Week 1 - Week 2 - Week 3 - Week 4 - Week 5 - Week 6 - Week 7 - Week 8 - Week 9 - Week 10 - Week 11 - Week 12 - Week 13 - Week 14
Sports betting is mathematically designed so that you lose money, if not right away then over a period of time. Streaks of good fortune are almost always more-than-eliminated once the odds — ever so carefully and magnificently stacked against you — have enough time to do their thing. But sometimes, at certain points along the way, it can be tempting to believe in your own hot hand, or think you found some magic.
This year's NFL value picks offer one such temptation. (I encourage you to resist.) As weekly readers are familiar, these value plays apply aggregate win and cover probabilities across the 50+ models found on The Prediction Tracker to prevailing moneyline odds and point spreads. We thus identify which picks carry the highest expected value in each week's games. The picks are as reliable as your faith in the models.
When picking against the spread, teams with the highest expected value are basically the ones with the highest probability to cover. The only thing that reshuffles that deck are the odds at which a given point spread exists. You usually have to risk $110 to win $100 (-110) — but sometimes it's $120 to win $100 (-120), which drags down expected value, or $100 to win $100 (EV), which provides an expected-value boost.
On the moneyline side, positive expected value exists when the models believe that a certain outcome carries a higher probability than the odds imply. Last week, the Arizona Cardinals were +650 to win at the Green Bay Packers. Those odds implied a 13.3% chance for an Arizona victory, whereas the models saw Arizona as having a 20.9% chance to win — which they did, paying bettors $650 for every $100 they risked.
It was this column's top moneyline value play of the week and second most profitable victory of the year, behind only Buffalo +1000 at Minnesota in Week 3. And it supercharged the column's running moneyline record to 22-43 — good for a hypothetical return of $1,534 if we assume $100 risked on five games per week. (All but two picks were modest to heavy underdogs — hence the high profit via a losing record.)
We'll save the real summary math and analysis for the end of the season. But if you're keeping score, that's $1,534 in hypothetical profit on $6,500 in money risked, which is good for a return of 23.6% since Week 1. Against the spread, the top five weekly picks are 35-27-3 (+$530). Assuming an average of $110 risked per game, that's $530 in hypothetical profit on $7,150 in money risked, or a net return of 7.4% since Week 1.
My disclaimers are that (a) this is a small sample size, (b) the floor will soon fall out, and (c) even if it doesn't fall out this year, the same method will get punished next year or beyond. So I caution against believing any of this. But it's fun to continue to track (and feel good about when it works). So that's what we'll do.
Probabilities to cover and win are sourced from the 50+ models on The Prediction Tracker. Against-the-spread expected values assume bets to win $100. Moneyline expected values assume bets that risk $100.
I. Against the Spread — Top 5 Value in Week 14
As I've noted since the outset of the season, my goal is to remove manual overlay from this process. The situations that screw that up are major injuries, particularly to QBs with shitty backups, which the odds and lines capture but the models largely do not. Washington carried high expected values in Weeks 12 and 13 because the models liked the team at given levels, without realizing that Colt McCoy was under center.
Over those two weeks, Washington made the top five against-the-spread (ATS) value list twice (went 0-2) and the top five moneyline value list once (went 0-1). Each time I offered a disclaimer but left them in. Had I subbed them out, there would have been no ATS replacement in Week 12 — there were only four teams with positive expected value — and Arizona +14 would have made the list last week and registered a win.
Now that backup Colt McCoy is also injured and Mark Sanchez is at the helm — Sanchez hasn't been a full-time starter since 2012 with the New York Jets — Washington's numbers are even more messed up. They're 3.5-point underdogs at home against the New York Giants and +160 to win the game. The models love them at those levels. But that isn't fair to the spirit of the system. So this week, I am taking them out.
(At season's end, I'll likely show summary results with Washington post-Alex Smith injury and without.)
1. CLE +1.5 vs. CAR, EV, 53.0% chance to cover, $5.9 expected return on $100 wager
2. SEA -3.5 vs. MIN, EV, 52.6%, $5.2
3. CIN +14 at LAC, -110, 54.6%, $4.6
4. ATL +6 at GBP, -120, 56.4%, $4.2
5. HOU -4.5 vs. IND, -110, 53.1%, $1.4
TOP 5 VALUE
Last Week: 3-2 (+80), Season: 35-27-3 (+530)
II. Money Line — Top 5 Value in Week 14
Washington would have owned the number two spot on this list, but again, they've been excluded. (Had Washington been excluded in Week 12, Arizona would have taken their place and lost. So keeping them in despite the quarterback situation did not have any impact on the running moneyline tallies down below.) Cincinnati (+700) is lurking in similar territory as Arizona last week — 13% by the odds, 21% by the models.
1. CIN +700 at LAC, 13% prob to win (odds), 21% prob to win (models), $65 expected return
2. OAK +500 vs. PIT, 17%, 22%, $34
3. ATL +200 at GBP, 33%, 42%, $26
4. MIA +300 vs. NEP, 25%, 30%, $22
5. BAL +250 at KC, 29%, 33%, $15
TOP 5 VALUE
Last Week: 2-3 (+530), Season: 22-43 (+1534)
All but two selections were modest to heavy underdogs
III. Searching for Value in NFL Futures Bets
After a couple of weeks of broken promises, I am finally back with an updated expected value futures grid. To the unfamiliar, this grid attempts to show where positive expected value lies when evaluating teams' odds to win the division, conference, and Super Bowl. Positive expected value shows up when teams' probabilities to win those crowns — according to three leading NFL prediction models (ESPN FPI, Football Outsiders, and Massey-Peabody) — outweigh the probabilities implied by their current futures odds.
As you can see from all of the red below, we haven't missed much by waiting a few weeks to revisit this. (Although who knows, maybe there was a mispriced gem lurking in the interim. But probably not.) As far as Super Bowl futures are concerned, the Kansas City Chiefs are the only team with positive expected value, and it ain't much. On average, the three models believe the Chiefs have a 20.3% chance to win the Super Bowl, while their +450 odds to do so imply an 18.2% chance. So there could be a hint of value there.
Kansas City offers a similar hint of value at +200 to win the AFC, with a 39.1% chance according the models versus a 33.3% chance according to the odds. (Remember, Kansas City released star running back Kareem Hunt after a video surfaced of him physically assaulting a woman at a hotel in Cleveland. This caused the Chiefs' odds to inflate a bit since last week, as observers wonder how the change will affect the team.)
Carolina has some theoretical value at +10000 to win the NFC, but that's mostly due to rounding error. The models give them a 1.2% chance to the win the conference, whereas the odds imply 1.0%. Other than the Colts at +9000 to win the AFC, which is fairly priced (1.1% chance), every single other conference or Super Bowl futures bet is mathematically configured so that the probability implied by the odds is higher than the average probabilities generated by the models — meaning they all carry negative expected value.
On the division side, Kansas City (-700, implied probability of 87.5%) and Houston (-800, 88.9%) are fairly priced. Pittsburgh offers the most value among favorites — the models give them an 83% chance to win the AFC North, while their -300 odds imply 75%. Not much, but not bad. Cincinnati and Green Bay price well as rounding-error long-shots, and Washington looks misleadingly good for reasons described above.
Westgate odds as of Dec 5. Vegas Insider division odds are added in for the Broncos, Lions, Texans, Colts, and Titans, Probabilities reflect the average across ESPN FPI, Football Outsiders, and Massey-Peabody. ESPN FPI does not publish probabilities to win the conference, so they are excluded for that category.