How To Fix Fantasy Football Scoring With Simple Advanced Stats

Image for article titled How To Fix Fantasy Football Scoring With Simple Advanced Stats

With only one touchdown in Week 16, Aaron Rodgers ranked 15th in fantasy points among quarterbacks during fantasy football championship week, likely costing many a Rodgers owner a fantasy title. However, Rodgers' performance was much more valuable on the real football field, where his contributions added an expected 12.2 points to the Packers' point total (6th among QBs).


This discrepancy is due to the fact that most "standard scoring" fantasy systems are grossly out of line with what plays are worth in real games, with the largest disparity being touchdowns. As most leagues count 25 passing yards as a point and touchdowns as four, the inch between the 0.1 yard line and goal is inexplicably valued as being worth equal to the 100 yards that precede it, and the randomness of goal line decision making often distortedly awards points to players involved in the play that just happened to be the final one of the drive.

This is obnoxious when "goal-line" running backs vulture lucrative scores from higher-profile teammates, as when last week's vulture-in-chief was Washington's Darrel Young, whose 2 carry, 2 yard, 2 TD line stole 12 points from Alfred Morris and placed Young 20th among RBs in fantasy points. it also makes fantasy football much more like a roulette wheel than a true representation of football.

What would scoring by EPA look like?

A more accurate way to allocate fantasy points would be to value events based on expected points added (EPA), a metric popularized by the website Advanced Football Analytics, which measures how many additional points a team is expected to score given a change in situation. For example, a team with 3rd-and-5 at their own 42 would have an expected net point advantage of 0.74 points, but after a 30-yard pass to set up 1st-and-10 at their opponent's 28, they would be expected to score 3.47 points—meaning that pass was worth 2.73 points.

As far as the touchdowns are concerned, once a team has first-and-goal at the one yard line, their expected points are already very near seven, so the act of actually crossing the goal line does very little in terms of increasing a team's expected point total.

Below is a hypothetical EPA-based fantasy scoring system, allotting points based on the average EPA of particular events. There are two EPA columns: "EPA" is the expected actual football game points added for each play (the stat described above). "EPA adjusted" is those same EPA numbers, but all scaled up by 1.43 so that 10 rushing/receiving yards = 1 point in both systems and one can use that as kind of a baseline to easily compare other values, e.g. TDs are only worth 0.78 points / 7.8 rushing or receiving yards in this system. To be clear, these are stand-in values based on average EPA, not the actual values from the games these players played.

Event Standard EPA Adjusted EPA
Passing TD 4 0.52 0.37
25 passing yards 1 1.00 0.70
Passing first down 0 0.29 0.14
Interception -2 -4.99 -3.49
Sack taken 0 -1.99 -1.39
Incomplete Pass 0 -1.06 -0.74
Rushing/Receiving TD 6 0.78 0.55
Rush/reception for no gain 0 -1.06 -0.74
10 Rushing/Receiving yards 1 1.00 0.70
Rushing/Receiving first down 0 0.29 0.20
Fumble lost -2 -4.57 -3.20
FG Made (50+ yards) 5 2.71 1.90
FG Missed (50+ yards) -1 -2.43 -1.70
FG Made (40-49 yards) 4 1.43 1.00
FG Missed (40-49 yards) -1 -3.43 -2.40
FG Made (0-39 yards) 3 0.29 0.20
FG Missed (0-39 yards) -1 -3.57 -2.50

*The standard scoring ratio of 25 passing yards = 10 rushing/receiving yards, passing TDs = 67% of of rushing/receiving TDs are kept constant.


**For those wondering, as far as PPR (point per reception) scoring, there is no EPA difference between gaining yards via a pass or a run, so no points should be rewarded for receptions. If anything, receptions should carry some small negative value, as they use a down. This is the same for rushes, which should carry slightly greater negative value given that the average depth of target for a reception is greater than the average starting rushing position a few yards behind the line of scrimmage. This is partly accounted for by the "rush/reception for no gain" category.

These are very crude estimations, as yard and first down EPA values can vary by situation, but should be far more accurate in assigning value than standard scoring.


