| |
BGonline.org Forums
21$.21 ??
Posted By: Stick In Response To: 21$.21 ?? (RobertFontaine)
Date: Thursday, 15 November 2007, at 11:33 a.m.
First thing to note on my site is the number of trials. Typically on my site when I roll the same position out with both bots, the number of trials with Snowie is less than GNU (due to time constraints/speed of the bots). Here we're lucky, GNU & SW have both been rolled out 15552 trials.
After that we should check if SW's rollout was with a live cube as GNU always is, or if it's cubeless, as I sometimes do w/Snowie. (time constraints again) Once again, here we're lucky as both bots have rolled out the position with a live cube.
As you can see from the following lines in Snowie's rollout:
live cube: -0.253 ±0.007.
live cube: -0.262 ±0.008.
The plays are still TCTC and the rollout should be extended even further before declaring a winner. The same can be said of GNU's results, since they're even closer, and I have firmly come to believe that GNU underestimates its own variance (as does Snowie).
I may one day extend these results but it will not be a day any time in the foreseeable future. I have many other things on my bg plate and I believe at the end of an endless amount of trials with this particular play it should come down to a matter of preference. Against weaker players I play the split, which may seem counterintuitive, but I don't want to give them any perfectas (hit and covers) that basically hand them the game immediately.
Overall the bots agree on almost all of the opening moves/replies so you shouldn't have to often make the distinction of "which bot to believe". If it's as close as it is here I suggest you play it out both ways a number of times until you realize which play you're more comfortable with.
Stick
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.