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

[Massimiliano . Maini]

>>the problem in gnubg's severe cube-rating may be related to the
>>computation of the actual/close number of cube decisions. If the
>>error is above very bad (something like 0.160 I think, it's
>>hard-coded somewhere in the code source) the cube decision is not
>>considered actual/close (hence the error per actual/close decision
>>rate will be higher)... maybe the threshold should be higher ...
>
>Correct, but adjusting the threshold because you don't like the
>size of the resulting error-rate makes no sense to me.

True, but having different weights in the translation error rate --> rating
somehow means that cube errors are less important than chequer errors.
Mathematically this doesn't make sense neither : if you lose 0.1 in equity,
it is 0.1 no matter if it's cube or chequer.

You are assuming that an Expert player is expert at chequer play and
at cube handling ... maybe we have just found that supposed world class
players are in fact just expert (gnubg rating) at cube handling.

Now the problem is : who is a world class player ?

2 possibilities :

1- a player with an overall error rating per (actual) decision smaller
than a give threshold

or

2- a player that has chequer and cube errors per (actual) decision
comparable to the ones of a bunch of players we decide are world class

1 is what gnubg is doing, 2 is what you propose.
Personally, I'll stick with 1 : I like maths and maybe in 10 years
cube handling by human will improve (thanks to gnubg, no doubt) and
a world-class player will be rated world-class by gnubg in chequer and
cube handling :))

On the other hand, it would be fairly easy to add a "Snowie equivalent
rating" (based on the already available "Snowie equiv error rate", using
the same translation used by Snowie, if known) to give Snowie-accustomed
players the same rating scheme.


MaX