NBA Championship Equity Based on Best Players

The Lakers recent NBA Finals victory got me thinking about how seemingly unusual their roster was shaped. The teams that win the championship almost always have at least a couple of high-end, All-NBA type talents on their roster (the 2011 Mavericks and 2004 Pistons are exceptions, but certainly not the norm). Still, this Lakers roster seemed especially top heavy, with LeBron James and Anthony Davis surrounded by a just solid crop of players. According to the newest Box Plus-Minus (BPM) model available at basketball-reference (the write-up of which can be found here if you are like me and want to get into the weeds), the third best player on the Lakers was JaVale McGee who was out of the rotation by end of the Western Conference Finals and did not play a game in the final round. Of the players in the rotation against the Heat, the third best player during the regular season was Danny Green, who added about half a point above average per 100 possessions during the regular season. McGee added 1.5; the difference between Davis (who posted a BPM of 8) and McGee was the second largest this century between the second and third best player on the NBA champion after the gap between the second and third best players on the 2012 Heat, Wade and Bosh. The gap between the Lakers this year and the third largest difference in the sample (incidentally between Kobe Bryant and Rick Fox on the 2001 Lakers) was the same as the difference between third and eighth of the list (which was the gap on the 2009 Lakers; the Lakers have won a lot of championships). 

It seemed like there were a lot of worthy challengers for the NBA crown this year who boasted much deeper rosters than the Lakers, including the Clippers, Bucks, Celtics, Heat (who they vanquished in championship round), the Nuggets (who they beat in the semifinal round), and even the Rockets (who they beat in the second round). Yet the Lakers came out as the champion on the backs of their two superstars. We know the NBA is a star-driven league, much more so than any other sport excluding maybe the quarterback on a football team. So it begs the question: how much championship equity does a team going into the playoffs based on its best player and its best several players? To investigate, I took every team season since 2000 and pulled out the best player on each team and the three best players on each team based on BPM and a minimum of 500 minutes played in the regular season. I built two models: the first is a logistic regression where the target was whether or not a team won the championship and the variable was only the BPM of the best player. The second is another logistic regression where the target is whether or not the team won the championship, but the variables were the BPM of the best player, the second best player, and the third best player. 

First, I took a look at the predicted championship probabilities of the model using data from only the best player. I will note I tried incorporating other data about the best player into the model, such as usage and shooting efficiency, but it actually made the model less accurate (based on AIC which gauges in-sample predictive power). For context, here is the distribution of BPM figures for the best players on championship teams versus all other teams in the sample: 

Unsurprisingly, the teams that win championships have top players significantly better than the average team. The exceptions are Chauncey Billups on the 2004 Pistons, Dirk Nowitzki on the 2011 Mavericks, Kobe Bryant on the 2009 and 2010 Lakers, and Kawhi Leonard on the 2014 Spurs. Notable seasons on non-title winners were LeBron in 2009 and 2010 on the Cavaliers, Chris Paul on the 2009 Hornets, Steph Curry on the 2016 Warriors, Giannis Antetokounmpo on the 2020 Bucks, James Harden on the 2018 Rockets, Kevin Garnett on the 2004 Timberwolves, and Kevin Durant on the 2014 and 2015 Thunder. When you look at the output of the regression model, the best player has an outsized affect on a team's championship equity. 

Interestingly, the marginal gains associated with your best player getting a little bit better have an outsized effect on your probability of winning a title at the higher-ends of the player production spectrum. For example, if a team had a best player that was about one point per 100 possessions better than average and added a player in the off-season who was about 6.25 points per 100 possessions better than average, that would would add about five percent of championship win probability in the average season this century (based on the model). The same championship equity would be added if the best player on a team went from about 8.75 points per 100 better than average to about 10. So on the upper end of the player ability spectrum, adding about 1.25 points is the same as 5.25 points when considering just championship equity.

When incorporating the second and third best players into the model, the model becomes more accurate (by AIC). This is not surprising because I am incorporating more information about each team's roster. Theoretically, adding in all of the players would be the most accurate, but there would probably be significant amount of diminishing returns after the seventh or eighth best players since rotations shorten in the playoffs. But in any sort of modelling process, there is a balance that needs to be struck between accuracy and the amount of information required to make a good prediction. A more complicated model that requires 20 inputs but adds only five percent more accuracy compared to a model with four  inputs is not really a better model. Marginal gains in accuracy compared to monumental changes in inputs is not good process. Thus, I figured for now looking at just the three best players would suffice do to its simplicity and intuitiveness. 

With the model trained, I could generate championship probabilities based on the play of the three best players on the team, just like the model that only incorporated the best player. My initial reaction was to look at the landscape of each team's top two players and outliers in championship probability. 

The Warriors show up a few times here, as they put together won of the most dominant stretches of play in league history. Interestingly, we also see the 2018 and 2019 Rockets with James Harden and Chris Paul at the helm. In 2018 specifically, the model indicated that the Rockets actually had a better chance at the championship than the Warriors. This is based on regular season performance and the Warriors after the Kevin Durant signing were known to loaf through games so the results should be taken with a grain of salt. Another couple of teams that unfortunately ran into the Warriors juggernaut were the 2015 and 2016 Thunder, featuring the aforementioned Kevin Durant and Russell Westbrook. The 2001 Jazz were the last hurrah for Stockton and Malone as serious playoff contenders. I also isolated teams that were similar to this year's Lakers, where the third best player was much worse than the top two. I filtered by teams with an implied championship probability of at least 20 percent and a third best player who was at best worth 2.5 points per 100 above average. 

