Bet the number, not the team. That has long been a motto for sharp bettors. You create a model of one form or another, you decide you'll bet on a given team if their number is, say, -9 or better, and you wait to pounce. You don't just say "I really like how Team A has looked lately," or "Team B is due a good result," or "Team C always loses big games on the road." Creating a model or set of heuristics for yourself and having a general understanding of the distribution of probabilities will probably take you a lot further down the road in the betting world. It's not so much "Bet the number, not the team," as it is, "Bet the number and the team," but that doesn't sound quite as good.
Using publicly available ratings and models can help, too. My SP+ ratings, for instance, generally grade out over 50% against the spread on all FBS vs. FBS games. I post weekly projections on Twitter and track them publicly. Projecting all games, with no adjustments for injuries, suspensions or other breaking news, SP+ tends to get the best of the books, however slightly.
• In 2018, it was 52.8% against the midweek spread, 51.2% against the Caesars closing line.
• In 2019, it was 53.4% midweek and 53.2% against the closing line.
• In 2020, it was only 51.2% midweek but 52.8% against the closing line.
• Early in 2021, despite some early misfortune on really tight lines -- it's 11-16-1 (41%) in games in which SP+ and the spread disagree by less than a point, which should even out over time -- it's back at 52.8% against the midweek spread and 53.4% against the close.
(Note: I'm using SP+ for my examples here because of my obvious and intense familiarity with it. Other systems, including ESPN's FPI, assuredly produce similar results. The website Prediction Tracker can help you find the right mix of spread success and average error that you are looking for. SP+ is at 12.8 points per game of absolute error early this season, which would rank very high.)
You can create your own system of heuristics based on SP+ or other reliable systems. For instance, if you had simply chosen to bet indiscriminately on all 2018-19 games in which SP+ disagreed with the books by at least six points, you'd have won 53% of the time (it's 57% early in 2021). That's not amazing, but it's above the generally accepted break-even point. If you had instead chosen games in which SP+ disagrees by three to five points -- games which are less likely to feature defining injuries or suspensions that SP+ doesn't account for -- you'd have won 58% of the time (55% in 2021).
You probably aren't going to bet indiscriminately on a large batch of games, however. If you were to avoid games featuring a key injury or suspension, or perhaps early-season games featuring teams with new coaches (which provides another layer of unknowns than normal), those averages would rise a bit further.
Of course, there's plenty of strategy and subjectivity to the number the books throw out
Books understand the gut instincts of the casual bettors, and they pay attention to what the sharps are picking, adjusting their lines to induce a certain volume of betting. They aren't relying on a strict formula, in other words, and changing a line from Away Team +1.5 to Away Team +1, Home -6.5 to Home -7, etc., involves specific logic. Like taxes, their movement is designed to drive certain behavior.
Knowing this, I got curious. Using closing lines from 2010-19, I looked into the variability of results for a given line over a large period of time. (I omitted the funky COVID-19 season of 2020 for these purposes.)
I found, for instance, that a home team with a +3 line tends to underachieve by an average of 0.2 points per game with a standard deviation of 16.5 points, a road team at +1 overachieves by an average of 4.5 points (!) with only a standard deviation of 15.2, etc.
Even using 10 years of data, the samples for a given line aren't enormous -- Home +3 had 171 instances, for example, while Road +1 had 87. But even accounting and adjusting for some of the smaller samples, slapping these numbers onto bell curves provided some interesting information. For instance, using the numbers above, you could project that a Road +1 team would cover about 60% of the time. (Home -1, therefore, could cover 40% of the time.)
In all, I found 16 examples of lines in which the team would project to cover at least 55% of the time:
Home -7.5
Home -15
Home -17
Road +6
Road +2
Road +1 (easily the highest odds of the bunch)
Road -4
Road -5.5
Road -7.5
Road -9
Road -10
Road -10.5
Road -11.5
Road -14.5
Road -15.5
Road -17
(It's pretty clear that the public doesn't bet on road teams nearly enough.)
The people who run sportsbooks are pretty smart on average; they probably have a good idea of what lines have which success rates, and some of these lines are hard to come by. Only 18 Caesars closing lines have featured one of these numbers through three weeks in 2021, and only 289 did so over the course of the three seasons from 2018-20, about 6-8 per week.
When you find one of these numbers, however, good things tend to happen. If you were to simply bet on all games with one of these closing lines, then you'd be at 10-8 (56%) this year, and you'd have gone 168-115-6 (59%) from 2018-20. That's quite a different, and effective, definition of "bet the number, not the team."
What if we combined these two ideas -- the model with the number?
