BGonline.org Forums

Hit or run?

Posted By: Timothy Chow
Date: Sunday, 16 January 2011, at 3:08 p.m.

In Response To: Hit or run? (leobueno)

As a tip for the future, you'll generally get more honest answers if you don't mention any bot info in your initial post. It's hard for most people not to be influenced in their comments by what they perceive to be the "right" answer.

Magriel's safe/bold criteria are perhaps the single most valuable guide in this sort of position, but unfortunately they give some conflicting advice. We have more checkers back and a race deficit, which usually indicate bold play. However, we have a weaker board, which usually indicates caution.

The match score is also a consideration. Usually when trailing in the match we are more inclined to play for gammons (although 5-away/4-away is a tricky score where sometimes the leader prefers gammonish positions, at least if the cube is turned).

We'd really like to contain White's back checker. Right now White's blot can move with 2's or 6's; if we hit then White's blot can move with 1's, 2's, and 4's, with 1's being good for us (pushing the blot further back) and 2's being bad (we get hit back). Of course if White stays out then we gain. Hitting should win more gammons because of the blitzing opportunities.

Running also has its merits. We are in danger of getting primed and so it's important to get those back checkers going if possible. We can run all the way with 24/14 or stop at the bar point and clean up one of our outfield blots. Either way we leave a bunch of shots (although 2's and/or 6's are somewhat duplicated), but if we are not hit then we have made progress on one of our main goals.

Finally, you didn't mention 11/5 10/6. This should lose the fewest gammons but will probably lose more games. This degree of caution is probably overkill except in match situations where avoiding a gammon loss is a big priority.

It looks like a difficult choice. I would probably run with 24/14 but I'm not at all confident that it's correct.

If you want to do a rollout, then I'd recommend 1296 trials with GNU 2-ply or XG 3-ply, including variance reduction but not truncated. There are "cheaper" settings but the results won't be so reliable. I'd also recommend that you roll out the position for money as a comparison. Then you can see what the effect of the match score is.

Messages In This Thread

• Hit or run?
  leobueno -- Sunday, 16 January 2011, at 1:30 p.m.
  • Hit or run?
    garyo -- Sunday, 16 January 2011, at 2:55 p.m.
  • Hit or run?
    Stick -- Sunday, 16 January 2011, at 3:01 p.m.
    • Hit or run?
      Timothy Chow -- Sunday, 16 January 2011, at 3:09 p.m.
      • Hit or run?
        Stick -- Sunday, 16 January 2011, at 6:03 p.m.
  • Hit or run?
    Timothy Chow -- Sunday, 16 January 2011, at 3:08 p.m.
  • Hit or run?
    Sebastiaan kuijs -- Sunday, 16 January 2011, at 4:55 p.m.
  • Rollout
    Stick -- Tuesday, 18 January 2011, at 6:26 a.m.