As I said last time, week 17 can be difficult to predict. This is also a nice way to cover up any potential bad performance by the models (maybe I should have this disclaimer every week). Let’s see how things turned out.
Over/under: Atlanta and Carolina hit 41 for a push, and both models went 7-8-1. On the season both models managed to have the same record, 119-115-5. This is just above chance, although not significantly so.
Moneyline: Mario liked Cleveland, Tennessee, Oakland, Washington, Cincy, Houston, San Diego, Arizona, and Seattle. Oakland, San Diego, Houston, and Seattle came though for $117.63 if you bet $100 per game. Luigi liked Cleveland, Minnesota, Tennessee, Oakland, New England, Tampa, Cincy, Jacksonville, San Diego, and Seattle. Oakland, New England, Tampa, and Seattle won for $25. On the season Mario was 45-80 for -$675, and Luigi was 46-79 for $315. That isn’t a lot of profit, especially betting $100 per game, but it’s at least on the right side. You might notice that the absolute records aren’t great, but that comes from thinking that the underdog is a good bet sometimes. They don’t need to win often to be profitable, as Luigi demonstrates.
Spread: Mario would have skipped Atlanta-Carolina, Detroit-Minnesota, KC-Oakland, Jets-Buffalo, and San Fran-Arizona. In the other games it went 7-4. Luigi skipped KC-Oakland, Washington-Giants, and SF-Arizona and went 8-5 in the other games. On the season Mario was 97-97-3 and Luigi went 98-93-2. That’s above chance, but probably not good enough to make money. If you were betting $100 a game, you would need to make $9300/98 = $94.90 per correct pick to counter the incorrect picks. That rate corresponds to getting each game at about -105. Since the ‘typical’ game goes off at -110 and more likely outcomes have worse odds (e.g. -115, -120), you would need to get a lot of even bets or better to make that work out.
Simmons: My goal this year was to pick better than Bill Simmons, not for any particular reason besides that he’s a prominent columnist and he publishes his picks for all to see. This week he went 6-10 against his own lines (I’m not sure where they come from), picking every game. Mario went 10-6 and Luigi went 10-6 as well. On the season, I have Bill at 131-119-6, which is above chance. Using my same logic above, he would break even if he got his games at an average of -110.08, which is pretty good. Mario went 116-120-4 and Luigi went 130-105-4. This is pretty fantastic. Luigi could get each game at -123.8 with Bill’s lines and break even. I’m not sure if the difference between Luigi’s performance against Bill and against the Bodog lines is due to the difference in the lines themselves or due to Luigi skipping on Bodog lines that are within a point of his predictions. Usually those marginal games are about 50/50 propositions, but Luigi might have happened to do very well with them this year. “My” picks consisted of a Frankenstein-type approach, using one model to pick in week 1, then using Mario early, then switching to Luigi later when it was obvious that Mario wasn’t ready for prime-time. I went 132-118-6 on the season, beating Bill by two games. Hooray! I’ll still track this in the playoffs, but I’m happy to declare victory now. Bill, if you happen to read this, let me know where you get your lines. I think there’s some money to be made.