# Are computers closing in on solving chess?

The title may be a little extreme, but I got very curious about the concept. I chose to use ICCF games to get a solid pool of computer-enhanced game results. The arrival of multi-core computers and ever-stronger software like Rybka and Houdini have definitely made an impact on the outcome of correspondence chess games. Consider the following data:

Draw percentages by year on ICCF:

Year - games - draws - white wins - black wins - white performance

2006 - 35790 - 39% - 35% - 26% - 55%
2007 - 39152 - 44% - 33% - 23% - 55%
2008 - 42735 - 45% - 32% - 23% - 55%
2009 - 41350 - 47% - 31% - 22% - 55%
2010 - 39430 - 47% - 31% - 22% - 54%
2011 - 12535 - 49% - 29% - 22% - 54%

As you can see, just in the past five years computers have increased the drawing percentage by an astounding 10%! White's expected performance went from an upper 55% to a near flat 54% in that time. Also note how 2007's arrival of multi-core desktop PCs jumped the drawing rate by 5% on the spot.

I also have determined much of white's (and black's) wins come from playing risky openings rather that 'safe' ones. An example would be the Sicilian offering black maximum risk-reward against 1.e4, while the Petroff affords black a rather dull but much safer chance to draw at the very least.

My own conclusion is that ultimately white cannot win if black plays for minimum risk in the opening.

• 17 months ago

We may have to use the methods used for Checkers to put some life back into the game.  At the higher levels of course.

• 18 months ago

Not solve in the sense of the entire game being calculated out to the end, but I do believe the way the current trend is going on ICCF, computers won't need to. Simply using an opening book based on how the ICCF games are playing out will eventually be enough to always ensure at least a draw. Once you're ensured of at least a draw, then that becomes a way of solving chess without having to calculate it.

• 18 months ago

I don't think computers will ever solve chess.