The "Heattles" show up here as do Durant's Thunder and the Stockton and Malone Jazz. Famously, Kevin Garnett carried a relatively barren roster (by contender standards) to the Western Conference Finals in 2004. The second and third best players on that team were Sam Cassell and Fred Hoiberg; this masterpiece from Garnett was one of the best seasons by a big-man in NBA history. Alternatively, here are some of the deepest contenders of the century (both the second and third best players were at least 3 points per 100 better than average): 

Here we see different versions of great Spurs teams, an organization known for its depth and ability to develop players for most of the Greg Poppovich tenure. The earlier Spurs teams were led by the trio of Tim Duncan, Tony Parker, and Manu Ginobili while some of the later teams featured Kawhi Leonard and LaMarcus Aldridge. The Chris Paul and Blake Griffin led Clippers also make an appearance. After recent renditions of the Rockets and Thunder, these Clippers were the best group to not win a title. 

When taking into account each team's championship probability every year since 2000, we can look at which teams exceeded or disappointed their expected championship output. First, all the teams that won at least one championship in the 21st century:

The Lakers lead the way in titles over expected. Some of this can be attributed to the Shaq and Kobe teams not taking the regular season seriously. The rest is attributable to the back-to-back championship teams in 2009 and 2010 not being especially strong title winners. Miami is second and for similar reason to the Shaq and Kobe teams they exceeded their expectations. Detroit and Toronto come out as surprising as "lucky" by this measure (a small ratio is expected to actual titles). For Toronto, they did not have much championship equity until they traded for Kawhi Leonard and in his lone season in Toronto Leonard often sat out regular season games to stay fresh. Detroit in 2004 was the worst champion by the model. Dallas had some great teams between 2000 and 2010 and oddly finally got a ring with Dirk when he was well past his MVP-level peak. 

As I mentioned above, the Thunder, Rockets, and Clippers were very unfortunate given the players they employed this century. Milwaukee's misfortune is concentrated in the past couple of seasons where Giannis posted some of the best seasons of the century. Minnesota's misfortune was also concentrated in a couple of seasons, 2004 and 2005. Also showing up here is Orlando (2008 through 2010), the Suns (mostly due to the Seven Seconds or Less teams) and Utah (the tail end of the Stockton and Malone connection). 

Finally, I looked at the strength of the top teams every season for the last 20 years. The model output does not consider other teams when computing championship probability. For example, the 2018 Rockets title chances are not affected by the fact that the 2018 Warriors existed. The model was not trained on individual seasons but the entire sample. Thus, the predicted probability is the probability of winning a championship given the players on hand in the average season in the 21st century. The issue for the 2018 Rockets is the average NBA season does not also include a team of the Warriors caliber. The top teams from each season have an outsized effect on the summation of title chances for all teams in a season. So when summing up all the title chances in each season, seasons over one can be considered top-heavy, while seasons under one have title contenders with worse players than the average contender. I call the sum of the probabilities the heavy weight index. The average heavy weight index in the 21st century is about one, which is not surprising. The landscape of the league, on the other hand, is noteworthy: 

Recent seasons, especially 2015 through 2018, were especially top heavy. Those seasons were dominated by the Warriors, Cavaliers, Thunder, Spurs, and Rockets. These teams had some of the most impressive top-end talent the league has ever seen and just happened to exist in the same universe so the Rockets and Thunder missed out on titles. All of these seasons had heavy weight indices of at least 1.5, so they were at least 50 percent more top-heavy than average. Seasons in the first ten years of the sample featured much less impressive talent, especially in the years where Kobe's Lakers took how titles. These seasons were almost 40 percent less top heavy than average. With the break-up of the Thunder, Warriors, Cavaliers, and now Rockets, the league is starting to balance itself out again after some extremely top-heavy seasons. I think, subjectively, having an index a bit over one (the best teams are slightly better than the best teams in an average season), but not so dominant that they bowl over the competition offers the best entertainment product. With many top players hitting the free agent market after the 2020-2021 season, it will be interesting to watch how the contenders sort themselves out and if any team adds talent to the extent of some of the great teams of the last 20 years.

Addendum to Investigation of Gary Sanchez's Struggles

In my last post I talked about Gary Sanchez and his 2020 struggles at the plate. I cited his excellent batted ball statistics persisting in 2020 and the main reasons for his demise being attributable to a bloated strikeout rate that is bound to regress and some bad luck on balls in play. After further reflection, my stance on his exceedingly low BABIP has not changed. It was one of the worst figures of the past five seasons and given how hard he hits the ball when it is put into play, I do not expect anything close to that figure to persist. Even though he hits the ball with so much authority, Sanchez may always post lower BABIP figures  due to his propensity for hitting fly balls and the proportion of his hits that are home runs (home runs are not included in BABIP), but nevertheless there will be some positive regression on this front. 

The strikeout rate requires more nuance. I noted his slight improvement in his approach; he improved his chase rate and swung at more pitches in the strike zone in 2020. His overall swing rate did not change much and he saw a slightly higher percentage of pitches in the strike zone. The issue was his contact rates, both inside and outside the zone. His zone contact rate declined a little bit from an already unimpressive figure, but the larger issue was Sanchez, relative to league average, could not put the bat on the ball when he chased. This meant that even though he was chasing less, his rate of contact when he did chase was such where swings on pitches outside of the zone had an outwardly bad effect on his results. I also included some analysis on the probability that this increase in strike out rate was purely the result of variance and found there was some non-zero possibility that this was the case. I threw the improved approach, substantial negative regression in contact rates, and variance into a blender and concluded that he was bound to be much better in the strikeout department in 2021. 

Thinking more about the strikeout issue, I thought I failed to offer context behind Sanchez's rising strikeout rate and look at other players who saw large changes in strikeout rate and how they fared in later seasons. It is easy to say that any massive increase in strikeout rate should be followed by a corresponding decrease towards the players "true talent". 

I pulled data on every set of three hitter seasons since 2015 where the hitter had at least 150 plate appearances in each season. I then took the calculated the changes in strikeout rate between year N and year N-1 and between year N-1 and N-2. 

