Interpreting the Giants’ 2-0 start and looking ahead using data

After two weeks of play, there are already only six undefeated teams in the NFL. Three are teams who have recently been the league’s juggernauts: The Buffalo Bills, Kansas City Chiefs, and Tampa Bay Buccaneers. Two others are the Philadelphia Eagles and Miami Dolphins, both of whom have beefed up their offenses and/or defenses with explosive early results.

Then there are your New York Football Giants. Few if any expected the Giants to be 2-0 to start the season. But here we are, Giants fans, with an unexpectedly significant game against the division rival Dallas Cowboys coming up on Monday night. How did the Giants get here? Is it sustainable? Kevin Cole of Pro Football Focus doesn’t think so, although he acknowledges that with the Dak-less Cowboys and Chicago Bears on deck, 4-0 is not out of the question:

Since Cole lives in the PFF universe, his opinions are based on numbers that try to capture the Giants’ performance in ways that are meaningful to winning and losing. Some fans are not keen on thinking about their team in these ways, subscribing to the phrase made famous by Mark Twain that there are “lies, damn lies, and statistics.” But let’s look at some of the statistics, anyway - not to bury them, not to praise them. Instead, let’s just assume that each statistic tells us something that is useful about the team and its players and try to figure out what that is.

Traditional NFL statistics

According to Pro Football Reference, the Giants are 21st in the NFL in total offense with 659 total yards (all teams have played 2 games by this point). That is not bad given the Giants’ recent futility on offense, but compare it to the Eagles (first at 941), Dolphins (854), Bills (827), and Chiefs (807), four of the five top offensive teams. The Buccaneers are 24th at 605 yards, so they are doing things a different way.

The NFL is a passing league today, though. Sure enough, Miami is first, Buffalo fourth, Kansas City sixth, and Philadelphia seventh in passing yards. Tampa Bay, again a bit of an outlier, is 24th. The Giants? Thirty-firs. Air Daboll, this team is not thus far. But the Giants are fifth in the NFL in rushing yards (341). Only Philadelphia, among the other 2-0 teams, is higher (379). So rushing has partly compensated for the lack of passing offense to date. That will come as no surprise to fans who watched Saquon Barkley’s performance against the Tennessee Titans.

The only statistic that stands out for the Giants on offense? They are the only team in the NFL with two game-winning fourth quarter drives.

On defense, things look better but not spectacular. The Giants are 12th in total yards given up at 634. Of the other undefeated teams, Buffalo is second at 430 yards and Tampa Bay fifth with 552. They look a bit better in several advanced statistics: eighth best in yards after catch surrendered (but trailing Buffalo, Philadelphia, and Tampa Bay), third in blitz percentage (42.3), and sixth in hurry percentage (15.5). But they are only tenth in overall pressure percentage because to date they have had relatively few quarterback hits and sacks.

PFF grades

PFF grades are a lightning rod for football fans because they sometimes do not seem to jibe with their impressions watching the game nor with the results. It’s important to understand what PFF grades are and what they are not. PFF analysts subjectively grade every player (on a scale from -2 to +2) for every play in which they are on the field to assess how they performed their own task relative to expectations, regardless of whether they were involved in the end result of the play. (Manning-to-Manningham in the Super Bowl is an example they give of a +2 play, presumably for both players. The fact that it basically won that Super Bowl does not enter the equation.)

If a receiver runs a great route but the quarterback is sacked, the receiver still gets a high grade. If every player does his job well but the play call is terrible, the players still get high grades. (For example, in principle the “surrender formation” plays the Giants ran in their final 2021 game could in principle have resulted in zero or even positive grades for the players involved if they did the job they were assigned on the plays.) PFF is not grading coaches, play calls, or play designs.

Composite grades for a player for a game are then scaled to fall between 0 and 100. There has been some inconsistency over PFF’s history about how to view these scores, but in recent years the most common interpretation is that grades in the 60s indicate an average player, 50s or lower a “replacement level” player, 70s an above average player, and 80s or higher an elite player. With that as background, here are the PFF grades for the six undefeated teams:

The Giants are by far the worst of the six teams by the PFF metrics. There is not a single category in which they excel within that group, and in several categories, e.g., run blocking (47.3) and run defense (50.9), they are well down into replacement level territory. Every other 2-0 team has above average passing game performance by its players. Notably, both pass blocking and run blocking are average or below average for five of the six teams. Only half the teams have had an above average pass rush, but four of the six have had better than average pass coverage. None of the teams has had excellent special teams play.

