The Fantasy Football Mythbusters: Flip the (Game) Script

Unlock Premium Tools + Bonus Episode

Join the #FootClan

In our previous entry in the Fantasy Mythbusters series, we explored the impact of weather on fantasy performance. This time, we will analyze another factor that enters into the weekly start-sit calculus: game script.

As the name implies, this essentially means the narrative of how a game plays out. We’re all familiar with common prototypes of NFL games: defensive slogs, high-scoring contests, blowouts, etc. For fantasy managers, the idea is to target players that are involved in positive game scripts and avoid players in negative ones. For example, you might prefer a running back on a heavily favored team, since the offense will likely be running the clock out in the second half. In the same game, you might target a WR on the underdog side to try to capitalize on all of that ‘garbage-time passing production generated late by trailing teams.

Let’s dive deeper into how much the ‘script’ actually impacts fantasy outlook. Data, unless otherwise specified, comes from nflfastr.

Is it Predictable?

First and foremost, for the game script to be a useful factor in fantasy decisions, it has to be predictable. Even if there is a significant benefit to targeting positive game script, this is no use if you can’t accurately forecast when these scripts will occur.

To check this predictability, let’s track the spread of the games (defined here as expected home score minus expected away score) from nflfastr vs. the actual result of the game for every single regular-season game since 2000:

  • Unsurprisingly, these spreads have a winning hit rate: the favored team wins 66% of the time. The pre-game spread has a .42 correlation with the final, actual spread of the game.
  • Spreads are even more accurate in heavy-favorite games: the team favored by 10 or more points wins 85% of the time. In a similar vein, home teams are favored by an average of 2.4 points (this is a good gauge for home-field advantage), which is why there is the most density in the top right quadrant of the chart.
  • There are some interesting divisional observations: spread predictions are slightly better in division games (66.7% vs. 65.6%), likely because divisional teams face each other more often and there is more data to make predictions with. Division games are also tighter (.25 lower spread on average) and lower scoring (1.6 points).
  • Two games are tied for the ‘greatest upset’ in this sample; both saw home favorites of 17.5 points lose to the visiting team. The first comes courtesy of Ryan Fitzpatrick, who bested the Patriots in Foxborough in 2019 to force the defending champs to fall to wild-card status in the playoffs. The second happened just this past year when the miserable New York Jets stunned the playoff-bound Rams in Week 15 to earn their first win of the season.

We see a similar result in the projected ‘total’ points scored in a game. The correlation to the actual points scored is .32; the total very slightly overestimates the actual score on average.

Note that the highest-scoring game in this dataset isn’t the Rams-Chiefs 2018 bonanza; this game comes in second to a 2004 58-48 Cincinnati Bengals win over the Cleveland Browns. The most disappointing game of the past 20 years from a scoring perspective came in 2000 when the Panthers beat the ‘Greatest Show on Turf’ Rams 16-to-3 in a game with 58 projected points.

Unlock Premium Tools + Bonus Episode

Join the #FootClan

It’s clear that it is very possible to predict, with reasonable accuracy, game scripts in the NFL. The next, crucial question remains: what does this tell us for fantasy?

The Fantasy Script

Let’s start with a visual on how NFL play-calling changes based on live game script. The y-axis here is the passing probability on any one play in games since 1999, while the x-axis is the number of points that a team is leading (or trailing) by. Each box represents one NFL quarter and the dots are sized by how often these scenarios occur.

  • NFL teams are often conservative in the first quarter, posting below 50% pass probabilities no matter the deficit. Note the huge bubble at a ‘lead’ of zero: this is because games start scoreless!
  • Each quarter, the ‘game script effect’ gets increasingly potent: teams in the lead run more and teams trailing pass more. This gets the most extreme, as expected, in the 4th.

We know, then, that game scripts are fairly predictable and that NFL teams let the game situation dictate their play-calling. The last step is to see if this translates to fantasy. Let’s chart the average (Half PPR in games since 2008) points scored by positions for both favorites and underdogs, with the x-axis as the point spread (i.e., 5 means that the favorite is projected to win by 5 points).

