|
BGonline.org Forums
Experiment
Posted By: AP In Response To: Breaking a match into smaller sets: Another advantages (Bob Koca)
Date: Wednesday, 28 August 2013, at 3:43 a.m.
You raise some interesting points, and I am not convinced either way.
The first thing I would want to have is a quantitative estimate of average cube complexity for a given match that captures the factors you listed. I trust the experts here could come up with some metric that penalizes a position when cube complexity is nonexistent and gives a bonus when cube complexity is high. For example, assign 0 to DMP/Crawford/post-crawford/etc, .5 when position is close to money cube action, and closer to 1 when the cube has power but the take points start to deviate from money. Average these values across all turns, and compare the metric between an 11-point match and 2 out of 3 5-pointers. It would be interesting to see this comparison in bot vs bot of equal as well as unequal strength. Also as a bonus, you could get other interesting metrics you are looking for, such as average # games per match and stdev of # games per match.
I don't know if it is feasible/easy to set up XG to repeatedly play itself in a fixed match length, outputting the relevant results at the end. If so, I'd be happy to donate some CPU power to the cause.
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.