RE: Re: Re: [Bug-gnubg] small suggestion and questions

[Ian Shaw]

From: Christian Anthon on 24 May 2006 16:09

> I tend to agree with you. However, we already have the 
> error-rate numbers to ensure long term consistency. The 
> error-rate to error-class translation is there for a single 
> purpose: to make it easier to relate the numbers to the real 
> world. If the real world world-class players are considered 
> world-class in checker play and expert in cube-play then I 
> think that this is inconsistent and arbitrary. We choose the 
> thresholds when we translate from error-rates to 
> error-classes and since cube and checker play errors are not 
> comparable the way gnubg calculates them, choosing the same 
> thresholds is both inconsistent, arbitrary and in denial of 
> empirical data.
> 
> If we in ten years find that the thresholds need adjusting, 
> we may do so, since we still have the error rates that are 
> consistent over the years.
> 

If you're not proposing to change how the error rates themselves are
calculated, then I'm in agreement with changing the way the textual
classifications are calculated. I'm in favour of gnubg "Expert, World
Class" etc matching up with Snowie's classification.

In that case, your empirical data seems the best way to go.

-- Ian