Hi All. I wanted to play 6/5 6/4 here; which is apparently a big mistake. The right move being 6/4 5/4, leaving a lamb for the slaughter on the 6 point.
I can only guess that I am supposed to be positioning for a safer bear-off next throw, due the fact that red doesn't have much hope even if he hits me?

Admit I am shocked on this one. I guess it must be a pay now or pay later situation. White's board is still weak but will get stronger, while black has strong gammon chances if white misses the double shot.

But after 6/5 6/4, black has 14 rolls to clear the 5 point without leaving a shot. I would have thought that is too much to give up on. Maybe needs a rollout?

Maybe the opponent's open high points could mess up the decision, but i cant see a reason to Leave a double shot

Your original play (6/5 6/4) is correct for all the obvious reasons. I think you need to upgrade your bot. Try Extreme Gammon, which has no problem with this sort of decision.

I actually use both: GNU on full tutor as a sort of training session, and XG to rate my progress (no tutor, just a review after each game to see where I went wrong).
I've notice GNU suggests some strange things in the very late stages of the game. A few even I could see were just "glitches" but most I assume I've just got horribly wrong.
Which makes me wonder: in these bear-off situations how reliable is GNU?
Had you not explained this to me, I would have tried to remember this position and made the same error again if something similar came up.

Back when I was teaching computer science, I was always amazed by the strong opinions my students had about which computer was best: Mac or PC. As best I could tell, the computer one owned was ten times better than the other. (Mac owners were a little more chauvenistic than PC owners.) The thing that impressed me the most was that the people expressing an opinion usually had little or no experience using the competition. How could they say one computer was better than another when they had only used one of them? Literally, they were not qualified to hold an opinion.

The debate between XG and GnuBg is a little bit like that. If you have not learned the ins-and-outs of both of them, it will be difficult to make anything more than a surface comparision. An uninformed bias may leave you open to assuming the worst.

I have seen both XG and GnuBg make some pretty silly board-breaking plays in holding games. Those generally occur behind an opponent’s anchor. That is different than the simple bear-off position we have here. Even a gullible person such as myself would find it hard to believe either XG or GnuBg would volunteer a shot in this position. If you find it easy to believe that one of them would, you may want to consider whether you have a bias.

GnuBg’s 0-ply eval is similar to a 1-ply eval in XG or Snowie. It is a weak setting that easily gets this position right. Needless to say, so do all the stronger settings.

The real question here is what went wrong for Kamba. Perhaps the two plays were evaluated using different plies. In that case, the one that used stronger settings would be displayed ahead of the other. That is something both XG and GnuBg do.

Mike

Thanks Mike. This is the hint I'm getting:



Any idea which setting I got wrong? Or if i can reset to the defaults?

I'd like to see the GnuBgID for the position. You can press Ctrl+C in GnuBg to copy the ID, and then press Ctrl+V to paste it into a message here.

Pending that, I can make some general recommendations, the most important of which is to use nothing weaker then 2-ply, "World Class" settings. If you have a fast computer, you can try 3-ply, "Grandmaster." There are three places you need to make the change:

1. Settings > Analysis > Analysis Level: World Class
2. Settings > Players > Player 0 > GNU Backgammon: World Class
3. Settings > Rollouts > Play settings > First play both: World Class

You should also add a checkmark so the tutor uses the same settings as the analysis:

Settings > Analysis > Eval Hint/Tutor Level: Same as analysis

Is your version up to date? You can check the top line of the Build Info window to find out. I am running the latest Windows version: GNU Backgammon 1.04.000-mingw 20141021.

Help > About GNU Backgammon > Build Info

Hope this helps.

Mike

Many thanks Mike. Unfortunately I've already changed the analysis settings to world class so couldn't tell you what my original settings were (i set up the position again by pasting the ID you posted above). Happily, the hint i'm getting now is the correct one.
I'll follow the rest of your instructions there to avoid any future issues.
I think this was clearly a malfunction of the operator rather than the equipment!
Thanks again.

All the settings below "Expert" work by adding random noise to the equities generated by Gnu's evaluation engine. This will sometimes cause the best play to be assigned a lower equity than some of those that are worse. The idea is to create an opponent that does not win so darn much. It is possible that you selected one of these sometime in the past.

That is one way to explain what you were getting. Whether or not that was the cause, I'm glad to hear it's working now.

Mike