"Well if they're big and you're small, then you're mobile and they're slow. You're hidden and they're exposed. You only fight battles you know you can win. That's the way the Vietcong did it. You capture their weapons and use them against them the next time."

Brill Lyle, Enemy of the State (1998)

Winning money betting on the NFL is hard. Due to its popularity, football invites more American bettors (with more money) than any other sport. Sportsbooks are incentivized to offer the sharpest lines on the biggest markets, and bettors with the highest volume help sharpen the lines with their action. This is why you’ll see people call the major NFL markets – spreads, moneylines, totals – the most “efficient” markets. They are hard to beat.

Regardless, betting on sports is fun. Having a financial stake in the outcome of games makes them more interesting to watch. The dopamine hit that follows a winning bet is real. You don’t need to have an edge to make a bet. Doing it for fun is perfectly fine, and probably advisable for most people. But winning is more fun, and allows you to play the game longer on your given budget.

You won’t win in the long run without a true edge. You might get lucky and go on a hot streak, but that could set up unreasonable expectations and lead to disappointment. So I want to talk about – and demonstrate – this concept. An edge is created by data, information, an angle, or perspective, that allows you to make a bet with a higher true probability than the implied probability of the odds offered by the Sportsbook. In other words, you only have an edge if you factor in something the market does not. The more efficient the market, the harder that is to actually do. But it’s not impossible.

Here is my simple two-step process to finding an edge. Then I’ll demonstrate with some examples.

Step 1: Understand the Betting Market

A lot of people make bets because “the public” or “the media” spouts an opinion on a team that they think is wrong. But you aren’t betting against the guy who Tweeted that the Bengals are clearly better than the Chiefs because they beat them twice last year, no matter how many likes his Tweet gets. And you aren’t betting against ESPN. Sometimes a majority opinion can influence the market through a massive volume of bets, but for the most part, these off-base takes do not move the needle in the betting markets.

You are actually betting against (1) the Sportsbooks' trading rooms and (2) professional bettors. Both of these groups of people invest substantial time and money into understanding the teams, situations, injuries, and other factors present in every game. They run statistical models to predict outcomes and hone those systems with machine learning. They have been doing it for decades. The more you know about the kinds of inputs these models use, and the type of data bookmakers and professionals use or find significant, the better you can understand the market.

Step 2: Differentiate

Once you understand the market, you can look for an edge. To find an edge, you need to be able to identify flaws in the betting market's biggest inputs and exploit those flaws. I like to use the market’s own data against itself by exploiting gaps in major statistical metrics that inform the largest bettors and Sportsbooks. Let’s use some examples from the Futures market to illustrate, since those are the bets I'm looking at in June.

Sophisticated bettors do not start with last year’s win total as a baseline projection for a team. You’ll hear “The Dolphins won 9 games last year and vastly improved so Over 9 wins is a lock.” That’s not how sophisticated bettors think, and so this type of thinking is typically not reflected in the market. Instead, bettors use more advanced metrics, such as pythagorean wins and strength of schedule. Most successful bettors look beyond these metrics of course, but they are used widely enough to impact the market, and that's what we are looking for.

Pythagorean win totals capture a team's performance based on total points for and total points against, instead of wins and losses. The number attempts to project how many wins that team should have won based on point differential, and does a good job of evening out the impact of a lot of close wins or close losses. Many bettors have modified the basic Pythagorean formula, but even modifications will suffer from similar flaws. Namely, point totals do not accurately reflect the true performance of a team. If you have a metric that evaluates performance in a game independent of final scores, you can use that metric to generate a more accurate win total. Where this win total differs most from the team's pythagorean win total, you may have an edge.

Bettors typically provide context to a team's Pythagorean win total by measuring strength of schedule. These strength of schedule comparisons often involve using team win totals. That’s highly flawed. Most sophisticated bettors are not using this number, but use power ratings instead to form their own strength-of-schedule ratings. Yet most of the content in this space – which helps drive bets by a high volume of bettors and can implicitly influence sharp bettors – involves win totals.

But even strength of schedule calculations based on power ratings are flawed because teams aren’t consistent throughout the season. Catching the Packers without Aaron Rodgers should not count the same as a fully healthy Packers team. Facing a team missing eight starters due to Covid virtually assures a win regardless of how good that team is on the season. Even situational factors matter, such as playing a road game on Thursday after playing a long overtime game on Sunday (like the Ravens did in 2021). So you have to factor in true strength of schedule at the time of the matchup.

My Method

Instead of using Pythagorean wins, I evaluate every game holistically and assign score values based on each team's holistic Effectiveness. Instead of using raw points for and points against, I use my own metrics to calculate Adjusted Wins, or how many wins the team would typically earn playing that level of football without variance. Over the course of the season, small gaps in individual results can compound to create a meaningful edge in differentiating from the market.

