| |
BGonline.org Forums
XGR++ RO and 1 ply RO..etc
Posted By: Daniel Murphy In Response To: XGR++ RO and 1 ply RO..etc (Henrik Bukkjaer)
Date: Wednesday, 14 December 2011, at 5:06 p.m.
Henrik,
I believe Gnubg 2-ply should play this position well, but I took Stick's comment to be critical not of Gnubg but of adding noise to simulate a weaker player. In my experience adding noise produces more or less random errors which in no way simulate the playing style or ability of a less than expert player -- as you yourself noted.
Now, after many years of bot development, we have a series of incrementally improved versions, which can be more or less incrementally scaled -- for example, Xavier's data showing that XG(latest) is X PR better than XG(latest minus one) which is X PR better than XG(Latest minus two), etc. It'd be nice to be able to play, say, GNUBG version 14 vs. version 10 vs. version 6 (these numbers are illustrative only and not intended to identify specific versions with modified playing engines) to simulate lesser players, instead of adding random noise. But I imagine that the problem with this approach is none of the versions released of XG, Gnubg or Snowie in the last 5 years play very much worse than expertly, surely not 100 rating points worse. Is there a bot playing engine that does?
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.