The NFL is going through an analytics revolution and at the forefront of it is the emergence of Expected Points Added (EPA). This article looks at the upside – and downside – of EPA in an effort to get you familiar with the statistic you will be seeing more of in this season than in any season before.
What is Expected Points Added?
The most common NFL statistic is yards gained. Since the National Football League was founded 100 years ago, yards gained has been tracked for every game.
• How many rushing yards did a team gain?
• How many passing yards did a team gain?
• What was the difference in total yards?
The problem with yards gained is that the metric is flawed as time remaining and game state will dictate the play calling of a team. For example, a team leading by multiple scores is more likely to play conservatively late in a game and sacrifice yards in exchange for time off the clock.
Expected Points Added was created to address the misleading nature of yards gained by putting value on down and distance as well as field position
Any team trailing tends to play more aggressively, and will be more willing to increase risk in exchange for yards gained.
Expected Points Added was created to address the misleading nature of yards gained by putting value on down and distance as well as field position.
For example, a gain of five yards on a 3rd down and 10, and a gain of five yards on 3rd down and 3, will both register as +5 in yards gained. However, the five yards gained with three yards to go is more valuable than the five yards gained with 10 yards to go, as it results in a first down and extends the drive.
Another example would be two plays that gain equal yards in equal situations. For example, if a team runs plays on two different series on 1st down and 10 that each gain four yards, the second play of each series results in the same situation of 2nd down and 6.
With expected points added, depending on field position, the success of the play will be graded differently. A gain of four on first down from a team’s own 30-yard line is much less than gain of four on first down from an opponent’s 15-yard line.
How is Expected Points Added calculated
Calculating expected points added is based on next points scored for all plays. By adding up the next points scored for and against the offensive team, the net point advantage can be determined for any down and distance.
For example, if a team runs a play on 1st and 10 from their own 20-yard line, the value of 1st and 10 from their own 20 is +0.4 points. If a team runs the football and gains 30 yards for 1st and 10 from midfield the next play from the 50-yard line is worth +2.0 points.
The difference between +0.4 and +2.0 is 1.6 points, which means that play from the 20-yard line had an expected point added value of 1.6 points. That point value will then be assigned to the players involved in the play.
The easiest way to access EPA data is to use NFLscrapR which is a library built for the programming language of R, and scrapes NFL play-by-play data. Ben Baldwin has written a brilliant tutorial to get up and running here.
On Pinnacle there is also a wealth of resources for using R.
Expected Points Added usage
When expressed as expected points added per game and plotted by offence and defence combined, the metric can be quite predictive of success in the league.
Expected Points Added can be used as a grading tool for talent within a roster
Last year the teams that finished with positive EPA per game on offence and defence included the Chicago Bears, Indianapolis Colts, Seattle Seahawks, Los Angeles Chargers, New England Patriots and Los Angeles Rams – each of which made the playoffs and two of which played for the Super Bowl.
Looking at offence alone, the top three teams were Kansas City, New Orleans and Los Angeles Rams who topped the league for most wins.
Expected Points Added can be used as a grading tool for talent within a roster. Each player who participated in a play will be assigned an EPA value.
Get the best NFL odds
Pinnacle's great odds and low margins cannot be rivaled!Sign up hereLog in here
When assessing replacement level talent or the contribution of individual players to their team, looking at EPA can identify players which the market over or under values.
For example, Andy Dalton had an EPA per dropback of 0.47 in 2018. Once he was injured, replacement Jeff Driskel and an EPA per dropback of only 0.35. The difference of 0.12 EPA per dropback is a great way to evaluate replacement level talent.
Expected Points Added can also be used to grade the effectiveness of NFL play calling on a team by team basis. For example, in 2018 the average passing play in the league was worth 0.05 EPA per play.
In comparison, the average rushing play was worth -0.06 EPA per play. While most NFL data will point towards passing being more efficient than running in almost all cases, it is important to note when teams run the football - and how frequently.
Downside of Expected Points Added
There are many parallels between Expected Points Added in football and Expected Goals (xG) in soccer betting. The biggest one is what the data does not tell you.
With Expected Goals in soccer, it tells you nothing about the defenders and the goalkeeper. A penalty kick taken from the spot results historically in 0.77 expected goals, but that number can change dramatically based on which keeper is in goal.
With the NFL being so matchup–based, where schedule difficulty is of the utmost importance in betting the league, using EPA week to week in betting can be misleading.
The initial reaction for many people seeing EPA for the first time is to back teams that are underperforming and fade teams that are over performing. Without accounting for the strength of opponent, bettors could end up making poor decisions.
The measure of expected points is linear. However, because winning a game by any scoreline proves equally valuable, teams will intentionally sacrifice yards in exchange for time as mentioned above.
The baseline for expected points is calculated on game situations in the first and third quarter where the clock will not be of influence, and only when the game is within ten points, to best account for neutral play calling tendencies.
As the calculation of EPA depends on the difference between down and distance, unusual down and distances prove to be extremely noisy.
What does the future look like for Expected Points Added?
There will be more expected points allowed discussion in 2019 than any season prior and I have no doubt that the process will be refined and developed into something useful in the future.
In its current state, I see much more value in individual player evaluation – especially for depth – than I do team performance and market projection.
Looking back on the trajectory of expected goals, as more and more data became available, innovation was quick to follow. Expected Goals now dictates the movement of many soccer leagues around the world on a match by match basis.
Considering how much recreational influence there is in the NFL betting markets, it is unlikely that EPA will have a noticeable impact on NFL markets in the immediate future, but the foundation it provides for creativity is very exciting.
To make the most of our best value odds make sure to read more of Pinnacle’s expert betting advice.