So how would this more accurate scoring system play out in a potential championship week matchup? Here's an example of a hypothetical matchup (defenses excluded) from last week using standard scoring:

Position Team A SS Pts. Team B SS Pts.
QB Cam Newton 22 Aaron Rodgers 14
RB Marshawn Lynch 23 Reggie Bush 15
RB C.J. Anderson 19 Damien Williams 12
WR Odell Beckham Jr. 26 Calvin Johnson 10
WR Torrey Smith 17 Julio Jones 10
WR Riley Cooper 17 DeSean Jackson 12
TE Antonio Gates 21 Charles Clay 11
FLEX Matt Asiata 20 Kenny Stills 6
K Sebastian Janikowski 15 Matt Bryant 15
Totals 180 105

It's a dominant blowout victory for Team A in this system, but here's how this championship week matchup plays out with the EPA-based scoring system:


In this system, Team B has a modest victory. This difference in outcomes is largely due to the touchdown discrepancy, as Team A's players lucked into 15 scores (an extra 88 points) compared to Team B's only finding paydirt three times (16 points). In standard scoring, this TD deficit between Team A and Team B represents a 72-point discrepancy, but in the EPA format, it's only a 9.4-point difference.

To illustrate how hollow some of Team A's TDs were, take C.J. Anderson, who got a handoff at first-and-goal from the 1 (5.96 EPA situation) and converted for the score (now 7.00 EPA - 0.34 EPA for the Bengals after kicking off). This 1-yard rush represented a mere 0.70 in EPA, and Anderson had done nothing earlier in the drive to get the Broncos to that point of the field, the vast majority of EPA instead coming from Omar Bolden's 77-yard return to start the drive (4.51 EPA).


Not all of Team A's short TDs were as worthless as Anderson's, though. The Eagles were at 3rd-and-goal from 3 (4.56 EPA situation) when Riley Cooper caught a 3-yard TD (7.00 EPA - 0.34 EPA for the Redskins after kicking off), meaning his 2.10 EPA TD was worth triple Anderson's, though still worth the equivalent of only 30 receiving yards (compared to 60 yards in standard scoring).

Team A also had some truly valuable TDs, such as Odell Beckham Jr.'s 80-yard bomb from 3rd-and-10 from the Giants' 20 (-0.57 EPA situation), which increased the Giants' EPA to (7.00 EPA - 0.34 EPA for the Rams after kicking off), making his TD worth 7.23 EPA, roughly 3.4x the value of Cooper's score and 10.3x as valuable as Anderson's. These long, valuable TDs are relatively rare, though, as the average distance of TD-scoring plays is only 15.6 yards, meaning most TDs are of the short, not-so-valuable Cooper/Anderson variety.


Could this actually work?

Again, it's important to note that, unlike the EPA values for the TDs described here, our EPA conversion chart at the top of this post is absent from context and does not take into account the length of the play or the down. It is simply a rough average of all plays. But if a fantasy provider were to come along and run with this idea, it would be relatively simple to license stats from Advanced Football Analytics— Brian Burke already publishes the top performers by position each week—or create an approximation of its own.


The other major EPA scoring discrepancy stems from the fact that fantasy football only uses count rather than rate stats, so players like Asiata who are terribly inefficient but rack up a ton of touches aren't penalized for repeatedly running into the line for no gain and stalling drives. In Week 16, this was most evident for Marshawn Lynch, who had four such no-gain carries, which totalled to -2.51 EPA, a penalty that severely hurt the Seahawks but would go unpunished in standard scoring fantasy systems. (And to be fair, this was largely overshadowed and more than made up for his 79-yard Beastmode run.)

The stuff you'd end up losing is the wild swings on fluke players grabbing a few garbage touchdowns, or someone like Ben Roethlisberger beasting out for 45 points—a figure that would come down significantly if pegged to actual EPA. Overall, would make for a game that is more realistic, but perhaps also less exciting, which could be a tough sell—especially with interest in full-season fantasy waning in favor of one-week games like Draft Kings. Then again, given the colossal value of touchdowns now and they way they immensely shape matchups, watching games purely from a standard scoring fantasy perspective can be somewhat unexciting if your players are at midfield or in their own territory, far away from the lucrative six-point end zone.


Fantasy football is already a game that depends very heavily on luck, and using such a grossly inaccurate point system only amplifies that randomness. Whether that randomness is good for the game or not is another question.

Jim Pagels is a regular contributor to Forbes. You can follow him on Twitter at @jimpagels