[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Bug-gnubg] How to calculate snowie error rate from logged database
Re: [Bug-gnubg] How to calculate snowie error rate from logged databasedata
Mon, 17 Aug 2009 21:36:21 +0200
I checked the code in relational.c and found out that the
matchstat.snowie_error_rate_per_move is the Snowie error rate already,
it just needs to be multiplied with a factor 1000.
So my query could be simplified; it can just start with:
select matchstat.snowie_error_rate_per_move * 1000 as error_rate
Perhaps the snowie_error_rate_per_move column should be renamed to
snowie_error_rate to avoid confusion.
I have checked only cubeless matches so far so I don't know if this
also gives the correct result when cube errors were made.
2009/8/17 Christian Anthon <address@hidden>:
> Please do, and report back, I remember being a bit confused myself
> when I implemented the reporting.
> On Mon, Aug 17, 2009 at 11:29 AM, Misja Alma<address@hidden> wrote:
>> Thanks for looking into the problem. So the snowie_moves number is correct
>> But the product of snowie_moves and snowie_error_rate_per_move is
>> still different than what gnubg reports itself.
>> Maybe I'll have a look into the code myself when I find some time,
>> 'cause I would really like to be able to query some average Snowie
>> error rates from my database.
>> 2009/8/17 Ian Shaw <address@hidden>:
>>>> -----Original Message-----
>>>> From: address@hidden
>>>> On Behalf Of Misja Alma
>>>> Sent: 15 August 2009 15:54
>>>> To: bug-gnubg
>>>> Subject: [Bug-gnubg] How to calculate snowie error rate from
>>>> logged databasedata
>>>> I've been trying to reconstruct my Snowie error rate from a
>>>> match that I've logged in the database with gnubg.
>>>> I then checked what the value for snowie_moves was for my
>>>> match, and it was the total number of (forced or unforced)
>>>> moves of me and my opponent added up.
>>> I can't help with the rest of your query, but I do know that this is
>>> correct. Snowie uses the sum of your and your opponent's moves as the
>>> divisor. Forced or unplayable moves (such as being closed out) are
>>> This is the main reason that gnubg's gradings are so much tougher than
>>> Snowie's. #
>>> Lot's of people dislike Snowie's method, because it is counter-intuitive
>>> to count your opponents' moves as the divisor for your own error rate.
>>> However, it is the rate that people are most familiar with (which I
>>> suppose is why you are trying to extract it.) Douglas Zare also wrote an
>>> interesting article at GammonVillage in which he argued in its favour.
>>> -- Ian
>> Bug-gnubg mailing list