Game script: A concept easy to understand in the abstract but difficult to distill into quantifiable data.
Those of us who play DFS understand game script through experience. A team gets down two scores early and begins to take more risks in the passing game. At the same time, the team with the lead begins to run the ball conservatively to burn time and keep the opposing offense off the field. Next thing you know, the winning team’s running back racks up 123 yards and a touchdown on 25 carries. Meanwhile, the losing team’s running back only manages 42 yards on 14 carries and 35 receiving yards on three catches.
Game script is a game-to-game phenomenon, which makes it vitally important when drafting players in DFS. But I believe we should pay a similar level of consideration to game script in redraft and best ball formats – especially when evaluating running backs.
In fact, this kind of game script discussion is already happening in the Twitter-verse with regard to Saquon Barkley. Public debate rages over who the overall RB1 should be in redraft: Barkley, Christian McCaffrey, Ezekiel Elliott or Alvin Kamara. Many analysts and casuals have weighed in with opposing views, but there consistently arises one knock on Barkley: poor potential game script.
The argument goes a little something like this:
@PanthersFanBoy81: “1.1 CMC, baby! Giants gonna suck this season, won’t be able to sustain drives. Saquon won’t even hit 200 rush attempts; big candidate for TD and all-purpose regression.”
@NYGfuheva: “OK, first of all, Saquon Barkley is the fantasy messiah: the second-coming of our one true fantasy Lord, LaDainian Tomlinson. Second, even if the Giants suck, Barkley will get 250 targets with New York always playing from behind. No more Odell Beckham, Sterling Shepard is already injured and Golden Tate is currently suspended. Crawl back into a hole, you Panthers homer.”
So, team allegiances aside, who’s right in this argument? Does the Giants’ poor offense cap Barkley’s rushing ceiling, or does it amplify his receiving ceiling? Moreover, if either of these theories is the “right” one, does that actually change his season-long fantasy projection in a substantial — and predictable — way?
This last question is the one I want to address in this series. In this article, I’ll be examining game script from a holistic perspective, and I’ll use historical Vegas win totals to evaluate how predictable and how sizable game script’s effect truly is. In subsequent series installments, I’ll elaborate on my findings and focus on which running backs may draw beneficial or detrimental game scripts in 2019.
League-Wide Pass-Rate Splits
Our first task is to confirm that NFL teams do in fact alter play-calling tendencies when playing with a lead or playing from behind. Furthermore, we need to quantify the disparity between leading vs. trailing pass-rates to establish a baseline for our discussion.
So, I gathered league-wide passing and rushing splits in each game condition from 2002 to 2018, the results for which are reported below:
The above line graph clearly demonstrates the profound effect game script has on play-calling tendencies. On average, teams’ pass-rate is approximately 17% higher when trailing compared to their pass-rate while leading. This finding was expected, but the magnitude of this difference is larger than I anticipated.
Another quick note from the graph above: We can clearly see the league’s increase in pass-rate across all game conditions since 2002, which I analyzed in detail in my passing revolution series. However, this pass-rate increase has mostly stabilized since 2013. As a result, all forthcoming analysis will focus solely on the 2013 to 2018 seasons.
Do Vegas Win Totals Reliably Predict Season-Long Game Script?
Okay; so, we confirmed that teams’ play-calling tendencies change depending on their game condition. But, that alone doesn’t help us project the usage distribution of specific players for an entire season. If we intend to capitalize on these play-calling splits, we need a method of predicting a team’s season-long game script.
In this case, the best tool for the job is Vegas preseason Win Totals. In theory, those are the sharpest pre-season approximation of a team’s relative strength entering the season.
So, I compiled Vegas Win Totals for all 192 teams in our sample. Then, I correlated those win totals with each team’s leading vs. trailing rushing and passing totals for the given season. Each of those correlational coefficients approached or exceeded r=0.40.
This means that Vegas Win Totals are a viable method of predicting a team’s projected pass or rush attempts when leading or trailing for the upcoming season. Win Totals obviously can’t account for future injuries, coaching changes or bad luck, but they’re nonetheless the best option we have.
For some context, a correlational coefficient of r=0.40 indicates that the variable of interest (Vegas Win Totals) explains about 16% of the variance in the outcome (team passing and rushing totals). A coefficient of r=0.50 translates to about 25% variance explained. A range of 16-25% may not seem like a lot, but fantasy football strategy is about capitalizing on many small edges, like death by a thousand paper cuts.
Analyzing Running Back Summary Statistics
Next, I compiled statistics for all running backs in our sample 1 into four quartiles based on preseason win total for each of their teams. I chose quartiles for the analysis, because the resultant sample sizes for each quartile were fairly even.
Quartile 1 is composed of players from teams with a Vegas Win Total of 9.5 or higher. Quartile 2 is 8.5 to 9.0; Quartile 3 is 7.5 to 8.0; and Quartile 4 is 7.0 and lower.
Let’s examine summary statistics for running backs from each of these quartiles to evaluate whether or not pre-season win total has a substantial effect on regular season PPR production.
Our sample population of running backs has 273 members, which is fairly substantial. However, subdividing that sample into quartiles – and then analyzing players based on other factors – cuts into that sample size significantly. As a result, we may expect our results to carry an elevated degree of variance due to sample size issues. Nonetheless, the trends reported herein paint a pretty interesting picture that’s worth our attention.
