Matthies/RubiChess

Bad move ordering at low plies

Matthies opened this issue · 0 comments

I did some statistics about how many moves need to be played in average in different plies:

2020-06-11 13:20:38,153<--1:(ST)   ply:  1  quiets-nonpv: ---   tactics-nonpv: ---  quiets-pv: 11.03  tactics-pv: 0.65
2020-06-11 13:20:38,155<--1:(ST)   ply:  2  quiets-nonpv: 0.93  tactics-nonpv: 0.12  quiets-pv: 11.56  tactics-pv: 0.86
2020-06-11 13:20:38,158<--1:(ST)   ply:  3  quiets-nonpv: 2.13  tactics-nonpv: 0.23  quiets-pv: 8.87  tactics-pv: 0.72
2020-06-11 13:20:38,160<--1:(ST)   ply:  4  quiets-nonpv: 0.99  tactics-nonpv: 0.15  quiets-pv: 9.51  tactics-pv: 0.74
2020-06-11 13:20:38,161<--1:(ST)   ply:  5  quiets-nonpv: 1.41  tactics-nonpv: 0.22  quiets-pv: 7.81  tactics-pv: 0.63
2020-06-11 13:20:38,164<--1:(ST)   ply:  6  quiets-nonpv: 0.92  tactics-nonpv: 0.18  quiets-pv: 7.74  tactics-pv: 0.66
2020-06-11 13:20:38,168<--1:(ST)   ply:  7  quiets-nonpv: 1.01  tactics-nonpv: 0.20  quiets-pv: 6.92  tactics-pv: 0.62
2020-06-11 13:20:38,171<--1:(ST)   ply:  8  quiets-nonpv: 0.84  tactics-nonpv: 0.19  quiets-pv: 6.74  tactics-pv: 0.63
2020-06-11 13:20:38,174<--1:(ST)   ply:  9  quiets-nonpv: 0.86  tactics-nonpv: 0.20  quiets-pv: 6.26  tactics-pv: 0.58
2020-06-11 13:20:38,176<--1:(ST)   ply: 10  quiets-nonpv: 0.76  tactics-nonpv: 0.20  quiets-pv: 5.77  tactics-pv: 0.59
2020-06-11 13:20:38,177<--1:(ST)   ply: 11  quiets-nonpv: 0.78  tactics-nonpv: 0.19  quiets-pv: 5.46  tactics-pv: 0.52
2020-06-11 13:20:38,180<--1:(ST)   ply: 12  quiets-nonpv: 0.71  tactics-nonpv: 0.19  quiets-pv: 5.00  tactics-pv: 0.58
2020-06-11 13:20:38,183<--1:(ST)   ply: 13  quiets-nonpv: 0.72  tactics-nonpv: 0.18  quiets-pv: 4.69  tactics-pv: 0.49
2020-06-11 13:20:38,186<--1:(ST)   ply: 14  quiets-nonpv: 0.67  tactics-nonpv: 0.18  quiets-pv: 4.28  tactics-pv: 0.53
2020-06-11 13:20:38,188<--1:(ST)   ply: 15  quiets-nonpv: 0.68  tactics-nonpv: 0.17  quiets-pv: 3.97  tactics-pv: 0.45
2020-06-11 13:20:38,191<--1:(ST)   ply: 16  quiets-nonpv: 0.64  tactics-nonpv: 0.17  quiets-pv: 3.74  tactics-pv: 0.48
2020-06-11 13:20:38,193<--1:(ST)   ply: 17  quiets-nonpv: 0.65  tactics-nonpv: 0.16  quiets-pv: 3.44  tactics-pv: 0.43
2020-06-11 13:20:38,194<--1:(ST)   ply: 18  quiets-nonpv: 0.62  tactics-nonpv: 0.15  quiets-pv: 3.34  tactics-pv: 0.43
2020-06-11 13:20:38,197<--1:(ST)   ply: 19  quiets-nonpv: 0.62  tactics-nonpv: 0.15  quiets-pv: 2.97  tactics-pv: 0.38

Shows that move ordering in low plies especially for quiet moves and in odd plies doesn't work very well.
Odd/even difference may have to do with cut nodes.