Week 2 NFL Pick Results: 12 of 16 ATS
No one has become rich picking 8 of 16 winners against the spread. In fact, many people have probably gone broke doing just that. Being bad, but not bad enough to realize how bad you really are, is one of my biggest fears when it comes to sports betting. I’d rather lose huge and know I’m no good than be perpetually teased by untimely variance making me believe I have an edge. In sports betting specifically, and predictive analytics more generally, a false belief in one’s ability to predict the future is an extraordinarily dangerous thing.
So last week, when the Machine went 8 of 16, I wasn’t thrilled. Very valid questions crept into my mind – maybe something was missed during backtesting and the model is bogus… Maybe the method I’ve dreamed up just isn’t that valuable… maybe my $3,600 of LVH Supercontest entry fees was a complete and utter waste… maybe, maybe, maybe. The reason I entered the contest in the first place was to prove out the method in a public forum that cannot be manipulated, a living testament to the predictive power of the Machine. Maybe after so much fan-fare it simple wasn’t predictive. It’s a possibility that every data scientist needs to be ready to accept: the data is in, your model is no good, back to the drawing board.
Of course, variance works both ways. We can’t go abandoning methods we believe in at the first sign of weakness, the first bad pick, or bad week. Going 50% for 16 games was well within the expected variance, albeit obviously lower than my expectations. With all these doubts going into week 2 it’s difficult to overstate how happy I was to bounce back in such an authoritative fashion, with the Machine going an impressive 12 of 16 against the spread for the week. My 7 unique Supercontest picks went 6 of 7 agains the spread. I always like to put sports-betting performance in perspective by comparing it to the probabilities of seeing the same level of performance if, instead of using the Machine, we were to pick games at random by flipping a coin. The probability of flipping a coin to achieve the mediocre week 1 performance, flipping heads 8 times or more in 16 tosses, is approximately 60%. We expect a coin to get 8 games or more correct most of the time. So going 8 and 8 isn’t an inditement of the Machine, but it also isn’t much of an endorsement.
Now, the probability of flipping 12 or more heads in 16 tosses – the week 2 Machine performance – is a meager 3.8%. Certainly 3.8% is a non-trivial probability. Even going 12 of 16 there is not sufficient statistical evidence to rule out the possibility that the process which I used to pick 12 of 16 against the spread isn’t a random process no better than a coin flip. With such a small sample size the jury is still out statistically speaking. But a random process is much less likely to go 12 of 16 than is a process that does have edge. So after week 2 I remain a skeptic, but my dread about having potentially wasted my LVH Supercontest entry fees are assuaged. I’m moving into week 3 with a modicum of optimism…
Here are the official week 2 results:
My cards now stand as follows:
TheSanction (tied for 406th of 1403): 6 wins, 4 losses
Rotoquant (tied for 223rd of 1403): 7 wins, 3 losses
You can see the official rankings here.
The Supercontest only pays prizes to the top-30 finishers, and my best card Rotoquant is still almost 200 spots out of the money, but this week was a big step in the right direction. It doesn’t take too many 5-0 weeks to rise to the top of the leaderboard. The top prize is $736,575, and 2nd place will likely be half that. My goal, as ambitious and potentially silly as it sounds, is to take 1st and 2nd for a cool million pre-tax. Hope you enjoy following along, it’s already the most exciting football season I’ve ever experienced.
The other bit of good news is that our site Victiv – the next evolution of daily fantasy sports – had a successful week 2 of the NFL season and will be rolling out it’s biggest set of contests yet in week 3. We are putting up $28,000 in free and guaranteed prizes this weekend and offering a 100% drip-in deposit bonus up to $1,000 if you sign up today.
I’ll review the performance of my DFS value plays from last week, put out this week’s Supercontest picks and value plays later on this week.
Follow me on Twitter @TheRotoquant, and please share these picks and this blog with your friends. If all goes well I wont be able to announce them so publicly for long.
Thanks so much, and good luck to everyone in their contests this week!