Tuesday, May 9, 2017

Annotations part 2

After one of my last games my opponent polity refused my invitation for a post-mortem. He didn't consider it useful and preferred to drink a beer quietly at the bar. Engines are today much stronger than any player so why wasting time at some lousy analysis. There is definitely some truth in this as with some simple clicks you can generate automatically an analysis which is many times more accurate. Besides in part 1 I advertised a method of annotating completely based upon evaluations of the engines.

The recent Penrose Chess Institute Puzzle demonstrates clearly the dangers of blindly trusting these evaluations. Engines show a winning score for black while any experienced player easily sees it is just a draw. One and a half year ago I wrote on this blog about computers achieve autonomy but this doesn't mean that we can't play any role anymore. The doom-scenario described in the recent article at chess.com "is this the future of chess" is just ballyhoo.

1 example of some fabricated position not looking close to any normal position in standard play doesn't refute the absolute dominance of the engines. Therefore some only consider positions from serious games relevant to judge about the supremacy of the computer. Do such positions exist which we as human can access quicker and more accurate than the current engines? If yes which ones?

In my article fortresses I already covered some positions of which we can prove that the computer-evaluations are inaccurate or even plainly wrong. However humans won't do necessarily better without using any tools. Nonetheless there exist some exceptions where we are stronger than engines. 1 group of endgames, opposite bishops stands out. An experienced player can often very quickly access correctly such position. In below position the engine is not eager to exchange the queens but Robert correctly values the endgame as harmless.

In the final position the engine still gives a small edge for me but I was already for a long time convinced this is a dead draw. Another recent example is shown below. Again the engine calculates the position as better for white as black loses the c7 pawn. White still could continue instead of repeating moves but the draw is not hard to achieve of course.

In both examples I consider it stupid to stick meticulously to my method of annotating. I exceptionally deviated from the evaluation of the engines and replaced them by my personal more accurate judgement.

In a recently played endgame I took it a step further in my annotations. Only a handful pawns are on one side of the board. The computer makes a complete mess when evaluating the played moves. Some moves are considered weak while there is nothing wrong. Others aren't annotated while there are clearly better ones. The original annotations linked to the evaluations of the engine can be found below.

After swapping all this with my personal more accurate evaluations we get a very different image of the endgame. I assume this also much better matches our intuition of such type of endgame.

This endgame shows there is also often a clear difference between understanding and actual play. We are still much more prone to blunders especially when we are running out of time.

Objectivity/ searching the truth still get absolute priority in my analysis. Engine-evaluations are used intensively but it is still good not to ignore your own chess-knowledge.


No comments:

Post a Comment