Here's this for what it's worth. I'm running at least 3 separate rollouts of an opening 21 for no reason other than to see them and get familiar with certain things. I'm as of now unable to export as HTML. This is a 3 ply Red rollout which is not the strongest 3 ply setting there is. There is also what seems to be a 3 ply step above this called simply *3 ply*. EXtreme starts at 1 ply, like Snowie. You can see that this particular rollout was a lot faster than either bot would be but I'm not impressed with the results or the amount of variance that still seems to be left. This was performed with variance reduction on.
The regular 3 ply rollout is taking much longer but also showing what I perceive as more accurate results. I will post them later today. The 4 ply rollout will not finish and should probably be avoided, it's way slow.
1. Rollout¹ 24/23 13/11 eq:+0.014
Player : 50.10% (G:14.07% B:0.68%)
Opponent: 49.90% (G:13.80% B:0.60%)
Confidence: ± 0.046 (-0.032Duration: 4 hours 55 minutes 35 seconds
2. Rollout¹ 13/11 6/5 eq:-0.024 (-0.038)
Player : 50.14% (G:14.28% B:0.67%)
Opponent: 49.86% (G:14.90% B:0.96%)
Confidence: ± 0.046 (-0.070Duration: 5 hours 06 minutes 03 seconds
¹ 15552 Games rolled. Dice Seed: 1296 Moves and cube decisions: 3 ply Red