BGonline.org Forums

Results for 2 "double mutant" experiments

Posted By: MK
Date: Thursday, 15 August 2024, at 9:32 a.m.

I posted the results for these two experiments 23 and 26 to my web page last week but couldn't take time away from other discussions in DailyGammon to announce them here.

These experiments are "double mutants" versions of cubeful Experiments 21 and 24. I ran them for the possibility that they may help me debunk a certain different "cube skill" fallacy.

Unlike bot making cube decisions for both players in Experiments 21 and 24, in these the checker play mutant also makes the mutant cube decisions as it did in Experiment 16, based on the same game stages formulas.

In Experiment 16, after 20,000 games bot won 37,854 (51.32%) and mutant won 35,900 (48.68%) of points, i.e. 2.65% less; bot won 10,003 (50.02%) and mutant won 9,997 (49.99%) of games, i.e. only 0.03% less.

In Experiment 21, after 20,000 games bot won 46,782 (92.54%) and mutant won 3,773 (7.46%) of points, i.e. 85.07% less; bot won 18,840 (89.71%) and mutant won 2,160 (10.29%) of games, i.e. 79.42% less.

In this experiment, it would be expected to see mutant lose the combined amounts of those two experiments. That is, 2.65% + 85.07% = 87.72% less of points and 0.03% + 79.42% = 79.45% less of games.

After 20,000 games bot won 79,255 (90.90%) and mutant won 7,934 (9.10%) of points, i.e. 81.80% less; bot won 17,571 (87.86%) and mutant won 2,429 (12.15%) of games, i.e. 75.71% less.

These results mean that, contrary to what would be expected, mutant lost only 81.80% of points instead of 87.72% expected and only lost 75.71% of games instead of 79.45% expected.

Similarly, in Experiment 24, after 20,000 games bot won 32,208 (57.85%) and mutant won 23,467 (42.15%) of points, i.e. 15.70% less; bot won 11,949 (59.75%) and mutant won 8,051 (40.26%) of games, i.e. 19.49% less.

In this experiment, it would be expected to see mutant lose the combined amounts of experiments 16 and 24. That is, 2.65% + 15.70% = 18.35% less of points and 0.03% + 19.49% = 19.52% less of games.

After 20,000 games bot won 47.912 (59.44%) mutant won 32,687 (40.56%) of points, i.e. 18.89% less; bot won 11,530 (57.65%) and mutant won 8,470 (42.35%) of games, i.e. 15.30% less.

These results mean that, contrary to what would be expected, mutant lost only 18.89% of points instead of 19.52% expected and only lost 15.30% of games instead of 19.52% expected.

The mutant had already performed better in cubeful Experiments 21 and 24 than in cubeless Experiments 22 and 25, which means that these "doube mutants" performed better than in both types of those previous experiments.

I ran 2 more pairs of cubeful vs cubeless experiments without mutant moves, i.e. bot vs bot. I will post them soon also. By now, some of you may have an idea as to where I'm going with these 6 pairs of cubeful vs cubeless experiments. I had already given a hint that it has to do with what the doubling cube does or doesn't add to backgammon. Any guesses and pre-comments..?