[Bug-gnubg] Predicted time for finishing the rollout

[Hardy Hübener]

Hi everyone,


don't know whether this has ever been discussed here. If so, disregard 
my mail.

When you conduct a rollout and find out that the result is not 
significant enough and resume this rollout with more trials, the time 
prediction how long this rollout will last is of no use.

I think GNUBG is calculating the remaining time from how long it took to 
reach the previous rollout. As the rollout started with the previous 
result, it took GNUBG 0 (zero) seconds to reach this amount of trials. 
So if I prolong a 1296 trials rollout to 2592 trials GNUBG is predicting 
that the rollout will be finished in a couple of seconds. However it's 
going to take hours (of course).

Perhaps another algorithm could be used to calculate the remaining time 
more accurate. (There are more important features, but as I do many 
rollouts, it's sometimes helpful how long a rollout is going to take 
approximately.)


Thanks,

Hardy


--
Hardy's Backgammon Pages --> www.hardys-backgammon-pages.com