[Top][All Lists]

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

Re: [Bug-gnubg] [fwd] 12 quickies (bugs and suggestions)

From: Jim Segrave
Subject: Re: [Bug-gnubg] [fwd] 12 quickies (bugs and suggestions)
Date: Thu, 30 Oct 2003 15:28:30 +0100
User-agent: Mutt/1.4.1i

On Thu 30 Oct 2003 (12:07 +0100), Achim Mueller wrote:

>   1.. Why can't I watch all rollout statistics while rolling out (not only
> equities but also closeouts, hits, cubes etc.)?

Screen real-estate will be a problem - if you're rolling out 3
candidate moves, or a doubling decision, you are asking for a lot of
additional information. For truncated rollouts, the figures aren't
even meaninful.

>   2.. When you just start GnuBg and press edit and try to set checkers up
> you can't. Same thing happens with new>position.

It's a known problem which stems from the way gnubg tracks games. This
is likely to change in the future, but it is not a small amount of work.

>   3.. Often crashes after doing a cube decision rollout, probably when
> opening the statistics window.

If you can come up with a position where this happens, please post it,
including your .gnubgautorc file and a description of exactly what you
were doing. This is not a known problem, but if it can be reproduced,
it will be fixed. 

>   4.. Where did the rollout-results of some of the moves and cubes in my
> match go after I saved it?

If you did rollouts as part of analysing a match (as opposed to the
hint window), then the results will be saved in the .sgf file and can
be redisplayed at will.

>   5.. Why only rollout statistics when doing a cube decision rollout and not
> when doing a normal rollout?

The statistics are gathered, as far as I know, no-one has shown much
interest in seeing them, so no method was ever added for displaying

>   6.. I still think it's annoying that there's no dialog to adjust the
> rollout settings before launching a rollout (analyze>rollout). 90+% of the
> time I check my rollout settings anyway (with
> settings>rollout). Don't you?

It's a matter of taste I suppose. In the analysis pane, next to the
rollout button is a small button which opens the rollout settings
window, so your mouse doesn't need to go far. I suspect a lot of users
(certainly I am one) don't often change rollout settings, so having
another dialogue box pop up which needs to be cleared would be an

>   7.. I still think "checker" is the correct word. NOT "chequer" as seen in
> some menus (J?rn?)

Go to Settings->Options->Other and select American English, save
settings and restart gnubg. Chequer and colour should pretty much
disappear from your copy. If you find places which have been missed,
report them and the American English translation can be updated.

>   8.. Why is there no indication on screen of what type of game you're
> playing? Maybe checkboxes (below the Crawford one) with "moneygame" "match"
> etc.

Match length unlimited == money game
Match length any number == match

>   9.. Do you have a database of positions where GnuBg's evaluations are
> nuts? Example: posID: "HgAAKLdtA4AQAA", matchID: "UQkAAAAAAAAA". GnuBg's
> evaluation seems totally random when you try different plys. Clearly correct
> here is double-pass.

I'm not sure if Joseph Heled (who has been doing the training of the
neural net) is collecting these or not. This is a classic sort of
position where a static evaluator looking at most 2 or 3 moves ahead
is going to have problems estimating equities, as most of the
interesting play happens when black reaches the bearoff stage, which
is several rolls in the future.

>   10.. When you set up a position with GnuBg on roll and try Hint you get
> the message "you cannot double" no matter what the position is. Confusing.

Why? You are asking for a hint on whay you should do, the dice haven't
been rolled, so you are asking about doubling decisions. But it's not
your turn, so "you cannot double". Change player 0 from being gnubg to
human and try this and you'll get doubling advice for you as player 0.

>   11.. Why posID AND matchID? Why not just one ID-string describing ANY
> postion (cube, score, hyper etc.)? It doesn't need to be of fixed
> (bit)length either.

I think it's historic. The IDs are already long enough to make them
not very human friendly, combining them would make for an even more
unwieldy string. There are ways to make the encoding more compact, but
the gain isn't all that large and, to quote from the manual:
  but I think you'd have to be a mathematical masochist to try it!

Jim Segrave           address@hidden

reply via email to

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