| |
BGonline.org Forums
How strong is this effect?
Posted By: Jim Stutz In Response To: In response to Timothy and Bob (Bob Koca)
Date: Tuesday, 27 June 2023, at 6:34 p.m.
I have no data to cite, as Timothy noted. With that in mind, here’s the way I see it:
A dead-cube tactical position like this one (ie: the Favorite likely ends up with the majority of the tough decisions) that goes from a ~1.04 pass at XG 4-ply to a 1.08 pass on rollout is indeed a practical take, all else equal, in a match between two human players who play their checkers to a 5 PR on average. It’s a solid take, in fact.
The gap in tactical playing strength between XG 4-ply and a good rollout (as measured by average PR, which is an uncannily accurate metric for playing strength, IMO, despite its limitations) is, what? -- maybe ~1 PR point across a thousand random matches? Meanwhile, the gap in playing strength between a human 5-PR checker player and a good rollout is ~5 times greater – and that’s across all classes of positions, complex and simple alike. If a tactical position is one-sided and sufficiently complex that a rollout’s performance on the side of the Favorite exceeds 4-ply’s performance by half-a-blunder’s worth of equity, then that same rollout is going to thoroughly trounce a human 5-PR player in a PR contest playing that same position out 1000 times. Notwithstanding Timothy’s valid point that PR differences are not linearly correlated with MWCs, the performance (playing strength) gap is so huge between 5-PR humans and bots in these tactically complex positions that MWCs are bound to diverge dramatically from the expectations indicated by bot-on-bot evals.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.