Objective Win Odds: The Foundation For Winning Football Pick’em Contests

If you want to win pools, you have to pick winners correctly and have a solid plan to avoid just going with your gut.

Win Odds and Pool Picks

The key to winning pools does not involve picking the Cowboys every week (Photo by Matthew Pearce/Icon Sportswire)

What is the key to winning pick’em pools?

Get closer, so you can hear this. It’s picking more winners than all of your opponents.

That may sound really simple, but it’s true. Of course, identifying enough of those winners is not easy, so we are going to talk about what you need to do in order to make the best decisions.

Before we get to any other game theory, or looking at how your opponents are picking, at the root of any strategy to beat your opponents, you must have good underlying objective win odds and projections for the games.

What Do We Mean By Win Odds?

Win odds are simply an objective estimate of a team’s chances of winning a particular game.

Our custom pick’em picks are based, at their core, on win odds that we create by combining power rating, team stats, betting market info, and other objective data. Are they always right? No, of course not. Individual games feature random events and outcomes. And sometimes changes in personnel, injuries, or other factors make the estimates a little more uncertain, at least until new players or strategies play a few games, and their value becomes more clear.

But in the aggregate, our win odds are fairly accurate, with a large sample size of games. For example, looking at the 2019 season, there were 72 games where our win odds put a team as having between 50% and 59% win. The actual win percentage of those teams in those games? 54.9%. So while there was plenty of variation in individual game results, on average, the favorites won as often as we expected them to. That’s what you want to see from objective win odds.

Let’s go through some of what goes into having objective win odds.

Models, Models, Models

We combine several different predictive models to make our win odds projections. Combining different types of models improves predictions because each model has its own strengths and weaknesses.

For example, a Power Ratings model is going to do a pretty good job by looking at past game scores to estimate the relative strengths of the teams. But it won’t be as responsive to things like injuries and changes in personnel (such as a new quarterback starting).

A Similar Games model, meanwhile, might be able to pick up relevant predictive information related to rest and travel by looking at the most similar matchups in the past based on the power ratings of each team, the relative rating difference, the location of the game, and how much time off each team has had.

We also have a Decision Tree model, which is a machine learning algorithm that looks through hundreds of input variables. In addition to rest and travel information, rolling averages of various stats and ratings are included. It also incorporates betting market information, such as money line odds, and game point totals.

Betting Market Odds

The Betting Market Odds are another good source of objective win odds. These are set by sportsbooks who have a lot of money at risk if they set a poor line, and are “bet into shape” by sharps as game time approaches. If the betting market is making a certain team a favorite, there’s usually a pretty good (and measurable) reason for it.

In addition to point spread lines which show the favorite, there are also money line odds on each game, from which you can decipher what the sportsbooks think the specific win odds for that game are. You do have to remember to remove the “vigorish” from the money lines, otherwise you will be estimating that the two teams combined have about a 105% chance at victory.

We incorporate the latest betting market information into our win odds, though we do not simply match the market, and do take positions on some NFL games. But incorporating the latest win odds coming in from the betting markets allows us to be responsive to news and injuries. For example, if news comes in on Sunday morning that a starting quarterback will miss the game, and it swings who is favored, that is valuable information if most of your pool is slower to react.

Objective Win Odds are Better Than Your Gut

The alternative to having an objective system to predict the game winners is to rely on your own gut or intuition. Even smart people can be subject to blind spots and biases. Year after year, we see that famous media and sports prognosticators cannot outperform our predictions. Sure, in a given year a few might, but over time, they cannot replicate it.

Chances are you also have successfully picked some upsets in your life. They happen in the NFL on a weekly basis, and if, say, the favorite has win odds of 65%, you can successfully predict an upset 35% of the time. These successful picks (along with a bit of selective memory) can create the illusion of being smarter than the market. But over a period of several weeks or seasons, most people will not beat the market when it comes to predicting underdogs to win on a consistent basis.

Forcing Upset Picks

Most people, when making their pool picks, probably do have some general sense of who the better team is. They might even glance at the point spread lines each week. But the allure of picking upsets and of nailing an upset call are powerful. As a result, most entries in pools pick too many big upsets, and it costs them.

Common Errors by Going to the Gut

There are also some areas where we see the public routinely make mistakes compared to the betting markets and objective win odds models.

The public tends to overestimate the impact on close wins or close losses on the predictive quality of teams, and thus has some bias toward teams that “just know how to win” and against those that lose close. That is, they value the win-loss record more than “how teams won.”

The public may also not properly value the impact of home field advantage when the “better” team is traveling to the “worse” team, especially when objective measures show the team quality is not that far apart.

Finally, when compared to objective systems that study such things, the public is far more responsive to and biased by recent results, often overreacting to what happened the week before.

Having objective systems that account for what is truly predictive, then, are important to counteract these biases and give yourself the best chance of winning the pool.