I happened on an article:
about chess players and how their play has been improved pretty much across the board by playing against chess programs. I passed it along to a friend who is in to chess and here is what he had to say:
Yeah, computers have greatly changed chess not just at the GM level, but at every level. I have used PC analysis of my games to find improvements. The nice thing is that you can set the “blunder check” level so that the computer doesn’t inundate you with improvements you don’t understand. For example, I set the computer to find mistakes that cost me the equivalent of a pawn. Then, if I’m feeling energetic I might reset the blunder check to 1/2 pawn and do it again. Grandmasters will try to understand at the 1/100 of a pawn level!
The biggest change for GMs is really psychological, though. Younger players (those growing up with computer chess) are much more comfortable with and willing to enter “risky” positions. That is to say, the Russian legacy that drove chess for many years was to prefer solid positions with clear, even if minute, advantages. Complicated positions that require massive calculation were risky — overlook one branch and you could lose an otherwise drawn or won game. But computer chess doesn’t have a “risk analysis” function. The computer plays what it thinks is the best move without regard to risk. Players that have grown up with computer chess are psychologically more willing to take these risks because they have learned from computers that the “risky” move is the better move. Plus, they are very confident in the calculations because that skill is more required when playing computers.
One point from the article that was invalid was claiming modern competitors are better because their ratings are higher. Unfortunately, the mathematics of the rating system mean that it is inflationary. A 2800 today is more like a 2650 in the Fisher-Spassky era. The system wasn’t always inflationary, but sometime in the 80’s the system was changed a bit and the by product of that change was ratings inflation.
There are two points that I find interesting about all this.
First, that this feedback loop has occurred. I can remember when the Deep Blue matches were going on because they were widely covered and there was the ‘man against machine’ story to it. It seemed clear at that point that the programs were going to exceed the human capabilities. I never thought of or foresaw that this would get turned around and improve human play. It makes perfect sense; playing against someone who is better than you in any competitive endeavor typically makes you a better player. Expanding on that, when looking at most situations where programs take over something that people previously did, I’ve never much had the thought that there might be circumstances where this could lead to improvements in people performance. This may have been obvious but sometimes it has to strike you in a certain way.
Second, the idea that we are learning a different concept of risk from what can be thought of as a foreign mind is fascinating. I can imagine a scenario where the programs have a probability component (not saying they do, just that I can imagine it) so I suppose that can be thought of as risk but it occurs to me that this isn’t really accurate. From the program’s perspective, there is no risk to a particular play because the branch of play off of that move can be played with perfect accuracy. The question then for the person is how accurately can you make the play and the branch off it as the program would? The less accurate, the more risk. If you can make the play and the branch with perfect accuracy then presumably there would be no risk. I’m curious as to how possible that is given human limitations keeping in mind that there may be shortcuts that extend human capabilities.