| |
BGonline.org Forums
Actually...
Posted By: Phil Simborg In Response To: Actually... (niuniuch)
Date: Monday, 27 March 2017, at 11:54 p.m.
I don't think programming a bot to play at a lower level will give us much insight here, as the bot will simply play everything a little worse. That is not how humans play worse. Humans clearly have more problems with certain types of games and positions than others.
I also don't agree with Stick about not having enough data. It's all there. We could, if anyone cared to spend enough time and trouble to do it, look at every match Mochy and Michy and Falafel and Victor have played against all non Giants (and we could include as Giants those players we know from BMAB or otherwise that play under 4 PR) and compare their W/L record to Stick's and other players at Stick's level who claim to consciously adjust their game significantly against non-Giants. I believe that would give a fairly decent sampling to allow us to draw some conclusions, but of course we also have to adjust for the fact that the above 4 play better than Stick to begin with, and if Stick played strictly for PR he probably is a little worse than those 4. (Sorry Stick, but I don't think I'm insulting you by putting you slightly behind those 4.)
But let's get back to the point. The point is that until we have real proof, we can only guess just how much adjustment is going to give a better player more wins against a worse player, if at all. It is going to remain subjective for a long time. And I also agree with Tim that there is no way to take a match and "re-adjust" your PR because of what you say you did because of your opponent. First, for all the reasons Tim said, and second, because you are only adjusting your PR down. You are not taking into account the times you went too far, or you didn't adjust when you should have.
Again, back to the point. I agree with Stick. He and others who play 3 or more PR less than their opponent should do some adjusting and probably will win more, and those of us playing someone better also need to adjust. The problem with the higher PR players is that we are less likely to be right about what the starting point is to adjust from, so even if we decide to take a little more chance on a cube, we are more likely to be wrong by taking a much bigger risk than we thought. So my conclusion is also that the worse you play, the better off you are trying to make the "correct" decisions no matter who your opponent is.
Exceptions: certain types of positions where you know for certain that humans are far worse than the bots, namely containment, back games, PVP, and blitzing. Namely, higher cubes where emotion runs higher as well.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.