Is the best move always the 'perfect' move?

Is the best move always the 'perfect' move?

Dec 15, 2009, 10:18 AM |

I will try to argue here that the "perfect" move is not always the "best" move.  To make this case, I will first have to articulate the relevant distinctions and provide examples.  These examples will serve as counter-examples and, so, as anomalies to the traditional theories and strategies on chess that presuppose that the perfect move is always the best move. 

These thoughts can be considered as a response to the suggestion attributed to Steinitz that links the idea of perfect play as crucial to chess strategy.  For more on this, see Kurt Godden's most recent blog on the subject and the ensuing comments below it. 

The Perfect Play Conjecture is the idea that, if both sides play perfectly (from the beginning), then the outcome of a game will be a draw.  In any given chess position, the game would be a forced win for one side or a forced draw. So even though Rybka might say (0.49) advantage for white, or a chess GM in a book might say "slight edge for white", the fact is that any and every chess position is actually either a forced win or a forced draw (even though, given our limited knowledge, we don't usually know which).  The 'perfect' move is the one consistent with attaining the forced win (or draw).  

What I am interested in here is the idea that chess theory and strategy have traditionally been geared towards this conception of perfect play with the unstated assumption that the best move for a human to make in any given position is the "perfect" move.  But what if that's wrong?  What if, sometimes at least, the best move is not the 'perfect' move?  If the best move to make, at least sometimes, is not the 'perfect' move, and if a given chess theory and strategy always recommends (and is designed to endorse) only the 'perfect' move, then those theories and strategies must necessarily be inadequate to the extent to which they fail to endorse the best move on those occasions where the best move is not the perfect one.

Let me provide examples of anomalies to perfect play that fall into four categories:


  • Time-management
  • Stylistic issues
  • Playing against a computer
  • Psychology and simplification


If White has a significant time advantage over Black and has a choice to make the 'perfect' move that trades down material and leads to a draw or an imperfect move that maintains the action and forces Black to spend a lot more time thinking for the rest of the game, then the best move could be one that prolongs the game, prolongs the tension, prolongs Black's confusion, and helps white win on time or helps provoke Black to blunder based on his time pressure.  In those situations, the best move can sometimes be an imperfect move.  In fact, if playing perfectly leads to a draw, and you want to win, and have earned a significant time advantage, it would obviously be silly for you to throw away the game by playing moves that quickly lead to a draw when all you have to do is prolong the game to win on time (or win via the errors caused by your opponent playing rushed chess).   In the specific context described above, the best move, at times, is not the perfect move.


Suppose you know your opponent is an expert in the Petroff and you know that you are an expert in the Cochrane Gambit (Nxf7 vs the Petroff).  Let's also suppose that the Cochrane gambit is not the 'perfect' play move (maybe it is, who knows, but for the sake of the example, it doesn't matter).  Even if the Cochrane Gambit is not the perfect way to play, it might be the best move for you to make given your preferred style of play vs your opponent's style of play. Regardless of if you win or lose, you give yourself better fighting chances by playing in the style that suits you, especially if it also a style you know to be uncomfortable for your opponent.  A similar point can be made when faced with a choice to keep a game open or closed.  Players have strengths and weaknesses, and it is common sense good strategy to try to play into positions that maximize your strengths and minimize your weaknesses while maximizing the weaknesses and minimizing the strengths of your opponent.  So in some of those kinds of situations, it is possible for the best move to be the imperfect move.


Related to the point above--it is well known that computers excel at tactical situations and can have trouble dealing with the long-term strategy of a closed position bereft of tactical possibilities.  In that case, a GM would be best advised to play in an anti-computer style that may very well include moves that, though not 'perfect' according to perfect play theory, but optimal or 'best' in the sense that they are the moves that are most likely to secure advantage.


Sometimes it works to your advantage to keep a position complicated and avoid simplifications to give your opponent the opportunity to make mistakes.  You can do this by invoking time-pressure, as mentioned above, but even when time pressure is not a factor, it is good to make moves that accurately gauge your opponent's skill.  You don't want to make a flat out cheapo attempt that, if unsuccessful will surely cost you the game, but sometimes, when given a choice between the perfect move and the best move, it might be that the best move is the one that, though not perfect, gives you the best chance to win in a given situation.  Part of chess is creating problems that you're opponent can't solve, and if you correctly size up the skill of your opponent in a given situation, then you can exploit that fact in making the imperfect move.  So a move you might make against a particular opponent is not the move you'd make against a GM or against a computer.  The best move in each case could well be very different in the exact same position.  

Each anomaly above takes into consideration context-specific features of the game, whereas any strategy based on always trying to find and play the 'perfect' move would necessarily recommend the sub-optimal move, on at least some occasions.  The reason should be clear; perfect play exists in a sort of idealized vacuum that disregards the very real context-specific features of chess games.  Real chess games involve psychology, time-pressure, and stylistic preferences, among other context-specific aspects of any particular game.  A chess theory based on endorsing always and only the 'perfect' move would necessarily have to disregard all context-specific features that would lead a real person to prefer to play the best or optimal move (the one most likely to help a person win a game or draw in an unfavorable position).  So now we see the upshot of the discussion on anomalies to perfect play strategy; a complete and accurate strategy should, therefore, include the context-specific features of chess and, in so doing, would recommend players make the best move and not the perfect one.  Therefore, a complete and accurate chess strategy must reject the thesis that one should always make the 'perfect' move.  The perfect move can be, but need not always be, the best move.