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

BGonline.org Forums

Supremo MET in the Real World

Posted By: Nack Ballard
Date: Wednesday, 16 December 2009, at 7:44 a.m.

In Response To: Supremo MET in the Real World (Rich Munitz)

"Okay, but what about the people that aren't practically speaking?"

First of all, I agree with both Neil and David that you have to start with a MET based upon perfect play and best information available, and then worry about adjustment. We rely upon bots to do rollouts of checker or cube play and humans will not play it the same either. But you need consistency, and it is essential to have a proper MET that represents bot play so that the rollouts make sense.

As for adjustment, sure if someone in live play can achieve comparable results to the bot, then they don't require an adjustment. But what about their opponent?

It seems my attempt at humor (having fun with the alternate meaning due to no commas before and after "practically speaking" in the original statement) fell flat.

On the issue itself, I agree with all of you that, given a choice, the theoretical MET is more valuable than an empirical MET (which, among its other inadequacies mentioned, would be a moving target as people's perceptions change).

At the same time, I think there is a key point being missed or misunderstood in this thread. It doesn't have to be one or the other. Surely an empirical MET would have some value, assuming enough data can ultimately be collected to sufficiently diminish statistical noise. If humans misplay certain scores more than others, during a match a player could adjust theoretical MET double and take points so that he reaches the human-strong-play scores and his opponent reaches the human-weak-play scores a bit more often.

One can choose to make either no adjustments or guess adjustments to the theoretical MET, but IMO an empirical MET has the potential to give a more accurate sense how far one should adjust and specifically to which typical weak spots (including the option of selectively enhancing or negating adjustments against certain opponents). Whether such a MET would be too much work to create or the interpolations too much work to learn to justify the equity gain is another matter.

Nack

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.