| |
BGonline.org Forums
Will XG2 have a 'View Statistics' feature?
Posted By: Timothy Chow In Response To: Will XG2 have a 'View Statistics' feature? (Daniel Murphy)
Date: Friday, 15 April 2011, at 10:59 p.m.
The pseudocubeless numbers are certainly close enough to true cubeless numbers, and cubeless checker play is often close enough to ATS cubeful checker play, that they're not useless. However, again and again we see people trying to get more information out of these numbers than they're designed for. People puzzle over why Play A beats Play B even though Play A "wins more games and gammons and loses fewer gammons." People puzzle over why the supposedly "cubeless" probability of winning differs in the Double and No Double halves of a rollout. People wonder why they can't get the numbers to work out properly. This shows me that people want this extra information and don't realize that they can't get it from the pseudocubeless numbers.
I called the pseudocubeless numbers "ridiculously primitive" for this reason. Not useless, just ridiculously primitive. In this day and age, why are we still struggling with the above confusion? Why don't bots show us cubeless numbers rather than pseudocubeless numbers? Why don't bots show us the cube statistics that we need for a proper understanding of cube efficiency? Surely the game is mature enough that the raw data needed to advance cube theory should be made easily available.
You point out that it's not easy to parse the raw GNU cube statistics. I agree; however, those statistics are just the beginning, not the end. You may recall that I started to develop a theory of "effective gammons" that allows one to make more refined analyses of positions with subtle recube properties. I think that if they were given access to the raw data, theorists would be able to figure out how to turn them into meaningful numbers that would nicely answer the kinds of questions people are always asking. I started taking some steps along this road but got frustrated because it was simply too hard to collect the data needed to start building and testing theories. For example, how do we test the notion that 60% or 70% recube efficiency is "normal"? If we had the raw cube statistics we could collect them for a variety of standard reference positions and find out what the efficiency actually is in those cases. Currently, we just don't have the tools to do this easily. And the only reason seems to be that people don't realize that there's a better solution, so they're not asking for it. This is what I'm trying to change.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.