This is a small sample. Over time, excellence has to be sustained for a team’s season to succeed. Here are the regular season grades for the six best (by record) 2021 teams:

(Note that the list does not include one of the Super Bowl teams, although the Cincinnati Bengals were not far behind.) Not surprisingly, most of the teams were above average in most categories for the season. (The comparison to 2022 is not a clean one, however. PFF values consistency in good peformance over the season more highly than alternating excellent and mediocre play and adjusts its season grades accordingly.) The one thing that apparently doesn’t matter? Run defense, which was subpar for all but one team. On the other hand, that one team was the Super Bowl champion.

So we can interpret the Giants’ early success in a few different ways:

• It’s two games, don’t get excited. A statistically insignificant sample.
• The teams the Giants have played are not good. This may well be true for Carolina. Tennessee was thought to be a high quality opponent but they were dismantled by Buffalo, so it remains to be seen whether they will survive the loss of A.J. Brown to the Eagles.
• The Giants are a collection of mostly mediocre players, but the coaching is compensating for that with good offensive and defensive schemes and good play calls at the right time to hide their deficiencies and put them into positions to succeed. This is plausible but difficult to quantify. Surely Wink Martindale’s aggression on defense and heavy fronts vs. dime packages has made a difference. Surely Mike Kafka’s offensive sequences that run different plays off the same early motion make a difference. Surely Brian Daboll’s decision to go for two made a difference.
• They’ve just been lucky. A missed 47-yard field goal. A made 56-yard field goal. Reverse those and maybe they’re 0-2.

All of these can - and may - be true.

Expected Points Added

As opposed to PFF grades, which are subjective assessments of individual player performance, Expected Points Added (EPA) per play is the signature objective statistic of a team’s performance by the analytics community. Analytics is a divisive topic within NFL circles - nerds taking over a man’s game, to some traditionalists. We all remember (well, at least this nerd does) former Giants’ head coach Joe Judge’s shot at the analytics people: “If Excel was gonna win football games, Bill Gates would be killing it right now.”

Interestingly, EPA does not trace its roots to a nerd. Its predecessor, Expected Points, was invented by NFL quarterback Virgil Carter, who started for the Cincinnati Bengals for a couple of years and had a seven-year career, going 16-14. The idea was published in 1971 by the Institute for Operations Research and the Management Sciences. (Admittedly, Carter had a co-author, Robert E. Machol, a professor at Northwestern University, so perhaps there was at least one nerd involved.)

Carter and Machol used a database of NFL statistics on outcomes of plays, the yard line at which they occurred, and eventual scores to quantify the intuitive idea that the closer you are to the opponent’s goal line, the better your odds of scoring either a field goal or touchdown are. For example, from the 50-yard line with first-and-10, on average teams score a bit less than three points on a drive - the “Expected Points (EP)” for that drive. These days, in addition to much larger databases, EP incorporates information on down, game situation, etc.

To evaluate a player, or a team, using this concept, you can calculate how much higher EP becomes if a team moves the ball 5, 10, or 20 yards, etc. closer to the end zone on a particular play. That increase in EP is called the Expected Points Added for that play. You can score a touchdown from the 50 with five short runs, five short passes, and five incompletions. Or you can score with one 50-yard bomb. The latter has a 15 times higher EPA/play. A negative EPA/play is one on which the offense is less likely to score after the play is over (a loss of yards, penalty, or unproductive down). So for a defense, a high negative EPA/play is desired.

The analytics community considers the one explosive play to be more a harbinger of winning than the 15-play drive, because you have 15 chances for something to fail and terminate the drive in the latter case. The philosophy is not much different from that which permeates baseball these days, in which home runs are valued more than trying to “build” runs with singles, bunts, stolen bases, etc. (Aside: Congratulations Aaron Judge!) EPA/play values scoring over ball possession. Not every NFL coach or fan agrees with that. And indeed in some cases, especially at the end of a close game, there’s a case to be made. Buffalo made the “mistake” of leaving 13 seconds on the clock for Kansas City after they took the lead in their playoff game last season and paid a price. But most football people believe these days that overall, plays that gain big chunks of yardage are better than three-yards-and-a-cloud-of-dust.