Positive integers indicate increases in strikeout rate, which are generally bad for hitters (but not always, could be an indication of a hitter being more selective or selling out for some power). As you can see in the direction of the trend line, increases in strikeout rate are often followed by decreases the following season. When you isolate Sanchez, the results are concerning: 

This is the same set of points with Sanchez highlighted. You do not want to find yourself on the top right of this chart, which indicates multiple seasons where your strikeout rate increased. Sanchez's 2020 is especially ugly in this regard. I will not that his 2019 season looks bad in this visualization, but in 2019 Sanchez posted his best numbers on contact. That might indicate that he was selling out for some power coming off a down 2018 season. Using the data presented above, I built a simple linear model predicting a player's change in strikeout rate. The model had two inputs: the prior change in strikeout rate and the player's age (we know changes in strikeout rate are partly a function of age). After fitting the model to the data, I wanted to look at how Sanchez looked in 2019 and 2020 relative to expectation (i.e. the output of the model). In 2019, based on changes from 2017 to 2018, we should have expected Sanchez to trim 0.57 percentage points off of his strikeout rate while in actuality he added 3.47 percentage points. In 2020, we should have expected him to lose about 0.72 percentage points off of his strikeout rate. Instead, he added eight percentage points. His 8.72 percentage point change over expected was in the 98th percentile in the entire sample. His two year total change is 7th in baseball from 2018 to 2020. Furthermore, I built another model that gauged the probability of a player trimming his strikeout rate based again on age and prior year strikeout change. Sanchez again sticks out. 

Player seasons in the top right quadrant are disappointing relative to expectation. Sanchez both seasons was expected to trim his strikeout rate and he did the opposite. I will note that most of the seasons where players added to their strikeout rate much more than expected came from 2020, due to the small samples. Still, I think there is some reason to be concerned with Sanchez and his increasing strikeout rate. If he wants to return to his 2019 level, he is going to have to buck the strikeout trend. Sanchez will be 28 going into next season and has been a major league regular for about four years now when you account of the fact that he did not play full seasons in 2016 and 2020. He is probably not going to get better from a batted ball perspective at this point; exit velocity peaks in a players mid 20s. Improvements will have to be made in his contact rates and correspondingly his strikeout rate. Whether or not he has the ability to make these improvements will largely dictate whether or not the Yankees tender him a contract going into his second year of arbitration. 

Gary Sanchez 2020 Struggles

Gary Sanchez has been the target of scorn among Yankee fans for a couple of years now and the distaste has only grown during the abbreviated 2020 season. Sanchez is much maligned for his subpar defense, which is always on display given the catcher is involved on every pitch. What this fails to recognize, however, is the value Sanchez brings with the bat and how that stacks up to his peers at the catching position. Since 2016 (he became a regular on August 3rd of that season), Sanchez has been the fifth most valuable catcher in baseball, per FanGraphs WAR. On offense alone he has provided the second most value while ranking just 10th in plate appearances among catchers. On defense he has been more middle-of-the-pack, but 2019 was actually his only season where he was below average with the glove, at least according to UZR and FanGraphs' catcher framing model. This is all to say I think the criticism of his play has been largely unfounded and lacks context, in that despite what fans may think he has been one of the best catchers in MLB since he became a full-time regular.

Having contextualized Sanchez's career performance, 2020 was still a disaster. Sanchez had a normalized batting line 31 percent worse than league average (based on wRC+) and struck out in 36 percent of his plate appearances, the 6th highest figure among all players who had at least 150 plate appearances. The 36 percent figure is by far the highest of his career and about eight percentage points worse than his previous high (which was 2019). This was the first time Sanchez was below replacement level over a 150 plate appearance sample in his entire career. The only stretch of play that resembled this was May of last season where he regularly posted batting lines about 10 percent worse than league average. For Sanchez, who has been about 20 percent better than average from 2016 to 2019, posting this type of line even over a tiny small sample of 178 plate appearances is surprising and unlike anything we have seen from him.

The million-dollar question (and the one every member of the Yankees front office will be thinking about this off-season) is whether or not this abbreviated 2020 season is reason to panic. Making a rash decision or evaluation over 178 plate appearances, on its face, seems foolhardy. Just last season, players such as Austin Meadows, Franmil Reyes, Robinson Cano, Yuli Gurriel, and Joey Votto posted 60 game lines similar to Sanchez in 2020. Meadows was one of the best hitters in the American League, Gurriel finished with a line 26 percent better than league average, Reyes almost hit 40 home runs, Cano was great when he played this year, and Votto took a step forward in 2020 after a rough 2019 season. Good, even great hitters, are capable of having stretches like this. Someone reading carefully might counter and say that I am cherry-picking good players when in reality, the list of players with stretches similar to Sanchez's 2020 is littered with more bad players than good. That person would be correct. However, we have a demonstrably larger sample of Sanchez being a very good hitter as opposed to a below replacement-level contributor.

Still, I have not answered the original question: should we be worried about Sanchez's performance going forward? We need to address the strikeout rate and his performance on contact. Even though he made some incremental improvements in taking walks, striking out 36 percent of the time is not conducive to being an effective hitter. From 2018 through 2020, among players seasons with at least 178 plate appearances, just 11 players were able to post above average batting lines while striking out in at least 33 percent (about one third) of their plate appearances. The players that are able to overcome massive strikeout rates are among the best in the league at hitting the ball with authority. Players like Joey Gallo (2018 and 2019), Miguel Sano (2019), Brandon Lowe (2019), and Ian Happ (2018) have struck out at similar rates to Sanchez in 2020 and were above average hitters in those seasons. The list of players also includes Willy Adames (who has about league average batted ball stats but ran a 0.388 BABIP in 2020), Jake Cave (a solid fourth outfielder/AAAA guy who actually had characterisitcs that support high BABIPs until 2020), and Tyler Austin (the quintessential AAAA player with pop but not enough to offset strikeout woes). The common thread here (besides Adames and his outlier 2020 performance) is these players have well above average exit velocity readings and hard hit rates (balls in play at or above 95 MPH). When they make contact, they make it count. Sanchez, unsurprisingly, did not post good enough results on contact and saw a sharp year-over-year decline in 2020.

