Week 1 Rankings & Game Flowbotics Preview
We’ve scaled things back dramatically here at TwoQBs, but I will continue to post weekly rankings along with my Game Flowbotics spreadsheet during the season. Scroll down to the bottom of the page if you want to go straight to my Week 1 rankings for fantasy football.
Once Football Outsiders releases their DVOA update after Week 1, the Game Flowbotics spreadsheet you’ve come to know will be back in action for Week 2 and every week to follow. If you’re unfamiliar with Game Flowbotics, what follows is the primer I wrote up on the spreadsheet from last season (copied from this article). The primer uses an old Titans-Steelers matchup as an example, so the player names won’t look right for the context of 2019, but the structure of the spreadsheet is still the same. Please follow @GameFlowbotics on Twitter to keep up with when the spreadsheet is published each week, and feel free to ask any questions.
Game Flowbotics Abstract
Game Flowbotics is something I started a few years back when I was writing for The Fake Football. The spreadsheet has evolved over time, but its purpose has always been to give a holistic overview of every NFL match-up on the upcoming slate, with the goal of helping us forecast game flow. If we can start to figure out how teams will try to attack each other, that should aid us in predicting usage for individual players.
Not only do we need to understand the strengths and weaknesses of a team, but we also need to understand how successful those angles of attack and defense will be against a specific opponent with its own set of strengths and weaknesses. To illustrate the games of Rock-Paper-Scissors-Lizard-Spock that NFL teams play, the spreadsheet amasses all sorts of statistics—from betting lines to Football Outsiders’ DVOA—and color-codes them so you can see how teams stack up against their opponents.
(Sidebar: Football Outsiders doesn’t have Week 1 data for real DVOA numbers, so that’s why the Game Flowbotics spreadsheet debuts in Week 2.)
Instead of continuing to ramble on about what the spreadsheet is, I’d rather just link to an example from last season so you can see it for yourself:
Is it starting to make sense? Yeah, probably not. Let’s break it down into detail…
Game Flowbotics Primer
The header at the top of the page indicates when the spreadsheet was last updated and provides relevant links for the matchup data shown below:
Now let’s go through an example matchup of Tennessee at Pittsburgh as it appears on the spreadsheet, from top to bottom, broken up into blocks.
Block 1: Game & Team Info with Total DVOA
Let’s skip the easy stuff in the two leftmost sections and skip ahead to the far left where Total DVOA and Total Weighted DVOA are shown. I urge you to check out Football Outsiders for an in-depth explanation of DVOA, but in short, it’s a measurement of team efficiency that adjusts for situation and opponent. Weighted DVOA assigns more weight to recent weeks, whereas normal DVOA weighs the entire season evenly. The “#” columns are the raw DVOA values, with 0.00% being average. The “Rk.” columns show where teams rank among all 32 NFL franchises for that particular metric. In this case, Tennessee was below average in DVOA, while Pittsburgh had performed well and ranked fourth in the league.
With that out of the way, let’s scan to the left and go deeper on the point spread, the over/under (O/U), and how they relate to the implied team total. (All spreads and over/unders are pulled from the Westgate listing on the ESPN page linked in the header.)
The spread is the perceived gap in points between the teams in a matchup. In this case, the Westgate has designated the Steelers as seven points “better” than the Titans for Thursday Night Football. The true value of home field is debatable, but to give a point of reference, evenly matched teams traditionally generate a spread with the home team favored by three points. Many other factors can influence the spread, and spreads can change in reaction to news and betting trends during the week.
Ultimately, sportsbooks like the Westgate are trying to get equal betting on both sides of the spread because they typically charge bettors a small fee for most wagers (called the vig, similar to the rake in DFS). With equal action on both sides, the oddsmakers break even on bet payouts while they rack up profit on the fees they charged to all bettors. As you might imagine, this means that spreads are not always an accurate projection of expected outcome. Sometimes, spreads are more indicative of the gambling public’s perception of teams.
Because some franchises are more popular than others (“public teams”), the spreads will often punish bettors more for picking those teams. The Steelers are a classic public team because they’ve been largely successful in the NFL for a long period of time (the Cowboys are another good example). The oddsmakers can’t go too far overboard with skewing the spreads for public teams, though, because if they do, sharper gamblers can bet heavily on the other side of the line and put oddsmakers at a disadvantage. Setting a spread is a balancing act between projecting the outcome of a game and projecting how people will bet it.
The Over/Under (O/U)
The same principles apply to the over/under (or “total”), which represents the combined number of points that could be scored by both teams. Instead of betting on a particular team with plus-or-minus some number of points, bettors simply choose if the combined total of points scored will go over or under a certain number. As with spreads, sportsbooks usually want equal action on both sides of an over/under, so some amount of projection goes into setting the number. As fantasy players, we need to be careful not to read too much into totals, because the they doesn’t discriminate between points from offense, defense, and/or special teams.