Here is how team EPA/play on offense and defense stands after Week 2:

Five of the six undefeated teams are outliers in a good way. Buffalo has a high positive EPA/play on offense and high negative EPA/play on defense. Kansas City, Philadelphia, and Miami have a high offensive EPA/play but only average defense. Tampa Bay has an excellent defensive EPA/play but a somewhat poor offense so far. The Giants are smack in the middle, with a neutral defense and slightly below average offense. That is a somewhat different picture than is painted by the traditional statistics of yards gained and given up, which suggest that the Giants have a bad offense and a very good defense. It does not, though, point to the Giants as being a likely 2-0 team.

Here are the quarterback EPA/play stats (credited to the QB even though offensive EPA is a metric shared by all 11 players on the field for a given play) through Week 2, plotted vs. PFF grade:

Four of the undefeated teams have gotten great quarterback play (as has Baltimore, which should be 2-0 but blew a big fourth quarter lead to Miami). Tampa Bay (hidden behind Tennessee) has been middling. The Giants have gotten slightly negative EPA/play performance from Daniel Jones - not a surprise, because Jones has thrown mostly short passes, whether because of conservative play calls, pass protection failures, his own failure to make downfield reads, or some combination of all of these. Note that the subjective PFF grades correlate pretty well with the objective EPA/play scores, but there is a decent amount of scatter. No metric is perfect.

One more EPA/play (expressed here as an “efficiency”) diagram, this one comparing contributions on offense from running vs. passing plays:

Five of the six undefeated teams have high EPA/play on pass plays. Only the Giants have a negative pass efficiency. The Giants’ offense, statistically anyway, is succeeding primarily in the running game, which was also evident in the traditional passing vs. rushing yards rankings.

Game decisions as seen by analytics

The most controversial application of analytics is its use to evaluate in-game decisions, specifically whether to go for it on fourth down and whether to go for two after a touchdown, depending on game situation. There is a publicly available fourth down decision calculator created by Ben Baldwin that uses a database of such decisions, resulting play outcomes, and game outcomes to calculate how these decisions affect the probability of winning. Here are two notable examples for Giants fans of decisions Brian Daboll has made. I input my best estimate of the game situation for each:

Game 1 vs. Titans, 2-point conversion at 1:06 in fourth quarter

According to the calculator, going for two was the right decision for Daboll, but not by much, adding only 0.7 percent win probability - which is about right considering how things played out. For the record, although the bot does not do well with end-of-game scenarios, it calculates that there was a 73 percent chance of Tennessee successfully kicking the 47-yard field goal.

Game 2 vs. Panthers, Gano 56-yard field goal at 3:34

With the Giants facing fourth-and-18 and a kicker as good as Graham Gano, there was little doubt about what the play should be. But just in case you were wondering whether flipping the field position with a booming punt by Jamie Gillan would have been better, the decision bot says no, the field goal attempt made winning 3.5 percent more likely, although even with the successful kick, the Giants’ win probability was only 66 percent at that point. The defense and then the offense each had to do their job one more time to secure the victory.

All these things went right for the Giants in Weeks 1 and 2: A two-point conversion that had only a 47 percent chance of success. A 47-yard field goal that had only a 27 percent chance of being missed. A 56-yard field goal that only had a 46 percent chance to be good and that only gave the Giants a two in three chance they’d win if it was. Why are the Giants 2-0? Luck had a lot to do with it.

Looking ahead to Monday Night Football

Could the Giants actually make it to 3-0? Their opponent, the Dallas Cowboys, has a strong pass rush led by LB Micah Parsons and an attacking though gambling secondary personified by CB Trevon Diggs. Given the Giants’ early struggles on offense, points may not come easy.