For context league average wOBACON (wOBA on balls in play or contact) in a given year is anywhere between 0.370 and 0.380. Expected wOBACON is based on the launch angle and exit velocity of a batted ball. The past three seasons, Sanchez has underperformed his expected wOBACON figures (though in 2020,  the leaguewide xwOBACON was much smaller than wOBACON which makes me think there was a calibration error in the model with the new HawkEye data or there was something weird with the baseball). Still, Sanchez actual results on contact were not what we should have expected based on his batted ball characteristics and much lower than league average and his career norms. Should we expect a player with barrel rates in the top five percent of the league in each of the past three seasons post below-average results when he puts the ball in play? Probably not. To say he was unlucky on balls in play in 2020 is an understatement. Sanchez posted a 0.159 BABIP, the third lowest figure in the past five seasons for player seasons with at least 170 plate appearances. The only worse seasons were Edwin Encarnacion in 2020 (0.156 BABIP along with a 13.2 percent barrel rate and 33 percent hard hit rate) and Ryan Schimpf in 2017 (0.145 BABIP along with a 16.5 percent barrel rate and 30.9 percent hard hit rate). Sanchez had a 17.4 percent barrel rate (97th percentile, after posting a 99th percentile figure last year) and a 49.5 percent hard hit rate (91st percentile). Not many players hit the ball as consistently hard as Sanchez while also posting barrels (the highest value batted ball type) at similar rates. Only six players in MLB posted both a higher hard hit rate and higher barrel rate than Sanchez in 2020. 

All of this is meant to show that we should expect a healthy amount of positive regression from Sanchez in 2021 from a batted ball perspective. Players who hit the ball like Sanchez are among the elite hitters in the league. Can Sanchez get back to that level in 2021? Even if he maintains this level of performance on balls in play, he needs to trim his strikeout rate to be an all-star level contributor. Is this 36 percent strikeout rate here to stay or is he the guy who strikes out about a quarter of the time, his career rate going into the year. The classic sabermetrician in me says he should probably fall somewhere in between those two marks in 2021, if anything closer to the 25 percent because for much of his career he performed at that level. The issue is we know strikeout rate is one of the fastest metrics that stabilizes quickly for a hitter. But do not mistake stability for predictability. Stability indicates the amount of time (in this case plate appearances) required for a stat to be able to adequately explain a player's talent over the prior sample. Predictability is the amount of time we need for a sample of a stat to explain the stat in a future sample with the same size. So we need to diagnose if Sanchez's strike out woes are the product of variance or something has changed in his "true talent" level and we should expect strikeout rates similar to 2020 going forward. Borrowing from my methodology in a previous post, I am going to look at the probability of a player posting a 36 percent strikeout over 178 plate appearances by pure chance for varying levels of "true talent". I simulated the 178 plate appearance sample 1,000 times for each true talent level. The following is the distribution of strikeout rates for varying levels of "true talent" strikeout rates: 

The dashed line represents an in-sample strikeout rate of 36 percent. Even for hitters with strikeout rates in the range of 25 to 30 percent, a hitter would be expected to strikeout at least 36 percent of the time fairly often. Furthermore, here is the percent of 178 plate appearance samples that showed strikeout rates exceeding 36 percent: 

A 30 percent strikeout hitter is expected to post strikeout rates at least as bad as Sanchez in 2020 about 35 percent of the time. For a 25 percent strikeout rate hitter (Sanchez's career rate going into the 2020), about a 15 percent probability. So posting a season like 2020, where Sanchez had 178 plate appearances, was definitely in the realm of possibilities just by pure chance. And if you look at his approach at the plate, nothing really changed in 2020. 

His overall swing rate barely changed from the past two seasons. He swung at slightly more pitches inside the zone while laying off more pitches outside of the zone compared to 2019. His chase rate was the best of his career. So I would argue his approach actually improved from 2019 despite much worse results. His zone contact rate was slightly down, but not so much so where you would see an influx in strikeouts. If anything, the increase in swings in the strike zone should have offset that and allow him to continue to put good hitter's pitches into play. Where Sanchez had a lot of problems was making contact when he chased pitches. While he has consistently posted out of zone whiff rates worse than league average, in 2020 Sanchez only made contact on 45.4 percent of his swings on pitches outside the strike zone, compared to a league average rate of about 60 percent. This was a large departure from his career rate going into the year (about 54 percent). The only way I can see this sustaining itself is if Sanchez is dealing with an issue of seeing the ball. But if he has a vision issue, how could he have both increased the amount of pitches he swung at in the zone while decreasing his chase rate. He improved at picking out balls from strikes. So the idea that he was not seeing the ball does not hold merit. I expect his contact issues to improve back towards his career norms going into the season, which should come with a corresponding decrease in strikeout rate. 

When digging into the data, Sanchez's awful 2020 results at the plate seem to be the result of poor luck and being on the wrong end of variance. Yankees fans have been especially tough on him and have called upon Aaron Boone to put him on the bench throughout the entirety of the shortened 2020 season. Given what know about his underlying performance and regression, I would expect Sanchez to have a bounce-back 2021 season. To believe that an all-star level contributor suddenly turned into one of the worst hitters in MLB would indicate a lack of understanding of the variance associated with outcomes in baseball and not appreciating the brevity of the 2020 season and its small samples. 

