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

BGonline.org Forums

Nine 5k results (report) - that why people need to check SI

Posted By: Timothy Chow
Date: Tuesday, 18 January 2011, at 5:30 p.m.

In Response To: Nine 5k results (report) - that why people need to check SI (eXtreme Gammon)

Part of what's subtle about this subject is that the decision about how long to roll something out can affect the statistical significance.

If everyone always rolled out positions 5184 times then it's likely that the confidence intervals would fairly accurately represent what we intuitively think they should, except perhaps in positions where the cube is getting repeated turned. However, that's not what happens in practice. People roll a position out for a while, then look at what's going on, and stop or continue based on whether they like what they see. This natural and seemingly innocuous procedure affects the statistical significance more than one might think. Unfortunately there's no easy way to program the bot to take this into account properly, because it depends on the details of how the rollout choices are made, and most people are not even conscious of the subtle choices they're making, let alone disciplined enough to follow a strict protocol every time.

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.