| |
BGonline.org Forums
Position of the Day 10-10-26 - Rollout No Jacoby
Posted By: Chuck Bower In Response To: Position of the Day 10-10-26 - Rollout No Jacoby (Colin Owen)
Date: Sunday, 8 November 2026, at 4:57 p.m.
I've delved pretty deeply into this and my conclusions are:
1) Jacoby does appear to have an impact on the equities, and 2) depending upon the level of evaluation and rollout you can flip the play when choosing Unlimited, No Jacoby.
Evaluation-wise, I looked at 1-ply through 6-ply. All chose 6/5. (My attempt at 7-ply seemingly caused XG to go into an infinite loop, which is inconsistent with past usage of that option when I've occasionally tried it -- previously it worked OK, just took a long time.)
XGR also liked 6/5 but then XGR+ not only flipped but also gave the bigger (0.014) preference for 10/9 compared to XGR++ which slightly preferred that by 0.002.
Next I ran 3-ply RO. I stopped after 1296 (for reasons you'll see below) with 10/9 favored by a mere 0.0005, only showing 52.9% chances of being best at that level of effort.
Now for the bigger guns -- 5184 trials at 4-ply said 10/9 better by 0.014 (0.0139 if you want to go the extra sig-fig) and likelihood of that being wrong at 1 chance in 157,000. So there's your answer? Most people would not have even gotten this far, but I'm not finished.
XGR+ rollout, 5184 trials reverts back to 6/5 although only by 0.0034. Confidence is 87.3% so more rollouts needed. (My Dell laptop groaned for 7 hrs 36 minutes on this rollout alone to reach this inconclusive result).
Finally I compared some of the individual equities, specifically between the 4-ply rollout and the XGR+ rollout, and noticed one glaring difference: for the 10/9 candidate, 4-ply RO's equity 0.5417 with standard deviation 0.0022 (half of XG's reported 95% confidence half interval) whereas XGR+ had that at 0.5235 with 0.0019 s.d. How likely is that if they both are playing the same? That's 6.343 joint standard deviation separation which happens randomly about one time in 444 billion! To be fair, I chose specifically the 10/9 comparison because of its size, and I could have chosen the other play (6/5) which only disagreed at the level of 0.0009 and 39% confidence. Thus you can divide 444 billion by 2 to conclude what I found is likely to be a 1 in 222 billion occurrence, if random.
So what's my conclusion? 1) the choice is close; 2) 4-ply and XGR+ clearly don't play the game the same way after 10/9. But which plays it 'better'? I don't know the answer to that and assume it would take much deeper investigation such as looking specifically at how these two levels play the following rolls AND then roll out out those disagreements. I think I'm done with this one. :)
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.