Anatomy of a Loss: Jacksonville Drops Second Game on Thursday Night

The Jaguars took on the Dolphins in last night's Thursday Night Football action. The Jaguars have been among the most talked about teams in the league going into the game. Going into the season, as the team gutted its roster, many were wondering how much the front office cared about competing this season and instead were looking ahead to the 2021 NFL draft and its ultimate two prizes, quarterbacks Trevor Lawrence and Justin Fields. Gardner Minshew, play-caller Jay Gruden and the rest of the Jaguars squad had other ideas. Catching double-digit points in week one, the Jaguars dispatched the preseason AFC South favorite Colts at home and followed up that performance with a narrow three point loss to the Titans in Nashville. The offense looked great on the surface scoring 57 points through two weeks of football. I will admit I was very surprised by this showing from Jacksonville. Jacksonville came into the game as a three point favorite despite their previous cellar-dweller expectations, but were thoroughly trounced by the Dolphins by three scores 31-13.

With that being said, the Jaguars torrid offensive start came with a large red flag, one that made me question their status as a three point favorite in last night's contest: their third down performance. Third down and fourth down performance is not an especially sticky measure, even in the realm of football analysis. The issue is multi-faceted. Play-calling is wildly different in these late downs situations because the scope of the play is changed. Instead of being able to sequence a play with another play or try to fool the defense and take a shot down the field, on late downs the goal is often just to move the sticks. Obviously on first and second down (the early downs) the offense still has the goal of another first down in mind, but the immediacy of needing that first down is totally different. Late down plays are more predictable because the goal of the offense on that play is much more clear to the defense. On early downs, for all the defense knows, the offense just cares about gaining a few yards on a run up the middle or it wants to catch the secondary sleeping for a chunk passing play. The other reason late down efficiency splits are not meaningful is the leverage. Offensive teams have the potential to add massive expected points totals on third and fourth down because the baseline expected point values are so low in these situations since more often than not the offensive will turn the ball over to the opposition in the form of either a punt or field goal. Since the magnitude of a single late down conversion (i.e. getting a first down and starting a new series) can be so large and late down plays are a smaller portion of the overall play sample for a given team compared to early down plays, late down splits should be regressed heavily towards the means.

We learn a lot more about teams based on how they play on early downs and how well they avoid these high leverage late down situations all together. The year over year consistency in late down performance is mostly noise, to the tune of a correlation coefficient of about 7.6 percent, based on data from 2010 through 2019. For more actionable information to help handicap the Jaguars going into last night's game, I looked at the relationship between a team's late down performance in weeks one and two and compared it to the rest of season late down performance for all team seasons from 2010 through 2019.

The first two week late down performance only explains 3.1 percent of the variance in late down performance for the rest of the season, so effectively it is noise. This is especially important when considering how Jacksonville's offense fared so well in weeks one and two and what that meant for week three and beyond. Even without the context of their early down offense, the Jaguars late down performance against the Titans and Colts was impressive. When you do incorporate that context, their rate of efficiency is even more eye-popping. 

The Jaguars, along with the Raiders, stand in a league of their own here. Despite middling early down efficiency figures, the Jaguars were the second best team in EPA per play on late downs. Furthermore through two weeks, the Jaguars ranked first in percent of third downs converted at 62.5 percent (about 50 percent better than league average thus far) and they converted their lone fourth down opportunity, bringing their total late down conversion rate to 64 percent which was still first in the league. This was all while posting middle-of-the-pack figures on early down EPA per play and early down first down rate. This was still better than expected performance for Jacksonville, but the Jaguars top-five offense through two weeks was a house of cards built on late down success. Here are there early and late down splits in weeks one and two compared to their opponents: 

Jacksonville, on a per-play basis, was out-gained handily in terms of yards and posted lower efficiencies in terms of success rate (percent of plays that produce positive EPA) and EPA. While the success rates on early downs are relatively close, the Colts and Titans combined posted almost double the expected points per play on offense compared to Jacksonville. On late downs, the margins more than just flipped: the Jaguars added more than three times that amount of expected points per late down play. Their late down EPA play was almost nine times their early down figure. 

I am burying the lead here, but this phenomenon completely vanished for the Jaguars against the Dolphins. Despite losing by three scores, the Jaguars were actually more efficient than Miami on early downs. The loss can be chalked-up to a reversal of fortune on late down plays: 

Jacksonville posted higher yards per play, success rate, and EPA figures on early downs. On late downs, the Jaguars could not even muster up two yards a play and lost almost an entire point per late down play, compared to Miami who posted figures similar to Jacksonville through week two. This game should have been much closer based on the more stable metrics when evaluating offenses, but Miami was able to get into manageable third downs and convert (based on the Dolphins elite efficiency compared to their low yards per play total). Jacksonville did the opposite and now both teams find themselves at with one win on the season. 

If Jacksonville hopes to remain competitive, the early down offense needs to be a tick better. They were never going to sustain their third down performance through two weeks, but something between that and what they did on Thursday needs to be the norm going forward. Given the personnel, middling offensive performance, and horrid defense, I would still say this is a Jacksonville team that is in for a long season. Maybe not a "first overall pick with a bullet" type season, but the Jaguars still have a lot of kinks to iron out before I would even consider them to be a playoff contender. 

deGrom's Run of Dominance has been Aided by Unprecedented Velocity Jump

Jacob deGrom is the best pitcher in baseball. He is coming off two straight Cy Young-winning seasons and is looking to add a third after a shortened 2020. By FanGraphs WAR, he has been the most valuable pitcher in MLB the past three seasons by almost three wins (18.7 WAR versus 15.9 from Scherzer) and in the last four he is second to Max Scherzer. deGrom's path to the majors was a bit unusual. He was a shortstop in college and made his major league debut at the age of 26. What has followed has been nothing short of brilliant. Not only has be been the most valuable, he is among the most consistent hurlers the league has to offer in terms of run suppression and keeping opposing hitters off the base paths.

