chess: time limits and level of play

[bayu]

When I watched Carlsen and Karjakin play their championship, a winning-% calculated by some engine was part of the broadcast. It started at 55% for white at the beginning and then it would rise or fall depending on play. The commentators I was listening to considered a value between 40% and 60% usually as drawish. During their long games (they lasted about 6 hours if they didn't agree on a quick draw) the winning percentage would usually only nudge in different directions, the "volatility" was a lot higher when they played the rapid chess tie break (~1 hour per game).

I interpretated the higher volatility with more and more serious mistakes (well mistakes is probably the wrong word. "suboptimal move" is better). Is that correct? And can these engines estimate by how many Elo-points the level of play sunk?

[dfan]

Yeah, there are more mistakes at faster time controls, unsurprisingly. Computer analysis of the strength of played moves suggests that players' Elo strength drops around 250 points when playing rapid as opposed to classical, and more like 600 points when playing blitz.