The nice thing about this type of position is that thanks to XG and its speed we can get a little closer to the absolute win % than before. I can't wait for the next version of XG to be released so we can use truncated rollouts (XGR+) within' the rollout itself to gain additional accuracy. What I did below was two separate rollouts for each ply. (3ply, 3ply huge, 4ply, 4ply huge, 5ply, 5ply huge) This brings the variance down to ~.002 with only 5k trials.
You can see the huge jump in time spent on the rollout from 4 ply to 5 ply even though the 5 ply rollouts have less trials. You can also see for this specific position it seems there is no real increase in win % between 4 ply and 5 ply. The move filter oddly enough seems to make no difference or even have a negative impact when it changes from the default to huge.
I will now roll out the entire block of positions you were inquiring about on 4 ply and report back when they're finished.
|
| |||
| XGID=aBBBBCCA------------------:0:0:1:21:0:0:0:1:10 | ||||
 to play 21 | ||||
is Player 2| 1. | Rollout1 | 7/4 | eq: -0.839 | |||
| ||||||
| 1 15552 Games rolled with Variance Reduction. Dice Seed: 24983242 Moves: 3 ply | ||||||
| 1. | Rollout1 | 7/4 | eq: -0.842 | |||
| ||||||
| 1 15552 Games rolled with Variance Reduction. Dice Seed: 24983242 Moves: 3 ply Search interval: Huge | ||||||
| 1. | Rollout1 | 7/4 | eq: -0.831 | |||
| ||||||
| 1 15552 Games rolled with Variance Reduction. Dice Seed: 24983242 Moves: 4 ply | ||||||
| 1. | Rollout1 | 7/4 | eq: -0.832 | |||
| ||||||
| 1 15552 Games rolled with Variance Reduction. Dice Seed: 24983242 Moves: 4 ply Search interval: Huge | ||||||
| 1. | Rollout1 | 7/4 | eq: -0.831 | |||
| ||||||
| 1 7776 Games rolled with Variance Reduction. Dice Seed: 24983242 Moves: 5 ply | ||||||
| 1. | Rollout1 | 7/4 | eq: -0.831 | |||
| ||||||
| 1 5184 Games rolled with Variance Reduction. Dice Seed: 24983242 Moves: 5 ply Search interval: Huge | ||||||