[Bug-gnubg] LAR, FIBS and MWC

[Mueller Achim]

Hi folks,

I have a sample of 410 games now (1-5 pointers, but mostly DMPs):

Actual result:  59.02   [wins/number of matches]
LAR:            55.05   [sum(LAR)/number of matches]
Fibs:           54.70   [sum(MWC]/number of matches]

using the Fibs formula "1-(1/(10^((YOU-HIM)*SQRT(ML)/2000)+1))".

LAR and Fibs are closer than I thought before, but both differ
from the actual result. Is anybody able to calculate the standard  
errors?

What I still don't get: two queries into one new table ... here
are my queries:

select a.session_id as No, b.length as Length, round(a.actual_result 
+0.5)
as Result, round(50+a.luck_adjusted_result*100,1) as Luck_Adjusted,
round(a.error_based_fibs_rating,1) as FIBS,  
round(a.snowie_error_rate_per_move*1000,2)
as Snowie, date(b.added) as Date from matchstat a, session b
where a.session_id = b.session_id and a.player_id = '${NAME_ID_RESULT}'
group by a.session_id;

select c.name as Name, round(a.error_based_fibs_rating,1) as R_Opp  
from matchstat a,
session b, player c where c.player_id = a.player_id and a.player_id !=  
2 and
(b.player_id0 != '${NAME_ID_RESULT}' or b.player_id1 != '$ 
{NAME_ID_RESULT}')
and a.session_id = b.session_id

$NAME_ID_RESULT is a query for a certain player name that I do before  
in a shell script.

At the moment I put the result of both queries into a spreadsheet.

Ciao

Achim