[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Bug-gnubg] JSD problem solved

From: Jim Segrave
Subject: [Bug-gnubg] JSD problem solved
Date: Thu, 15 Mar 2007 23:05:53 +0100
User-agent: Mutt/1.5.13 (2006-08-11)

I've just commited a fix. In the threaded code, the flag indicating if
this was a rollout of a cube decision or a chequer move was being
misinterpreted to indicate if you wanted Cubeful or Cubeless equities
to be used to calculate rankings and JSDs, so all chequer moves were
calculated cubeless.

Chase's bug report and followups made this pretty easy to track down.

This has no effect on Chase's rollout data, it only affected the
display in the rollout window (and of course might make the rollout
stop too soon). I think that's all fixed now.

By the way, do developers realise that the rollout code can rollout
different moves in different board positions (although stopping on
JSDs is no longer interesting)? I've seen people ask for a way to
rollout a set of moves without having to do it manually for each
move. If someone figures out an interface, the code supports it.

You can also do horrible things like the following:

Say there are four moves, two likely best moves you want to rollout
2ply Supremo and a couple of other unlikely candidates that you
wouldn't mind rolling out expert, just to see if they actually are
better than you thought.

Select the two good moves, set the rollout for two play and rollout a
trial or two, then stop.
Select the two unlikely moves, set the rollout for expert and rollout
a trial or two, then stop.
Select all four moves, set a reasonable number of trials and start the
rollout going. The 2 ply moves are extended at 2 ply, the expert moves
at 0 ply.

I don't know if that's good for anything, but there it is.

Jim Segrave           address@hidden

reply via email to

[Prev in Thread] Current Thread [Next in Thread]