The Implied Total
Over/unders don’t discriminate between teams, either, but that doesn’t stop us from using the spread to project implied team totals. Essentially, an implied total assumes both the spread and over/under are accurate projections for a game. In this case of Tennessee at Pittsburgh, if 44 total points are scored and the Steelers win by seven, simple math says their final score will be 25.5 to Tennessee’s 18.5 (25.5 + 18.5 = 44, the total; and 25.5 – 18.5 = 7, the spread). In reality, teams can’t score fractional points, but implied team totals give a nice frame of reference when comparing different matchups. In theory, the more points a team is implied to score, the more fantasy production we can project for that team’s players.
Block 2: Overall Offense vs. Overall Defense
The second block shows overall matchups of offense versus defense in a contest. The “scoring” section shows teams’ points scored, points allowed, and points per game differentials. I include actual scoring numbers because they provide a nice sanity check against implied totals and DVOA matchups. In this case, the PPG values don’t line up especially well with the implied totals from the first block. Pittsburgh projected to score over 25 points, despite the fact that they averaged less than 21 per game while Tennessee allowed less than 24 per game at the time of their matchup. Perhaps this indicated some public team bias in favor of the Steelers, as discussed above.
The DVOA section on the right side of this block is similar to the one in the first block, except the ratings apply specifically to team offenses and team defenses. Again, the “#” columns contain raw DVOA and Weighted DVOA values, while the “Rk.” columns indicate how those values rank among all NFL teams. This will be a recurring theme going forward, but the matchups are color-coded. In yellow, we see the Titans’ offense matched up diagonally with the Steelers’ defense. In purple, we see the Steelers’ offense matched up diagonally with the TItan’s defense. Note that for defenses, negative DVOA values are a good thing. On offense, negative DVOA values are bad.
Altogether, this block plus the first block should set up a basic expectation for the end result of a game. In this particular matchup, the Steelers are clear favorites. As we progress through the remaining blocks, we can try to estimate how each team will attack the matchup to either reinforce or disrupt the expected result of a convincing Pittsburgh win.
Block 3: Passing Offense vs. Passing Defense
The third block focuses only on passing. Again, the matchups are color-coded. Orange plays against orange and blue plays against blue. In addition to DVOA stats and rankings, this block includes Adjusted Sack Rate data to show how well offensive lines block pass rushers and how well defensive lines get after passers.
Against Tennessee, Pittsburgh had advantages in the trenches on both sides of the ball, but especially in pass protection. Ben Roethlisberger should have had plenty of time in the pocket to attack Tennessee’s 24th-ranked pass defense thanks to his top-ranked offensive line. Considering how much the Steelers liked to throw deep last year, maybe it wasn’t so crazy to think they could hit their implied total of 25+ points in this matchup.
Block 4: Rushing Offense vs. Rushing Defense
The fourth block is very similar to the third, except this one applies to rushing, not passing. Red plays against red, green plays against green. In the running game, offensive and defensive lines are measured by Football Outsiders’ Adjusted Line Yards metric. As with all sections, the “Impacted Players” listings are only updated periodically, with the most recent changes date-stamped in the header at the top of the spreadsheet.
On offense, Tennessee rated better than Pittsburgh at running the ball, with the Titans eight spots ahead of the Steelers in DVOA and four spots ahead in Adjusted Line Yards. On defense, however, the Steelers were the better team against the run, especially up front. The Titan’s defensive line ranked 17th in Adjusted Line Yards, and that small failing should have spelled doom for them against Le’Veon Bell and a Pittsburgh O-line ranked 11th in Adjusted Line Yards.
Block 5: Receiver Types vs. Passing Defense
In the fifth block (only partially shown above), we finally get to receiver-specific matchups. Football Outsiders gives us DVOA values and rankings for defense against each possible type of pass-catcher: number one wide receivers, number two wide receivers, other wide receivers (WR3+), tight ends, and running backs as receivers. They also give us data on passes per game and yardage per game allowed to each of those receiver types. I’ve added columns to show the differentials in these volume stats between team defenses and NFL averages. There’s no color-coding in this block. The matchups read straight across.
In this game, we can see how Tennessee struggled against No. 2 wideouts. They ranked 28th in DVOA against those types of receivers, allowing 2.0 passes per game and 13.4 yards per game above league averages. In other words, JuJu Smith-Schuster had a good shot to perform well in this Thursday night contest.
On the flip side, the matchup for Rishard Matthews was a good example of DVOA not “agreeing” wholeheartedly with a defense’s volume stats against. By DVOA, Pittsburgh ranked middle of the road against No. 2 wideouts (21st among all teams), but those wide receivers only averaged 32.5 yards per game against the Steelers, compared to the NFL average of 45.2 yards per game allowed to the position.
At this point, you may be wondering why Rishard Matthews was listed as the Titan’s number two guy, and not their number one guy (as he played ahead of Corey Davis for the most part in 2017). On their defensive DVOA page, Football Outsiders admits to the arbitrary nature of their statistical designations with the following disclaimer: “Note that our decision of which receiver is ‘number one’ and which receiver is ‘number two’ is somewhat subjective.” Don’t blame them for what you see on the spreadsheet, though.
Football Outsiders is only responsible for how the different receiving stats are allocated to different receiver types. I personally dictate how the listed players slot into different roles, making educated guesses at FO’s intentions based on snaps played, target share, and general usage philosophy for each player. If you disagree with my choices, use your imagination and pretend the names are switched around to suit your own sensibilities.