mrdct
29-12-06, 02:16 PM
I've just had a fiddle around with my datums spreadsheet so that it now calculates cross-imps as well as a butler score versus the average of the middle two scores from the four tables that played the boards. The results for Cayne03 are:
Butler X-IMP
seamon-cayne 0 (10)
paul-mike (33) (47)
ish-vince 62 58
sarts-hughes 10 (0)
bob-zol 0 10
migry-westheimer 33 47
bruce-avi (62) (58)
tony-keiran (10) 0
At both butler and cross-imps, there is quite a discrepency between the match result of -23 and net butler and cross-imp results of -33 and -37 respectively; but there was some duplication of poor results - most notably on board 1 where Oz-One went for 800 at one table and 500 at the other for a loss of "only" 16 imps.
I'm doing some analysis at the moment of expected error rates when butler scoring a small number of tables (such as four) and believe that in a 28 board match the error rate is about +/- 12 imps. I'm basing this on a single sample I took of 28 boards played on BBO 16 times on which I calculated datums for 25 random combinations of four tables which, on this sample, had a standard deviation of 11.4 imps.
I'm not a great expert in statistics, so I'd appreciate any input people may have as a meaningful way of calculating, reporting and interpreting results from four tables.
Butler X-IMP
seamon-cayne 0 (10)
paul-mike (33) (47)
ish-vince 62 58
sarts-hughes 10 (0)
bob-zol 0 10
migry-westheimer 33 47
bruce-avi (62) (58)
tony-keiran (10) 0
At both butler and cross-imps, there is quite a discrepency between the match result of -23 and net butler and cross-imp results of -33 and -37 respectively; but there was some duplication of poor results - most notably on board 1 where Oz-One went for 800 at one table and 500 at the other for a loss of "only" 16 imps.
I'm doing some analysis at the moment of expected error rates when butler scoring a small number of tables (such as four) and believe that in a 28 board match the error rate is about +/- 12 imps. I'm basing this on a single sample I took of 28 boards played on BBO 16 times on which I calculated datums for 25 random combinations of four tables which, on this sample, had a standard deviation of 11.4 imps.
I'm not a great expert in statistics, so I'd appreciate any input people may have as a meaningful way of calculating, reporting and interpreting results from four tables.