, hasn't something as simple as queen and king against queen and king been beyond the powers of calculation?

I don't think what I'm saying is really incompatible with "usual practice". I'm not saying "equal" and "drawn" mean the same thing in all respects, just that they mean the same thing with respect to the evaluation of a position in terms of the expected outcome with precise play on both sides. ...

I see no reason to believe that annotators are attempting to evaluate a position in terms of the expected outcome with precise play on both sides. As far as I can tell, expressions like "stands slightly better" are explicit references to the real-world difficulty that an opponent might have playing precisely.

kindaspongey

Oct 21, 2018

0

#53

Ahptoemiz wrote:

… And I'd have to disagree with you about N+N+KvK or …

My assertion about N+N+KvK was that "one would say that the vast majority of king and two knight against king positions are drawn". What objection do you have to that assertion?

kindaspongey

Oct 21, 2018

0

#54

Ahptoemiz wrote:

... Q+KvK. I don't think, even going a long ways back, theory assessed these positions based primarily on intuition. I think analysts started with an end mating position and then tried to figure out if they could engineer getting there by force. If they could, it was considered a win. If they couldn't, it was considered a draw. I'm no historian of endgame theory (that's for sure) but endgames have that special advantage that retrograde analysis is viable in a way not typical of middle games or openings

My assertion was about "queen and king against queen and king". What end positions are appropriately considered and, without a computer, how does calculation-without-intuition identify them?

Ahptoemiz

Oct 21, 2018

0

#55

@kindaspongey Don't forget ZZZ ;-)

kindaspongey

Oct 21, 2018

0

#56

Ahptoemiz wrote:

... I think what (most) annotators try to do is evaluate the position based on their best guess of what the result would be from optimal play. That's the intuition part.

I mean if I annotate a position on move five as =, I mean by that it's my intuition that the 32 piece tablebase with show "drawn in 102", or something similar

Again, as far as I can tell, symbols like the stands-slightly-better symbol are used as explicit references to the real-world difficulty that an opponent might have playing optimally even if the position is drawn with optimal play.

kindaspongey

Oct 21, 2018

0

#57

Ahptoemiz wrote:

... if [chess] can be solved then isnt there is a best move (set of moves) for any position? …

As near as I can tell, "best" is not a reference to what would come from a computer solution of chess. It seems to me to be a reference to what would give one the best chances of winning against a human opponent. There can be many drawing moves in a position, but some will give the opponent more difficulty than others. Put a Black king on f8, a White king on g3, and a White pawn on d4. Would you rather play Kh2 or Kf4 ?

haroldschris

Oct 21, 2018

0

#58

Ahptoemiz wrote:

IMBacon wrote:

You expect someone to read all that?

I expect SOMEONE to, just not you obviously.

Good comeback!

ARCS2016

Oct 23, 2018

0

#59

kindaspongey wrote:

ARCS2016 wrote:

I think that the objectively best move in a deterministic game like chess would be the move that has the most likely chance of ending in a victory or draw. You could theoretically, from any position, map every possible move and use all of the endings to come up with a probability of victory or draw. However, most engines take a shortcut and only anticipate probably moves. Does this answer your question? I am willing to debate

It seems to me that there are two problems with this sort of thinking. (1) Is it useful to have a definition that can only be "theoretically" applied? (2) In order to "come up with a probability", what could one do other than (a) make the doubtful assumption that all moves are equally likely, or (b) use human behavior to assign probabilities to moves? 2b would result in probabilities that vary from one player to another.

1) Yes. In this instance, theoretical means that while it is not practical to do so, it is possible.

2a) In a deterministic game, this is usually the best way to determine the probability of victory. However, is a fallible method if you are not playing the side that has a 'winning strategy'.

2b) As described above if you are playing the side with the 'winning strategy', you do not need to factor in human behavior. If you are playing the other side, a winning strategy does require personalization to the player

kindaspongey

Oct 23, 2018

0

#60

ARCS2016 wrote:

kindaspongey wrote:

ARCS2016 wrote:

I think that the objectively best move in a deterministic game like chess would be the move that has the most likely chance of ending in a victory or draw. You could theoretically, from any position, map every possible move and use all of the endings to come up with a probability of victory or draw. However, most engines take a shortcut and only anticipate probably moves. Does this answer your question? I am willing to debate

... Is it useful to have a definition that can only be "theoretically" applied? ...

1) Yes. In this instance, theoretical means that while it is not practical to do so, it is possible. …

Do you have any reason to believe that it is currently "possible" to "map every possible move" in chess?