Re: [Bug-gnubg] Fwd: Listing JSDs in rollouts

[Ian Shaw]

Since no-one else has commented I'll chime in here:

Adding the JSDs is a great idea, if it can be done tidily.

Now I'm going to hijack this post somewhat, and develop the topic furhter:

Gnubg has a larger problem in it's rollout reporting. Anyone who has seen XG 
rollouts knows thay are much more succinct and readable than gnubg's.

TO some extent, this is because XG has fewer options than gnubg (JSD-halting is 
a good example), so gnubg rollouts are always going to have to contain more 
data. 

However, XG does a great job of displaying the data so that it's easy to 
compare the rollout results between plays without the rollout context 
interfering. It does this by
(a) using a small font for the rollout context
(b) relegating the rollout context to below the results and marking them with a 
superscript 1,2 etc.

Gnubg would benefit from something similar.
A smaller font would be quite easy. Other improvements would be trickier. 
Because gnubg can stop early, there might be a different number of trials for 
each play.
However, a lot of information gets repeated, e.g. rng, seed, plies, etc. This 
could be condensed somehow.

Long before XG existed, I suggested a differential report of rollout context. 
That is, gnubg exports only display context information which differs from the 
play below it. This would cut out a lot of redundant rng, seed, ply information.

An alternative solution would be to print the number of trials and jsd for each 
play, then reference the rest of the context with a superscript to a footnote, 
XG-style.

Here is a link to a rollout I performed recently. Four plays were rolled out at 
2-ply, and five more at 0-ply. https://sites.google.com/site/gnubgexport/

The original is on the left. Look how messy it is.

On the right I have tweaked it to reduce redundant information, and include the 
suggested JSD information. I think the primary data - the rollout results - are 
a lot more legible.

What do people think of this?

There is scope for a lot more improvement, such as:
Labels for cubeless and cubeful equity
Labels for wins, gammons, backgammons
Making the confidence intervals less intrusive
Any others?


-- Ian



-----Original Message-----
From: address@hidden [mailto:address@hidden On Behalf Of address@hidden
Sent: 25 May 2011 19:10
To: address@hidden
Cc: address@hidden
Subject: [Bug-gnubg] Fwd: Listing JSDs in rollouts


Wondering whether y'all have decided on the suggestion that the rollout
results indicate the plays' JSDs.  Note that because we can manually stop
rolling out specific plays it is not clear in the outoput listing whether a
rollout was stopped manually or whether the JSD limit was reached.  In other
words, by looking at the output, you cannot tell whether the JSD limit was
reached.

Thanks.