[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Rollouts For Dummies/Gnu

Posted By: Timothy Chow
Date: Saturday, 27 March 2010, at 12:56 a.m.

In Response To: Rollouts For Dummies/Gnu (Seth)

The threshold of 3.0 jsd is commonly used, but there is nothing magic about it. If you're just comparing two plays, then 3.0 jsd's is usually interpreted to mean that there's less than a 1% chance that the plays are actually equivalent and that the apparent difference between them is just due to random chance. This interpretation, however, depends on certain statistical assumptions that are not quite true, especially when the cube is in play.

Often when you do a rollout, you're interested in several candidate plays, and you're interested not only in statements like "Play A is better than Play B"; you're also interested in how much better Play A is than Play B. If you stop the rollout as soon as it hits 3.0 jsd's then you will usually not be able to say with any confidence how much better Play A is than Play B.

The bottom line is that if you're short on time and you just care about verifying which play the bot likes best, then stopping at 3.0 jsd's is usually a reasonable criterion (though I personally prefer to run a minimum of 1296 trials). If, however, you want more detailed information about the actual equities or difference in equities, then you will usually need to roll the position out longer than that.

Messages In This Thread

 

Post Response
Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:
 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.