| |
BGonline.org Forums
Stick vs. mfic final game cube action
Posted By: Henrik Bukkjaer In Response To: Stick vs. mfic final game cube action (Christian Munk-Christensen)
Date: Monday, 19 April 2010, at 12:12 p.m.
The Woolsey Heinrich MET was based on data from live human games.
I know that the games recorded and used in that MET, are so old that bots probably have changed the style and play (and thus the current data would differ).
I also think that the MET itself changed things for the best players, rendering itself obsolete when published!
Such work should therefore be done in iterations, until the delta is under a certain threshold. That's pretty easy for bot vs. bot, but it definitely takes more time for live games to be collected.
It would be hugely valueable, if someone could collect and process live open/masters flight games, and re-do the Woolsey Heinrich table (or simply do a new table based on recent matches.
It must be possible to make a program that can calculate a MET when fed a database of games in txt format, so and all recorded games could go into the database, and an updated MET could be published every quarter!
It sounds like hard work, but if someone would care to put in the hours and make the framework, then you could have a lot of people (tournamament organizers, or unselfish people with too much spare time, an enormous interest in BG (the Stick types :-))) submitting matches to the database through the net.
I wonder how many matches it would take to get a significantly safe result? What's your experience when doing bot vs. bot METs? When to you typically see the sample size make a stable MET?
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.