What if, instead of those 289 games, we looked only at the ones in which SP+ agreed with the pick from 2018-20? That cuts the number down to 178 games ... and raises the win percentage to 64% (112-62-4).
If you chose only the games in which SP+ disagreed with the line by at least three points, you're down to 96 games and up to 72% (68-26-2). If I were a handicapper with a subscriber service, I'd give those games a fancy name, like FIVE-STAR PICKS or BIG CASH GUARANTEES or something.
(Actually, "Prime Numbers" would work pretty well, wouldn't it? With the math tie and all? Let's go with that.)
Week 4 Prime Number candidates
Sharp bettors tend to seek out good lines. Instead of sitting down and making your picks at a defined point in a given week, you monitor the lines, likely from multiple books, and strike when you find the number you're looking for.
With that in mind, here are the Week 4 games that are both (a) currently within range of one of the 16 lines above and (b) backed up by SP+ projections.
Range: Home -7.5
Marshall at Appalachian State (-7) -- SP+ projection: App State by 7.6 (diff: 0.6)
Cal at Washington (-7.5) -- UW by 14.9 (7.9)
Texas Tech at Texas (-8) -- Texas by 9.3 (1.3)
Miami (Ohio) at Army (-8.5) -- Army by 11.4 (2.9)
SP+'s strange faith in Washington paid off with the Huskies' big (and prolific) win over Arkansas State last week, but it's clear that there's still quite a bit of cushion between SP+ and the books' view of UW.
Range: Home -15 or Home -17
Colorado at Arizona State (-14.5) -- SP+ projection: ASU by 18.3 (diff: 3.8)
Kent State at Maryland (-14.5) -- Maryland by 21.0 (6.5)
West Virginia at Oklahoma (-16.5) -- OU by 20.7 (4.2)
Maryland is another SP+ favorite and, like Washington, the Terps are awfully close to one of the prime numbers.
Range: Road +6
Texas State (+6.5) at Eastern Michigan -- SP+ projection: EMU by 5.6 (diff: 0.9)
Kansas State (+6) at Oklahoma State -- OSU by 3.3 (2.7)
FAU (+5.5) at Air Force -- FAU by 3.6 (9.1)
Nebraska (+5) at Michigan State -- MSU by 2.3 (2.7)
The 9.1-point difference between SP+ projection and spread on the FAU-Air Force game is the largest of the week, and it's usually a sign of a key injury affecting the lines. I haven't found any specific injury news, however.
Range: Road +2 or Road +1
Middle Tennessee (+3) at Charlotte -- SP+ projection: MTSU by 0.7 (diff: 3.7)
New Mexico (+1.5) at UTEP -- New Mexico by 2.5 (4.0)
In both of these instances, SP+ projects the road team to win outright.
Range: Road -4 or Road -5.5
Toledo (-4.5) at Ball State -- SP+ projection: Toledo by 9.8 (diff: 5.3)
UCLA (-5.5) at Stanford -- UCLA by 10.7 (5.2)
Kentucky (-5.5) at South Carolina -- UK by 6.6 (1.1)
Liberty (-6) at Syracuse -- Liberty by 6.8 (0.8)
Even after last week's frustrating loss to Fresno State, UCLA still carries quite a bit of value per SP+. Of course, SP+ is also grading Stanford artificially low because of the Cardinal's Week 1 egg-laying against Kansas State. They chose the correct QB in Week 2 and have looked much better since.
Range: Road -7.5 or Road -9
Boise State (-9) at Utah State -- SP+ projection: BSU by 14.4 (diff: 5.4)
Utah State is 3-0, but based on the key, predictive stats each game produced, SP+ viewed the Aggies' wins at both Washington State and Air Force as games that should have each produced losses by about six points. It has been slow to warm on USU, then, only upgrading it from 114th to 100th overall, and assigning a solid edge to BSU.
Range: Road -10, Road -10.5 or Road -11.5
Clemson (-10) at NC State -- SP+ projection: Clemson by 11.5 (diff: 1.5)
North Carolina (-12) at Georgia Tech -- UNC by 15.0 (3.0)
North Carolina has looked every bit as good as it was supposed to look since Week 1's offensive no-show against Virginia Tech.
Range: Road -14.5 or Road -15.5
Louisiana (-13.5) at Georgia Southern -- SP+ projection: UL by 18.7 (diff: 5.2)
Buffalo (-13.5) at ODU -- UB by 17.6 (4.1)
Honestly, I have no idea why UL is favored by less than two touchdowns. You might want to hop on this one at whatever number.
Range: Road -17
Hawaii (-17) at New Mexico State -- UH by 20.8 (3.8)
NMSU actually has overachieved against projections a bit this year, so I'm less sure about this one than SP+. Either way, if the number is good, the number is good.