We can start with quarterbacks:

And onto running backs:

Unlock Premium Tools + Bonus Episode

Join the #FootClan

Finally, wide receivers:   

In all of these charts, we see for games that are projected to be close (spread of 2.5 or less) fantasy players score relatively the same if they are on the favorite or the underdog. However, as the spread widens – the favorite becomes a heavier favorite – we see that players across all positions score more when they are on the side of the favorites!

The implication here is what matters most is pointspicking players on the side of the favorite is more important than trying to play the game-script narrative such as picking a WR to rack up yardage on the side of the underdog in garbage time.

This doesn’t seem to gel with what we’ve seen so far, though: NFL teams do pass more when they are trailing, so why don’t we see QBs and WRs perform better when on the side of the underdog? What’s going on?

All About the Pie

If you listen to The Fantasy Footballers Podcast, you know that ‘pie size’, or the total production generated by NFL teams, is a crucial factor in fantasy projections. For example, it’s generally better to target players that are second or third on the depth chart in a highly productive offense than the top target in an anemic offense (think Cole Beasely vs. Jakobi Meyers last year).

In a nutshell, this is what is happening between favorites and underdogs. For example, while wide receivers on underdogs of 7 points or worse own a slightly larger share of the pie (66.5% of their team’s total yardage flows through wideouts vs. 66% for favorites), the size of the pie is small enough to offset these gains. Wide receivers on the side of the favorite see an average of 42 yards more than wide receivers on underdogs. Running backs on underdogs do have a smaller share of the pie – which agrees with the traditional game-script narrative – but it’s a tiny discrepancy: 33.5% for underdogs vs. 34% for favorites. Again, the average size of the pie – 27 yards larger for RBs on favorites – is the deciding factor here.

We can see this graphically by looking at the average cumulative yards by minute in an NFL game broken out by the favorite (blue) and underdog (red) in games with favorites of 7 points or more since 2008. Even though the passing clip accelerates for underdogs towards the end of the game, and they make back some ground, it’s not enough. The blue line favorite makes steady gains from the start of the game on, enough to more than offset the ‘garbage time’ points racked up by underdogs at the end of the game. Note that the favorites also do a better job going into halftime: they have a steeper incline in the final few minutes of the first half, likely due to productive possessions in the waning minutes.

Unlock Premium Tools + Bonus Episode

Join the #FootClan

The story is similar in rushing attacks (the blip downward at the end is because of kneeling; yes, even for the underdog, they win sometimes too!). The incline becomes steeper for favorites at the end of the game when they start to grind the clock out, but it doesn’t really make much of a difference: they already had a significant edge in rushing yardage.

Myth…

Mostly busted! I was surprised by this, and the conclusion is a bit nuanced. In this paradigm, it’s more important to target points than game script. Put another way, you want to have fantasy options on the teams that are projected to score more NFL points (the favorites) than trying to target a player because the flow of the game might fall their way.

In general, the main point of conventional wisdom challenged here ‘targeting QBs and WRs on underdogs as opposed to their counterparts on favorites’. While it’s true that favorites will generally choose to run more at the end of games (and underdogs will usually choose to pass more), by then the damage has already been done. QBs and WRs on the team projected to win have already generated their fantasy production, usually more than the underdogs will make up in the final minutes of the game. The ‘points’ factor dominates the ‘game script’ factor in most cases.

Similarly, it remains prudent to, all else being equal, prefer running backs on favorites vs. underdogs, but this isn’t necessarily because of game script. We saw that, while RBs on favorites have a slightly higher share of the pie, they really generate their edge before the fourth quarter clock-grinding drives.

_____

Did I miss something? Interested in putting another myth to the test? Message me on Twitter.

Comments

andy says:

excellent and interesting new pov

Leave a Reply

Your email address will not be published. Required fields are marked *