Average End-of-Season PPR Position Rank
I’ve grouped running backs in each quartile according to their ADP position rank, and I’m reporting average end-of-season PPR position rank for each ADP range. This chart, and each subsequent chart, is heat-mapped based on ADP category.
The most striking takeaway from this table is Quartile 3’s phenomenal performance compared to other quartiles. Running backs from this Vegas Win Total range (7.5 to 8.0) boast the best end-of-season PPR position rank for ADP RB1s, RB2s and RB3s. If it was just one ADP range, maybe we could write it off to variance. But three of them? That’s indicative of a systematic trend.
Most impressively, ADP RB3s from this range finish as mid-range RB2s on average. That is staggering.
But, why? What could cause Quartile 3 running backs to excel like this? Perhaps running backs on these teams benefit from particularly positive game script, because their teams are below-average but competitive. Think about it: These aren’t the bottom-dwellers of the league, but they’re not world-beaters either. They’re teams that are on the outside-looking-in at playoff contention, which means they’re likely locked in competitive game scripts on a regular basis throughout the season. That may lead to shootouts, which could in turn elevate the ceiling for all running backs on such teams.
Notable Quartile 3 running backs currently being drafted among the top-36 at the position include Christian McCaffrey, Le’Veon Bell, Leonard Fournette, Derrick Henry and Tevin Coleman.
ADP Position Rank “Over-Rate”
Alright, setting up this next summary statistic is going to be a bit clunky; just bear with me here. In the gambling sphere, over-unders are a common bet type. You may wager that a team will go over or under its Vegas team total, that a particular matchup will go over or under its Vegas points total or even that a player will go over or under his rushing yardage total that week.
I’m applying the same logic to ADP position rank. Below, I’ve reported the over-under record for running backs in each quartile based on running backs’ ADP position rank. If by season’s end a player outperforms his ADP position rank by at least four ranks, he went “over” his ADP position rank (aka: he exceeded expectation). Conversely, if a player failed expectation by at least four ranks, he went “under” his ADP position rank. If a player’s end-of-season PPR position rank was within plus or minus three ranks of his ADP position rank, it’s a “push.”
Yet again, we see Quartile 3 running backs excelling. 62.5% of Quartile 3 running backs exceed their ADP by at least four position ranks. And that includes all players – even those drafted outside the top-40.
Interestingly, it’s almost the opposite story for Quartile 2 running backs. Running backs from these teams (Vegas Win Total of 8.5 to 9.0) report an abysmal over-under record with just a 37.3% over-rate. This finding isn’t substantial enough to warrant fading all running backs from this tier blindly, but it does raise a significant red flag.
Notable Quartile 2 running backs include Ezekiel Elliott, James Conner, Dalvin Cook, Nick Chubb, Devonta Freeman, Aaron Jones and Mark Ingram.
Weekly PPR Finishes
For our final summary stat, let’s take a look at average PPR finishes for running backs in each quartile. I compiled every running back’s weekly PPR position rank for each season in our sample. Then, I counted how many top-12, top-24 and “bust” 2 weeks each running back managed in the given season.
Finally, I summed each of these categories and produced average top-12, top-24 and “bust” weeks expected from fantasy RB1s, RB2s, RB3s and RB4s in each quartile. I’ll report each ADP position class one at a time, beginning with ADP RB1s:
First, let’s get the obvious take out of the way: Quartile 3 is balling once again. Running backs from this quartile are tied for the top average top-12 weeks, boast by far the most top-24 weeks and also report the lowest overall bust rate.
Quartile 1 running backs (Vegas Win Total of 9.5 or higher) aren’t too shabby either, but they also report the worst bust-rate of any quartile. Perhaps this is due to variance, maybe it’s due to blowout victories, or maybe top teams simply don’t employ running backs as voluminously. No matter the cause, that bust rate is a concern.
This table looks pretty different than our previous one. ADP RB2s from Quartiles 2 and 4 excel here, boasting higher average top-12 and top-24 weeks than their other cohorts. Additionally, “bust” weeks are pretty similar for all four quartiles, so there’s similar downside risk associated with all ADP RB2s in general. This should serve as encouragement to draft for upside among running backs in this ADP tier.
Woah. This table is the most dramatic of them all, and once again it’s Quartile 3 that’s making some serious noise. I don’t think I need to break this result down too much, because the production level for Quartile 3 ADP RB3s jumps off the page. There’s some serious value here, and there’s only one Quartile 3 running back currently being drafted from RB25-RB36: Tevin Coleman (RB30).
What to Expect in Part 2
Admittedly, that was a lot of data and analysis for a single article. So, where do we go from here? How is there still more to uncover in future installments?
Well, the analysis in this piece is very general. It deals with global trends, correlations and averages. These are summary statistics. They may help us identify potential sources of value, but they don’t do a great job at pinpointing which players to target in each category.
So, in future articles, I’ll be analyzing running backs from each quartile based on the way they’re used in their respective offenses. I’ll break down how team rush share, target share and opportunity share affects running backs in each quartile – and in each game condition (leading vs. trailing).
In Part 2, we’ll tackle 2019 running backs from Quartile 1, which includes the following teams: Saints, Chargers, Rams, Chiefs, Colts, Patriots and Eagles.