| |
BGonline.org Forums
Unlimited Game - Cube Action - XGR+ Evaluation
Posted By: Stick In Response To: Unlimited Game - Cube Action - XGR+ Evaluation (Daniel Murphy)
Date: Wednesday, 13 October 2010, at 9:54 p.m.
This site existed before XG and its speed, that's the reason a lot of rollouts are 'subpar' in length or variance left in the rollout.
For many positions, 324 trial would be a big a waste of time.
Again, not with XG, not a 'big' waste of time, which is why I recommend that as a minimum. I won't go as far as saying I've never stopped a rollout before 324 trials but in general, you won't waste the amount of time it takes to preheat the oven if you let the rollout go that far. I believe that's a good idea because even if you get a play 100% certain by 50 trials for example, I like to have a better idea of exactly how bad the other play was. If the variance is still +/- .120 on two plays then what you mentally consume may be way off.
(Me) Usually what you want out of a rollout is for the decision (here the cube action) to be 100%.
(DM) No, that is not what we usually want
Ok, that is what I want. I used to talk of 95% CI and JSDs ... again, all before XG. Now though you'd be hard pressed to find mention of them. 40 is the new 30 and 99.5% confidence is the new 95%.
Stick
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.