Re: [Bug-gnubg] Suggestion

[Ian Shaw]

Hi,

 

If 4-ply is overturning some of your mistakes, then it would probably also overturn some of your correct moves, too!

 

It’s always tempting to appeal to higher and higher plies, and rollouts, until the bot agrees with you, but it’s not really very scientific.

 

You can’t currently set gnubg to automatically do extra analysis on the errors. You might be able to do it with a python script, if you’re any good at such things. If not, you might be able to come close by messing with the analysis settings.

 

There’s all sorts of things you can enable and disable, if you delve deep into the advanced settings. It would be a question of trading off analysis time and thoroughness.

 

Personally, I run a 2-ply analysis as my standard. I’d skip 3-ply, because of the known odd-even ply effects, and do a 4-ply analysis or 2-ply rollout if I want a deeper check.

 

Thanks for the suggestion; I agree it might be a useful option to include. There are other analysis options that could be added, such as mini-rollouts, but nothing has been changed in quite a long while.

 

n  Ian

 

 

From: bug-gnubg-bounces+address@hidden [mailto:bug-gnubg-bounces+address@hidden On Behalf Of address@hidden
Sent: 22 December 2015 13:27
To: address@hidden
Subject: [Bug-gnubg] Suggestion

 

Hi!

 

First, thanks so much for this incredible free software. I had an idea, I am not sure if it's already possible but I couldn't find this feature.

 

When you analyse a game (3-ply), you often find afterwards when you 4-ply the moves marked as errors, that either they weren't as bad as they were evaluated by the 3-ply or they weren't errors in the first place.

It would be great to tell the software to automatically 4-ply the errors, so when you review a game you don't waste time double checking manually every error.

 

My 2 cents,

 

Thanks!