$$$ NCAA Saturday Games Thread $$$

[iggymcfly]

[ QUOTE ]

Somebody in one of the threads mentioned something about teams playing their first 4 games at home and how they performed in that 5th game coming on the road.

[/ QUOTE ]

Is that really that weird though? It seems only natural that when the squares tend to underaccount for home advantage anyway, that they'd tend to overrate a team that had played all their games at home. That's certainly not the kind of trend I'd deride as meaningless.

[CarlSpackler]

Would it be possible that something as say moving from the standard 11 game schedule to a 12 game schedule in the cfb regular season would be enough to throw off Dr. Bob's model significantly? I was thinking about this the other night.

Back in the day, before the 12 game schedule, one of the strongest trends he used every year was to fade teams who were playing their 5th game before October. The reasoning behind this trend was pretty simple: these were teams that played the week before the regular season started in the Kickoff Classic, etc., and had no bye weeks before they played their 5th game, and were playing an opponent who had played only 2 or 3 games previously, and thus were fresher, more well rested and recovered from injuries.

Now he used further analysis to decide games which qualified for this trend were plays or not, but even if a monkey blindly followed only this trend over the years, he would of made a lot of $$$$, as it was really strong.

Well, now every team plays 12 regular season games, so this trend is useless. Now forget about the trend and think about how this extra game impacts cfb. Teams now only have 1 bye week, instead of 2. In theory shouldn't this increase the variance in the outcome of games, since teams now don't have that extra week off to rest and heal? Combine this with the ever increasing parity in cfb, and I think Dr. Bob's model is almost certainly outdated to some significant degree.

[NajdorfDefense]

[ QUOTE ]
20 years, huh? Funny how the WSJ only quoted since 1999.

'Since 1999, Bob Stoll has recommended 658 bets on college football, or about 81 per season. Here are his results. (When betting against a point spread in Las Vegas, bettors must win 52.4% of their wagers to make a profit.)'

According to him, he made another 800 bets that the Journal didn't feel were worth mentioning. Hmmmm. Or that he's been losing 3 out of last 5 years now.

[/ QUOTE ]

Okay, using DrBob's 1500 picks and 57% number, I ran a confidence interval [I think right, but done quickly] and compared it to his results since 1999 and 2003.

Using 2 standard deviations, DrBob would be above 55.9% winrate 97.5% of time, [or 40-1 against it being pure variance if he is not.] I have him in the mid-55% range since 1999, so statistically unlikely he is 'running cold' since that falls outside 2 std dev test.

Using 3 S.D. confidence intervals, his results are within the range, so he is more likely to be this cold than a 200-1 shot, using a 1-sided test.

I.e. less than 40-1 shot his underperformance is solely due to bad luck, if he really is right 57% of the time.

So you need to ask yourself, is this the 1 in 60 or 1 in 90 5-year or 8-year periods he will statistically run this cold? Or is something else happening? [Which could be that he is not 57%, if he is really 55.75% or whatever, that answers the question.]
[someone with more time should feel free to adjust any errors, I think we're in the right ballpark.]