
BGonline.org Forums
Improved Cube Handling in Races
Posted By: Taper_Mike
Date: Tuesday, 17 June 2014, at 9:26 a.m.
In Response To: Improved Cube Handling in Races (Axel Reichert)
Regarding the Isight count, I am not sophisticated enough mathematically to judge what Axel Reichert has done. I did, however, read and enjoy his paper today. I think it is well written, especially in comparison to the paper published this year by Michelin Chabot. Provided the formula works as advertised, I think it has the potential to be one of the most important developments ever in the history of racing formulas. The reason is not so much because it improves on the Keith Count, although that is a boon, but rather because it can be applied in a form that gives the Cubeless Probability of Winning. Heretofore, only the Kleinman Count could do that, and the applications to tournament backgammon are extremely important.
My concerns are:
 I don't understand the heuristics used by Isight to trim the search space. This means I am not qualified to judge mathematically whether his results are "optimal."
 Some have commented that the database of positions is slanted towards shorter races too much. This would certainly impact the parameters selected by Isight. In practice, we often apply these formulas in longer races, and even in contact positions that might become races (here I do not mean just minimal contact at the midpoint).
 I am not satisfied that summing the absolute values of the adjustments made by each of the parameters in a race formula is the best way to determine the "effort" necessary for using that formula. I worry that a formula that is better than the Isight formula may have been discarded because it requires too much "effort."
Experience may well show that all of these concerns are unfounded.
Kudos to Axel Reichert for creating and sharing the new Isight formula.
Mike

BGonline.org Forums is maintained by Stick with WebBBS 5.12.