How has he managed to be so consistent across multiple seasons? Major league hitters are obviously the best at what they do. In order to be so dominant, a pitcher like deGrom has to constantly be evolving. Whether it be by adding or tinkering with a pitch or altering your pitch mix, pitchers have to overcome the opposition's adjustments as they age. In conjunction with the league gaining familiarity on a certain pitcher, that pitcher also often combats the loss of velocity with age. Velocity is the ultimate equalizer in the world of pitching; it allows pitchers to get away with pitches right over the heart of the plate and can also make a fastball a viable two strike offering to put away batters. Most of the best pitchers in MLB have above-average to elite velocity and deGrom is no different. Even still, deGrom's velocity for a starter is highly unusual. 

deGrom has continued to add velocity as he has aged. In this shortened season, his average fastball checks in at about 98.5 MPH better than all pitchers who have faced 200 batters this season, save for Dustin May and his super sinker. When you account for his age, what deGrom's velocity for a starter is unparalleled. May is 22 years old, while deGrom is 32. Of the top 20 starters in average fastball velocity (minimum 200 batters faced) only five are above the age of 30. Velocity is an attribute that pitchers start to lose basically the moment they enter the league. The following represent the average year over year changes velocity by age bucket and the aging curve for fastball velocities: 

The first chart in the pair represents the average change in fastball velocity based on age year over year. The age on the x axis corresponds to the age in a given season. So, for example, age 25 indicates the change in velocity from age 24 to 25. The age averages were calculated using a weighted average of all pitchers in the age bucket. The weights are the total batters faced for each pitcher. The second chart shows the aging curve, where 0 represents the maximum fastball velocity and anything less than that is the velocity in the age bucket relative to that maximum. As I alluded to before, pitcher velocity peaks early and is in decline basically the moment that pitcher steps foot on a major league mound. Recall the chart with deGrom's average fastball velocity readings. deGrom, coming into the league at 26, has added velocity in each of the past five seasons (Statcast data goes back to 2015). For context of how unusual this is, here are the same charts of the league aging patterns versus deGrom: 

I cannot emphasize enough how amazing this is. deGrom is taking everything we understand about pitcher aging and literally flipping it upside-down. By continuing to add velocity at this point in his career, deGrom is ensuring his viability as an ace. Even if his velocity ticks down over the next few seasons, he will still be among the leaders for starting pitchers. If his fastball still plays up, his main secondary offerings in his slider and changeup will continue to rack up swings and misses. Whether or not he wins another Cy Young award, watching deGrom dominate in a unique way well into his 30s will be one of the most fun stories to watch for the rest of the decade.

The Cowboys Made an Unconventional Decision

The Cowboys won a wild 40-39 game against the Falcons this past Sunday. After fumbling and losing the ball on three of their first four drives (the other was a three and out), the Cowboys found themselves down multiple scores early. After the third fumble, the Falcons kicked a field goal with about a minute left in the first quarter, to bring their lead to 20 points despite going three and out on their first two possessions themselves to start the game. Atlanta was able to go up 20-0 without having to piece together a drive of more than 52 yards (the scoring drives were 22, 52, 5, and 31 yards in the first quarter). Atlanta was in the drivers seat from this point through the end of the first half, where they went to the locker-room with a 29-10 lead over Dallas.

From then on, Atlanta did everything in their power to give the game away. Dallas opened the second half with two long touchdown drives (74 and 83 yards) in just over five minutes combined. Atlanta then scored a touchdown of their own and a field goal bringing the score to 39-24 with 7:57 left in the fourth quarter. The Cowboys then scored with three minutes left in the game, shrinking Atlanta's lead to nine points pending the PAT. What followed was one of the more surprising coaching decisions in recent memory. Conventional wisdom says the Cowboys kick the extra point, putting them down eight points with three minutes left and hope to get a stop and a chance to score a touchdown and tie the game with a two point conversion. The Cowboys instead took the road less travelled. They attempted the two point conversion following the touchdown and failed to convert, resulting in a nine point deficit with three minutes to go. What happened after has been talked about ad nauseam: Atlanta went three and out, the Cowboys marched 76 yards in just one minute and eight seconds, recovered an onside kick in the most ridiculous manner possible (Zuerlein used no kicking tee and the Falcons seemed to avoid the ball like it was a punt) and the Cowboys swiftly moved the ball into field goal territory, sealing the game as time expired. One of the main talking points after the game was of course the brutal collapse by the Falcons. But since this game involved the Cowboys, I would argue the most bandied about topic was the fact that the Cowboys chose to go for two after cutting the deficit to nine points. Many in the media lambasted the decision by Mike McCarthy, including noted NBC pundits Mike Florio and Chris Simms. I wanted to look at the decision and discuss both how unusual it was and how it was definitely the correct decision.

First, teams almost never behave as the Cowboys did when trailing by nine points. Since 1999, there have been 23 instances of a team cutting their deficit to nine with between eight and five minutes left (basically all situations similar to what Dallas faced on Sunday). In just three of those situations, the team elected to go for two: Carolina against the Rams in 2001, Buffalo against the Ravens in 2019, and Cleveland against the Ravens in 2019. So before 2019, in the timeframe where I have access to play-by-play data, a team elected to go for two just once. In all three of the cases, the team that elected to go for two lost the game. In the other 20 instances, when the team kicked the extra point, those teams won seven times and tied once, for a 0.375 win rate. To further highlight how unusual this decision was, since 1999 when a team scored a touchdown to trail by nine points at any point in the game, the team kicked the extra point 90.2 percent of the time (130 times out of 144).

But let's go back to the situation at hand. Given we only have three instances where a team went for two in a similar situation to Dallas, using the fact that none of them won the game does not hold merit (and that is without accounting for the fact that the Rams in 2001 and Ravens 2019 both went 14-2, thus proving to be a better-than-average opponent). Without a sufficient sample size where we can just analyze the results on the field, we have to turn to both logic (for the explanation) and win probability added (for evidence).

