Comments on Advanced NFL Stats: Week 9 Game Probabilities

Washington is strong across the board, but their i...

2008-10-31T22:27:00.000-04:00

Washington is strong across the board, but their insanely low turnover rate is driving the numbers.

I'd be worried about Baltimore. Both starting cornerbacks are out for Sunday's game.

I think the only other upset the system is predicting is MIA over DEN. Denver can score, but they have no defense. Should be a good match-up.

Im still extremely suprised about those Wash-Pitt ...

2008-10-31T15:27:00.000-04:00

Im still extremely suprised about those Wash-Pitt numbers..

21-79 %!!!!!!! crazy. I know this is SU but wash is only a 2.5pt fav...

I just see Pitt "playing" much better than their numbers i guess. Hence why im taking them this week.

That Bills game seems dangerous too...considering ...

2008-10-30T22:34:00.000-04:00

That Bills game seems dangerous too...considering how the Bills have been declining since week one.

I'd throw caution to the wind in that Titans-over-...

2008-10-30T22:31:00.000-04:00

I'd throw caution to the wind in that Titans-over-Packers forecast...Green Bay has been improving since week three. Their highest lead, lowest deficit and margin of victory or loss has been increasing...I'll bet if you plotted out stats on a graph you'd see the same improvement.

Matter of fact - is that something you could do? Just plot out a team's stats from week to week? I'd be interested to see how they moved.

Anon-Thanks. 52.2 isn't so great. I guess that's a...

2008-10-30T18:44:00.000-04:00

Anon-Thanks. 52.2 isn't so great. I guess that's about break-even for bettors.

That brings up another point about SU predictions. Some years are much easier to predict than others due to match-up disparities. For example, when divisions tend to be tightly bunched in team strength, SU predictions are very difficult. Take the NFC East right now--they play each other twice, and they're all pretty good. The NFC West is almost as hard to predict because the teams are of similar strength. But with a team like NE last year, it's easier to predict. When divisions are widely spread in strength levels, it makes for easier predictions.

2006 was a relatively difficult year. The spreads predicted the winner only about 65% of the time. 2007 was far easier as the spreads correctly favored the winner about 70%. This year seems more like '06 than '07.

Accidentally posted this on Week 8 instead of Week...

2008-10-30T17:57:00.000-04:00

Accidentally posted this on Week 8 instead of Week 9:

I count a 36-33 record ATS for this season, around 52.2%, due to an inordinate amount of favorites not covering the spread (something around 75% of winning favorites have covered, down from the typical 82.7%).

That's a good suggestion. I've done that in the pa...

2008-10-30T17:49:00.000-04:00

That's a good suggestion. I've done that in the past. In 2006 I basically doubled the weight of the most recent 4 weeks. If I recall, it helped overall accuracy very slightly, maybe a game or 2.

It multiplies the complexity of the system considerably more than you'd think. I'd have to also over-weight recent opponent strengths. Each team requires its own special calculation for most of the season because of byes.

I really need to automate a lot of this much better and not spend so much time on the weekly upkeep chores. Once I do that, I can tinker more with over-weighting and other improvements.

Should you try to include moving averages (last 3 ...

2008-10-30T17:19:00.000-04:00

Should you try to include moving averages (last 3 games)? So more weight is giving to closer games.

The short answer is I don't track it ATS. I only u...

2008-10-30T15:29:00.000-04:00

The short answer is I don't track it ATS. I only use the spreads as a benchmark for accuracy in predicting SU winners.

The long answer is: My system is designed to estimate the probability for a SU win, and does not factor in spreads in any way. Historically, it is slightly more accurate than the spreads at predicting winners. However, because SU winners usually also beat the spread (at about 80% or so), if you are good enough at picking winners, you'll have a marginally favorable record against the spread too. Last season the system would have been correct 59% ATS, just by choosing the SU favorites. To be honest, that's even higher than I would expect. I think there's some favorable luck involved. I don't advertise my model for gambling purposes, and there are never any guarantees, but everyone is free to use it as they wish!

So far this season the system is 47-22 SU, which is still 2 games above the consensus favorites. That's 68.1%. Assuming the 82.7% "SU winner also beats the spread" rate from the past 6 seasons, that suggests my system should be about 56.3% ATS.

But again, that's not what it's designed for.

Thanks! Hey can you post a running total of your r...

2008-10-30T13:06:00.000-04:00

Thanks! Hey can you post a running total of your record against the spread, given it was something like 60% in past years if I recall. What is it so far this season? Thanks!