Because we are looking for edges over market metrics, let's look at the biggest gaps between my Adjusted Wins and Pythagorean Wins. My numbers have incorporated all strength of schedule and hidden schedule quirks as well. I ran the numbers on every team and the two biggest outliers in 2021 (one in each direction) were the Raiders and Dolphins:

Miami’s win total and point margin were both inflated by extreme in-game variance and favorable situational strength of schedule quirks. For example, in Week 1 they beat New England 17-16 despite a very unfavorable Effectiveness Rating (6.12-5.11). They had fewer first downs (24-16), were less efficient on a per-play basis (5.6-5.0), and had fewer red zone trips (4-2). The Patriots moved the ball much more easily, converting 11 of 16 third downs to the Dolphins’ 4 of 11. But the Patriots settled for three field goals and lost two skill-player fumbles. The first was on a 9-yard catch on 1^st down. The second was on 1^st down on the Miami 11-yard line with 2 minutes to go, down by 1 point. The fumble allowed Miami to secure an unlikely win.

They had similar variance in Week 18 against the Patriots, in a 33-24 win despite another unfavorable Effectiveness Rating (5.56-4.55). This time, The Patriots outgained the Dolphins significantly (6.4 yards per play to 4.5), but Miami squeaked by with a win due to three Patriot turnovers, two of which were returned for touchdowns, and a heavy penalty advantage.

Outside of these two misleading final scores, the Dolphins had zero games in 2021 where they lost outright despite outplaying their opponent in terms of Effectiveness. They won every game they deserved to win, which is actually quite rare for an entire NFL season. They avoided negative variance as much as they benefitted from positive variance.

On top of that, they played the Saints down three offensive line starters with Ian Book, the Jets down two offensive line starters with Zach Wilson, the Giants with Mike Glennon, the Panthers with Cam Newton off the street, the Jets with Joe Flacco, the Texans with Tyrod, and the Ravens down two offensive line starters on 4 days rest after playing nearly 100 snaps on Sunday. These schedule quirks inflated both their strength of schedule and Pythagorean wins.

The Raiders, on the other hand, caught the short end of the variance stick in several games. Despite an ostensibly lucky record at 10-7 (including 4 out of 4 overtime wins), their actual performance was not reflected in their final scores. I'll highlight five example games:

• They beat Miami by 3 points in overtime but significantly outplayed them overall (5.17-4.64). They edged Miami in first downs (28-22), yards per play (6.1-4.2), and red zone trips (4 to 1). They overcame a pick-6 deep in Miami territory and a turnover on downs in their own territory. Variance went against them and it should have been a more comfortable win.
• They lost to the Giants 23-16 but were the more Effective team overall (5.30-4.62). They racked up more first downs (24-16), better yards per play (6.0-4.6), and more red zone trips (6 to 2). But three turnovers, including a pick-6, some unfortunate high-leverage failures, and a missed 25-yard field goal set them back.
• They lost 32-13 to the Bengals in what was actually a fairly close game. The Bengals dominated time of possession in large part due to a massive penalty discrepancy (7-77 to 1-5) and two more turnovers by the Raiders. But it was 16-13 with just over 5 minutes left in the 4th before a late flurry of fairly meaningless points. These teams eventually played a rematch in the playoffs that more closely reflected the respective quality of the teams, as the Raiders had a chance to tie that game late and lost by a touchdown.
• They got blown out by Kansas City to the tune of 48-9 but experienced massive negative variance. They had 5 turnovers, including 4 lost fumbles, 2 on first-down catches and 1 on a kickoff that was returned for a touchdown. Kansas City wins this game almost every time, but the 39-point differential skews the Raiders' pythagorean win total.
• They only beat DEN 17-13 but absolutely dominated the game (5.78-3.78). With an edge in first downs (22-8), yards per play (5.1-4.0), red zone trips (3-1), the Broncos only stayed in it due to three turnovers. The Raiders finished the game kneeling in Denver territory, so it could have easily been a bigger win if they needed it to be.

Conclusions

With very similar records last year (10-8 for LV and 9-8 for MIA), similar win total projections for this year (around 8.5 wins), and similarly low pythagorean win totals, these two teams make for an interesting comparison. Both improved in the off-season. Yet both are projected by the market to win fewer games than last year. Schedule has something to do with it of course, but so does last year's results. In Miami's case, this element is justified. But for the Raiders, it arguably is not.

Now, I don't necessarily recommend betting on Miami's win total under and Las Vegas's win total over. I put a lot more into my analysis before firing off a bet. But I use these outliers to demonstrate one way that a bettor can find a potential edge over the market. Given that any edge in an efficient market will be small, I recommend only making a bet when you can compound a few edges in one bet. And hopefully, this gives you an idea of how to go about doing that for yourself.

Now for my pitch. This all takes a lot of work. Below, Members have access to a table with every team's Pythagorean Win total and Adjusted Win total from 2021. This is just a small snippet of the actionable information I provide to Members. If you're interested in learning more about what I do, you can join my free mailing list, or if you'd rather pay for the fruit of my work, you can become a SharpClarke Member and get all my picks and analysis. You can find both options on my Membership page.