10

I've downloaded and parsed the king-base chess database (about 1.5 million games) and am looking at distribution of moves for games. When plotting the distribution for different outcomes, I see spikes at 80, 100 and 120 moves (and smaller spikes for draws at 20, 40 and 60 moves).

enter image description here

Why are these spikes occurring? Are people more likely to resign at these round numbers? Or are there rules in professional games that mean games are more likely to end after multiples of 20 moves?

3
  • 4
    Are these half-moves or moves (a move by both sides)? I'd expect a spike at 40 full moves. Apr 7, 2017 at 11:45
  • @RemcoGerlich I grouped white and black moves into "move-pairs" to smooth the data. Otherwise the data is very spiky. For example White is much more likely to win on an even move (white gives checkmate) rather than an odd move (black resigns). Apr 7, 2017 at 12:45
  • @RemcoGerlich - ah sorry, these are half-moves Apr 7, 2017 at 13:15

1 Answer 1

8

This is due to time control:

the World Chess Federation FIDE sets a single classical time control for all major FIDE events, according to the handbook: 90 minutes for the first 40 moves, followed by 30 minutes for the rest of the game, with an addition of 30 seconds per move

Other time controls are used, but are usually based on multiples of 20 moves.

Mikhail Botvinnik, a former world champion, once said:

The amount of errors made at time control are legion

So gross errors accumulate around the 20th, 40th, 60th etc. move, and so would resignations, loss on time and draw offers.

2
  • 7
    And the peak for draws after 30 moves is likely due to Sofia rules (In some tournaments no draw offers are allowed before move 30). Apr 7, 2017 at 12:20
  • 3
    It's not just that errors are more likely to be made. It makes no sense to resign in time trouble -- you might miss a way to save the game. Better to wait until the time control, and resign at leisure.
    – TonyK
    Apr 7, 2017 at 12:55

Not the answer you're looking for? Browse other questions tagged or ask your own question.