| |
BGonline.org Forums
Special purpose or general purpose (Was: Bear off cube action)
Posted By: Axel Reichert In Response To: Bear off cube action (Bob Koca)
Date: Sunday, 16 August 2015, at 7:54 p.m.
Hi Bob,
true enough. The Isight method is meant as a general purpose tool. We had this discussion some time back when Joachim Matussek said I abused his EPC approximation by checking how accurate it would be for positions with more than 8 checkers.
I goes without saying that a bag of tools (one method for cub-offs, one for rolls positions, one for pips positions, one for long races, one for short races, one for positions with lots of wastage, ...) can in theory give better accuracy than a one-size-fits-all solution.
However, with a bag of tools you will need decision criteria for when to use which tool. These decision criteria will themselves introduce further errors, perhaps even to the point that the errors of the general purpose tool will be more than outweighed.
Also, from a practical point of view, you will probably be less familiar with the more rarely used tools and make a mistake while applying them. And did I mention the additional mental load that goes with a multitude of tools?
This is why I firmly believe in my general purpose method. It is far from perfect, but in my opinion it is the best general purpose method we have. The EPC approximation in my paper may be good for this particular position, but across the range of positions in Tom's database it is much worse than my Isight method with adjusted pip counts, see table 9 in my article (1909 vs 1064).
When I saw Mochy's position and the blunder caused by applying the Isight method, I had the idea to gather positions resulting in blunders and to try to find a pattern behind them. It might be feasible to augment the Isight method with further parameters/penalties to cater for these patterns and hence reduce the total error.
Best regards
Axel
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.