The Giants’ best hope for this game comes from the defense. Dallas has a potentially explosive offense with wide receiver CeeDee Lamb and perhaps the return of Michael Gallup, in addition to a strong two-pronged rushing attack with Tony Pollard and Ezekiel Elliott. Can the Giants slow them down? Here are two possible keys, both based on simple, purely objective analytics information. Each one revolves around the question of whether backup quarterback Cooper Rush can duplicate his heroic performance that led Dallas to victory over Cincinnati last Sunday.

Key 1: Show CRush a defense he’s not used to seeing

Two-high safety defensive looks before the snap have become the rage in the NFL. With one big exception:

For the most part, NFL teams are running the same defense in 2022 as they did in 2021, hence the diagonal orientation of most of the cluster of points. Not the Giants, though, who have gone from 60-65 percent two-high safety pre-snap looks under Patrick Graham to less than 30 percent two-high under Wink Martindale, increasingly a unicorn among NFL defensive coordinators. Only the 0-1-1 Indianapolis Colts are showing as much single-high. Two-high presents its own challenges to a QB, especially when a defense rotates into single-high at the snap to disrupt the QB’s initial read. But single-high defenses like the one Martindale is running are often accompanied by added pass rushers and blitzes such as the one that led to the game-saving sack of Baker Mayfield by Julian Love on Sunday.

Thus far Rush has seen Tampa Bay (for about 5 minutes), which runs two-high a bit more than half the time, and Cincinnati, which does it a bit less than half the time. Last year, Rush beat a Minnesota team that ran two-high a little more than half the time. The Giants will be the biggest taste of single-high safety looks that Rush has seen so far. Can they pressure Rush often enough into bad passes or sack him? Can Adoree’ Jackson and whoever plays CB2 hang with Lamb, the best wide receiver the Giants will have seen thus far, and Gallup if he plays?

Key 2: Keep the Cowboys from getting off to a fast start

Most teams run a set of 15 or so pre-planned, or “scripted,” plays at the start of a game to get the offense going. For some teams it makes a big difference in their offensive efficiency. At least it has through the first two games:

Dallas is one of those teams, having a 0.2-0.3 offensive EPA/play on scripted plays at the start of the game and a -0.3 EPA/play for the rest of the game. It makes sense that an inexperienced QB like Cooper Rush would execute play sequences he’s known about and practiced for days better than those that are introduced in reaction to events that happen during a game.

Against Cincinnati, Rush opened the game with a 12-play, 75-yard TD drive. The next time Dallas got the ball, Rush ran a 6-play, 75-yard drive for another TD, and the score was 14-3 Dallas by the end of the first quarter. Rush would only put 6 more points on the board the rest of the game. To his credit, he got 3 of those 6 in crunch time to secure the win. But the opening drives put the Bengals on their heels right away.

The only team that has prospered more in the scripted portion of the game relative to the remainder of the game is the Tennessee Titans. Against the Giants, Tennessee opened with a 5-play, 45-yard drive for a TD and followed it with a 10-play, 65-yard drive for a field goal, and the Giants found themselves in a 10-0 hole.

Dallas did a lot of their business in their opening drives against Cincinnati with RBs Elliott and Pollard. The Giants need a plan to neutralize them early to put pressure on Rush to win through the air. But do they do so in Martindale’s favored single-high safety scheme with lots of blitzing if both Lamb and Gallup play?

Leonard Williams will miss time, weakening the interior defensive line, and the Giants’ linebacking corps has problems in coverage of running backs, or inexperience, or both. Will Kayvon Thibodeaux and/or Azeez Ojulari make their season debuts, and if so will they be able to set the edge on running plays? Will Martindale bring Xavier McKinney and/or Julian Love down as pseudo-linebackers again as he did against Carolina and risk Lamb and Gallup getting behind the defense with only Dane Belton back to help out the cornerbacks?

Could newly signed Jaylon Smith, who immediately becomes the Giants’ best cover linebacker, possibly be activated by Sunday? It seems impossible to imagine him learning the defense that quickly. Fabian Moreau played for the first time last Sunday, 13 days after being signed to the practice squad, and made a key fourth quarter tackle on a long run by Christian McCaffrey. Smith did have a 79.4 PFF coverage grade and 77.7 tackling grade against Dallas last year for the Giants, and he knows the Cowboys from his time there. The odds are small, but you never know. We are learning to expect the unexpected from Brian Daboll and his staff when it comes to usage of personnel.