| |
BGonline.org Forums
Fast Rollouts
Posted By: Bob Koca In Response To: Fast Rollouts (Daniel Murphy)
Date: Sunday, 6 December 2009, at 4:39 p.m.
Adding to those quotes, if the VR is bad it will be reflected in the reported std error being a larger value. Suppose that a RO is odne of a DMP situation without VR. Results such as +1, 0, 0, +1, +1, ... is obtained. The std error is just the std dev of that sample. If VR reduction is used and the bot nderstands the position then it will appropriately give the players result a boost after a bad roll and lessen it after a good roll (in an unbiased way). Results such as 0.61, 0.62, 0.61, 0.60, 0.61, 0.64, 0.61, 0.60, 0.59, 0.62,... may be obtained. The reported std error is the dev of this sample. If the bot has major misunderstandings about the position then the VR may actually be detrimental. It may think a roll was good and lower the result even though it was a bad roll. (It is still unbiaseed though, basically opposite situations could occur also). REsutls such as 1.2, 0.7, 0.2, -0.5, -0.3, 1.3, .9, ... may be obtained. The avg of the reported results is an ubiased estimate for the win% but the std. dev may be much larger. But that is indicated in the results. As a side implication comparing the std errors fro a RO done with VR on and the same RO done but with VR off can give a rough idea of whether the bot understands a position.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.