With regards to logic, I will turn to a brief column supporting the decision from Brian Burke, who is a member of ESPN's Stats & Info group and the preeminent forerunner in the football analytics movement. While citing the piece might seem like an appeal to authority, his logic is sound and I encourage you to read it. It basically boils down to this: when down 15, a team knows it needs to score two touchdowns where one is followed by a converted two point attempt to tie the game. So the team at some point is going to have to go for two, assuming it scored the requisite two touchdowns. Most teams, as I alluded to above, prefer to kick the extra point first and delay the potential of being eliminated from the contest and then wait to go for two after a potential second touchdown. Burke's argument (which is generally accepted by analytics-types) is that instead of waiting, immediately going for two gives the team immediate feedback on its chances of winning the game, forgoing the period of uncertainty about the success of going for two, and allows the coaching staff to operate in such a way that maximizes win probability for the remainder of the game without the uncertainty of the result of the two point conversion. Either way the Cowboys would have to go for two at some point if they hoped to tie the game down 15. Going for two early gave the Cowboys additional information and allowed Mike McCarthy and Kellen Moore to make decisions and play-calls that maximized their chance of winning the game. Coaches too often are scared of embarrassing themselves and losing a game earlier than many would expect (this is the sort of logic that led to Anthony Lynn punting the ball to the Chiefs in a sudden-death overtime). McCarthy tossed optics aside and made the decision that gave his team the best chance to win.

If this sort of logic does not seem sound, or you scoff at my appeal to authority, let me present you with some data to back up Burke's logic and show that going for two is the correct decision. First, let us assume that when kicking the extra point, the Cowboy would be successful. This is not actually true: since 2015 when the extra point was moved back, teams have been successful on about 94 percent of the attempts. For ease of analysis, however, we are going to assume that the Cowboys would automatically get the point if they kick the PAT. Kicking the PAT would put the Cowboys down eight points. By going for two, the Cowboys would be faced with one of two scenarios: they would be down seven points or the deficit would stay at nine points. Since 2015, teams have been successful on about 47.9 percent of their two point attempts. For rushes, the success rate is 54.9 percent and for passes a 45.6 percent success rate. For additional context, the Cowboys did run the ball on their two point attempt, but lets assume that they are an average team in this situation with average play-calling tendencies, so the probability of success is 47.9 percent. Next we consider the win probability added in each situation relative to the win probability at an eight point deficit. Since 1999, teams that are down eight points at a similar point in the game to the Cowboys have a win probability of 13.5 percent. Down nine in the same situation, win probability goes down to 12.3 percent. When down seven points, the team's win probability increases to 18.5 percent. For this information in tabular form, see the following:

Converting the two point attempt would add about five percentage points to the Cowboys win probability. Going for two and failing would cause the Cowboys to lose just 1.2 percentage points of win probability. Therefore, the magnitude of win probability added in the case of success is 4.17 times larger than the magnitude of win probability lost due to failure. Using a similar formula to the one for calculating the break-even point of attempting a stolen base, but instead of using run expectancy we use win probability added, the break even rate for going for two in the Dallas' situation is just 19.4 percent. Remember, teams convert on two point conversions at a 47.9 percent rate. That is about 2.5 times the size of the break-even rate and that is without considering the extra benefits of running the ball on such an attempt. So, without using any of Burke's logic, the win probability arithmetic shows that going for two in this scenario is actually the overwhelmingly obvious decision here. Kicking the PAT in this situation makes no sense and McCarthy should be commended for both defying conventional wisdom and calling for the smart play, despite not actually converting the two point attempt. And yet, the Cowboys still won. 

Many are going to point to this unorthodox decision and criticize McCarthy because what he did flew in the face of conventional wisdom. In a sport where front offices still choose running backs in the first 10 picks, teams run the ball often on early downs, and coaches make decisions based on the optics if that decision fails, seeing McCarthy make this decision in such a pivotal game for the Cowboys was refreshing. Hopefully other teams learn from this and see how it was the correct decision even though they failed. The fact that the Cowboys eventually won the game, for better or worse, might nudge others in the right direction (making decisions based on results instead of process puts you at the mercy of small samples and/or variance in outcomes). Even if teams choose to go this direction because of the result, us viewers will be in for a better viewing experience with more efficient football. 

play-by-play data via nflfastR

Anatomy of a Loss: The Chargers Fall in Overtime to the Chiefs

The Chiefs and Chargers squared-off in the Chargers' first game in the newly built SoFi stadium on Sunday afternoon. The Chiefs entered the game as heavy favorites to the tune of 8.5 points, carrying an implied win probability of about 77.5 percent. After Kansas City's thrashing of the Texans in the opening game of the 2020 season, the Chiefs struggled to put away the Chargers. Despite these struggles, Kansas City was still able to pull out a narrow 23-20 victory in overtime.

In some ways, the Chargers performance was admirable given the circumstances. Tyrod Taylor was a late scratch so 6th overall pick Justin Herbert was tasked with starting his first NFL game against an offensive juggernaut in the Chiefs. Expecting a rookie quarterback to keep pace with Patrick Mahomes is a daunting task, but Herbert was up to it matching Mahomes' performance throughout the contest.

While it was not the best performance we have seen from Patrick Mahomes, especially through the air, it was still impressive to see Herbert post a better line through on pass attempts, when accounting for penalties. Mahomes was able to add a ton of value on scrambles, most notably on third and 20 with 54 seconds left in regulation, adding about 2.20 points of value and bringing the Chiefs into field goal range so they could tie the game.

While it was nice to see the Chargers, led by Herbert, play well-enough to win against the Super Bowl favorite (according to the Vegas odds as of the writing of this post), the goal of any game is to win. A team that was not favored to be in the playoffs before the season started even with the addition of a third wildcard cannot afford to drop games where it is winning for the majority of the snaps, especially against a division foe. A couple of end game decisions by Chargers coach Anthony Lynn (who has been among the least aggressiveness head coaches on fourth downs and has had prior issues with clock management) peaked my interest and I wanted to check if they were contributing factors to the Chargers losing this game.

