Predictions Audit

Checking in on the TAPE predictions, and it looks like they're not bad, but not as good as they should have been. Overall, TAPE has correctly predicted the outcome of 76.2% of games, which is just a couple points shy of its predicted rate of 77.9%:

Conf. Range	# Games	Predicted	Actual
50-55%	117	61.5	63
55-60%	101	58.3	58
60-65%	109	68.3	73
65-70%	105	70.9	69
70-75%	124	89.7	82
75-80%	126	97.5	98
80-85%	99	81.5	80
85-90%	144	126.1	115
90-95%	151	139.9	135
95-100%	225	220.3	218

It turns out that a whole lot of those misses were in the first week of predictions. That's not really all that surprising given the fact that at that point most teams had only played 4 or 5 games, and some had played as few as 1 or 2 games. The sample size just wasn't big enough yet, and that led to false confidence. If we take out that first week, though, and look at the results from 12/6/2010 through 1/4/2010, TAPE is spot-on. It's correctly predicted the outcome of 77.5% of the games, against an overall confidence level of 77.6%.

Conf. Range	# Games	Predicted	Actual
50-55%	96	50.5	52
55-60%	76	43.8	47
60-65%	80	50.1	52
65-70%	82	55.4	56
70-75%	94	68.0	63
75-80%	97	75.1	76
80-85%	71	58.3	59
85-90%	105	92	88
90-95%	114	105.7	105
95-100%	167	163.5	163