Re: [Bug-gnubg] Errors tagging in analysis

[Philippe Michel]

On Thu, Jan 10, 2019 at 12:24:13PM +0000, Ian Shaw wrote:

> I'd go with 0.1 for the Blunder. I think that's common usage, so it 
> would make sense. Then 0.05 for Error is a nice threshold ,and as you 
> suggest, 0.02 or 0.03 for Mistake.

I liked values multiples of 0.06 for a possible further tweak, after 
something pointed by Nack Ballard years ago in the bgonline forum (and 
possibly earlier in his book on opening play or other writings).

He noted that, on average, errors in early play are divided by about 2 
at dmp compared to money play, and by 1.4 at gammmon-go or gammon-save.

>From this, one could use thresholds divided by 2 at dmp and by 1.5 if 
the cube is dead but gammon values are not 0, with the mostly round 
values 0.015/0.03/0.06 and 0.02/0.04/0.08.