NFL Wildcard Review and Sony Open Fantasy Golf Picks!
At the end of last week I swore I’d give myself a break from this blog. I needed time to digest the happenings of the season, collect my thoughts, possibly revamp the algorithm and triple check the code for small errors. Work was pressing hard on me, and I didn’t know if I’d have time to make the projections, let alone post them… I mean – the point was projecting the NFL regular season, right? I did that, and did it well – why not leave well enough alone?
To be honest – I simply love the idea of predicting sports outcomes. Last week while watching the wildcard playoffs with a good buddy who is aware of the blog he alerted me to the Las Vegas Hilton Super Contest. It’s a season long pick-em pool against the spread. Each week participants submit their 5 most confident picks; a correct pick is worth 1 point, a tie is .5, and a loss is worth 0. The cost to participate is $1500 and this year the winner took home over $500,000, picking roughly 67% of his games correctly against the spread. The contest had over 1,000 participants and it paid something out to each of the top 30 handicappers for the season.
I was entranced. This is exactly the kind of thing I need. Picks can be entered by proxy – so I never have to actually go to Vegas. The winner of this contest is widely considered the best handicapper on the planet, and past winners have parlayed their success into careers in either sports betting or sports analytics for professional franchises (mostly baseball). Usually though, winners have been of the former ‘sports-betting’ variety; not concern with statistical treatments of the games but instead making their picks on old-fashioned intuition, instinct, heresay, line-movements and myriad other nitty-gritty non-statistical things to consider when picking an NFL game against the spread.
So I was all-the-more captivated when I researched the background of this year’s winner. David Frohardt-Lane is a 36-year old Wall Street (well, Chicago anyway) trader who specializes in predictive modeling. He used an algorithm to pick every game. He was 67% accurate. And he won half a million dollars.
This is a video lecture about his NFL-handicapping algorithm that I found on Vimeo, which he posted back in July of 2013. If he knew that a mere 6 months later he’d be collecting $500 grand for this method, I wonder if he’d still want it outlined and made so publicly available. What’s most striking to me about David’s methods is their apparent simplicity; he uses nothing more than linear regression and his method is of the common ‘power-ranking’ variety; the algorithm tries to understand how good each team is based on their performance, and using this to predict expected outcomes. He claims to use only data that is freely available in box scores … very impressive indeed.
I couldn’t help but get excited at all of this. This is what I do! What’s more – I’ve done it better than he did, at least for the last 5 weeks of the season. From a qualitative perspective, my methodology is on the opposite extreme of his – it’s so complicated and uses such advanced statistical methods that I could describe exactly what’s happening internally to someone in words and be relatively confident that they could not reproduce the method. The majority of the computational heavy lifting is done using algorithms of my own design – other people don’t even fathom the method, let alone have the means to reproduce it. In addition, each season I pay for a data provider to provide me with the highest-quality detailed NFL statistics, and I use these statistics to extrapolate still more proprietary variables (i.e. unique ways of combining the standard variables that help better understand performance). The point of mentioning all of this is not to impress the reader (“Oh, wow – how cool you are Mr. Rotoquant… you know what you didn’t do? Win a half million dollars!“), but rather to get across exactly what my thought process was as I was watching this lecture… I can do this… I have to do this!
Needless to say, next year’s Rotoquant blog will focus on my performance in the super contest. It’d be crazy not to try, and it’d be too fun to avoid sharing the experience with the world and documenting it for all to see.
Also needless to say, seeing Mr. Frohardt-Lane’s performance has reinvigorated my computational spirit. With project deadlines pending and plenty of code still to write I’ve had one helluva work-week (my day-job is being a director of a computational consulting software company). The girlfriend is out enjoying time with friends tonight so I did the second most relaxing thing I can do to take my mind off of work … well … more work! I sucked it up and put together the grid of of last week’s NFL wildcard action – that part was easy; the machines are already built I just have to download data and run a script.
What I didn’t anticipate doing until at least few weeks after the NFL season was publishing my fantasy golf projections. But I couldn’t help it… Since I’m actually using these to build my teams, and since I had already made them anyway for the Sony Open, I might as well start posting them now. At this point my opponents are no doubt starting to pay attention, so I’ll be posting all the golf projections after the rosters have locked (after the first golfers tee off for round one). It’s a four-round tournament and I’ve got a track-record of posting predictions well before competition kick-offs, so I hope the reader will trust that all projections truly were made prior to tee-off.
Without further adu, let’s get into it!
Wildcard Playoff Week Analysis
Of course, as the betting God’s demand, I can go 77% in the regular season (see last week’s full season analysis) and as soon as I venture into a friendly wager with my dear old dad the pick-machine goes cold. The only pick my dad and the machine disagreed on was the IND vs KC game – where he took KC plus 1.5 points. The machine picked IND minus 1.5, and in amazing fashion IND managed to win the game… 45 to 44. Oh well – there are worse fates to face than having to buy dinner for my dad. Keep an eye out for the round two picks to be posted sometime tomorrow or Saturday morning at the latest.
Sony Open Fantasy Golf Projections
To better understand this grid – the Rank.Index is a measure of how well the player will perform when compared to the average player in the field. I can’t go into much more than that without giving away part of the algorithmic goose – but suffice it to say that a higher Rank.Index is better. So here, Adam Scott was predicted to win the tournament (he is predicted as the best player in the field, by far, which I was glad to see since it coincides with his Official World Golf Ranking of 2 – putting him only behind Tiger Woods who is not in the field this week). On the downside I thought it was pretty interesting to see that last year’s winner Russell Henley is predicted to be terrible, finishing second to last in the field. It’s not that I’m incredulous about the odds of Henley underperforming – it’s just that many experts would probably expect him to, at least, above average. After round one, Henley indeed looks like he’s on course to miss the cut, posting a +4. The thick red line here represents the approximate cut-line.
The algorithm used to make these predictions is a modified version of the one used to create the Cover.Index during the NFL season. I use it to build my own personal fantasy golf rosters in competitions on sites such as DraftDay.com and DraftStreet.com – the same sites on which I participate in fantasy football competitions – but with so many NFL players I am unable to use this advanced statistical method to predict them all. With Golf I can predict each of the 130 or so players in the field with as much attention as I’d pay to an NFL game. I’m eager to see how the machine performs.
Ok, that’s all for now!