The first decision was kicking a 23 yard field goal from the Kansas City five yard line to put the Chargers up 20-17 with 2:27 left in the fourth quarter. My initial reaction was this was a mistake. The Chargers, as I mentioned above, were heavy underdogs going into the game. Scoring a touchdown in this situation would put them up seven (assuming they hit the ensuing extra point) and would give the Chiefs no room to win in regulation. The Chiefs, with 2:27 left, would have to score a touchdown to even get a chance at overtime. By kicking a field goal, the Chargers allowed the Chiefs an additional avenue to win the game. If the Chiefs were to score a touchdown, they would almost definitely win. To send the game into overtime, all they needed to do was kick a field goal. I went back through all the play-by-play data I had access to (back to the 1999 season) and isolated games where a team was in a similar situation as the Chargers. I filtered for games where a team kicked a field goal of less than 40 yards when they had less than six yards to go for a first down and there was between a minute and a half and three and a half minutes left in the fourth quarter. I had to make the time interval large enough to extract a large enough sample. Teams (of which there were 19) in the Chargers' situation won the game about 63.2 percent of the time and went to overtime 26.3 percent of the time. The issue I had when trying to do the same analysis with going for a touchdown was that every team in the Chargers situation decided to kick a field goal. To try to get around the issue, I looked at drives that started with between two and three minutes left in the fourth quarter and where the team on defense was up by seven points. I found 33 drives that met this criteria and the team on defense only won about 57.7 percent of the time.

This is not intuitive. How can a team have a worse chance of winning if it scores a touchdown? There are two issues. The first is sample size. The sample of teams that kicked a field goal in this very specific situation was 19 and they were only 33 teams in my touchdown sample. In both situations, making sweeping judgments based on such a small portion of all the games since 1999 is bound to lead to some faulty conclusions. The small samples also lead to factors heavily influencing the overall results of the samples, such as the specific teams involved. This brings me to the second issue: the teams involved. The Chiefs, since Mahomes became the starter, have scored a touchdowns on a shade over 40 percent of their drives. By kicking a field goal, the Chiefs would have the opportunity to essentially seal a win in regulation by scoring a touchdown. When you factor in the rate of drives that end in a field goal for the Chiefs, we are talking about a slightly greater than 50 percent chance that they either win the game or send it to overtime, the latter of which you would still favor the Chiefs to win . One might quibble that this does not account for the lack of time remaining (about two and a half minutes) but I would counter that the time constraint forces the Chiefs to have to throw the ball more than they would otherwise and the Chiefs passing offense with Mahomes has been among the most prolific in NFL history. When you consider my small sample analysis of teams in the Chargers situation and the team the Chargers were facing, I find it difficult to be too hard on Anthony Lynn in this spot. I can see the argument in either direction, so pinning the loss on that one decision might be a bit harsh. Furthermore, when using the win probability calculator at pro-football-reference, based on the changes in win probability for scoring a touchdown and kicking a field goal, the offense would have to be able to score a touchdown more than 57.4 percent of the time. This is higher than even the two point conversion rates. Thus, kicking the field goal was actually the correct decision here.

The second decision major decision by Lynn was much more indefensible. After winning the coin toss in overtime, the Chargers started their drive on their own 25 yard line after a touchback. The Chargers ran the ball on first and second down then threw the ball for six yards, leading to a fourth and one on their own 34. Lynn had two choices: go for it on fourth down or punt. By going for it and not converting, the game would effectively be over because all the Chiefs would have to do was kick a field goal to end the game. A field goal from the 34 would be about 51 yards, which has been converted at a 56.4 percent rate since 1999. At any distance closer, that percentage increases. Move the ball 10 yards and the rate is 78.3 percent. Another 10 and it's 83.4 percent. Obviously a touchdown would end the game also. Basically, no matter how you slice it, going for it and not converting almost definitely means a loss. Punting the ball also means all the Chiefs have to do is either kick a field goal or score a touchdown to win. Like I said before, the Chiefs over the last two years have scored in some form on a shade over 50 percent of their drives. What this does not account for is garbage time. With the game on the line, you would assume the Chiefs would be very aggressive throwing and moving the ball downfield. They have spent such a large portion of the past two years ahead, thus many of their drives involved trying the churn the clock as opposed to trying to score. When the score is tied, following a punt the past two years the Chiefs have scored on 72.7 percent of their drives, per pro-football-reference, a figure 11.9 percentage points better than the next best team. So by punting, the Chargers were facing overwhelming odds to win the game.

This brings me to my next point. By not going for it in this situation, all Lynn was thinking about was not losing the game in that moment. Going for it close to your own endzone and might seem bad at first blush. In that case you are almost guaranteeing a loss. But punting is not much better; given the opponent you are also giving yourself almost no chance to win but you avoid the optics of a potentially disastrous fourth down play. And I have not even considered what happens if you do convert fourth down. On fourth down plays with one yard to go in between your own 30 and 40 yard line, NFL teams have converted at a rate of 72.2 percent since 1999. A 72.2 percent chance of continuing the drive makes going for it much more appealing than simply punting the ball back to this Chiefs team, given their scoring prowess. Using the win probability calculator at pro-football-reference, the Chargers would have a 51.1 percent chance of winning the game if they converted. Using that same calculator, the Chargers win probability was just 38.1 when punting and 19.4 percent upon a failed fourth down. When you account for the swings in probability for winning the game by converting and failing to convert the fourth down, the break even conversion rate for the Chargers is just 41.1 percent, more than 30 percent lower than the odds of them actually converting. Not going for it was so obviously the wrong decision I am still shocked Lynn decided against doing so. Though the decision did not totally seal the Chargers' fate, it effectively did so. It is amazing in the year 2020 we still have coaches making nine figures still make these types of decisions because they are worried about the optics of failing a conversion. If the Chargers are on the outside looking in for the playoffs at the end of the season by just a game, this decision against a division rival will loom large on Lynn's conscience.

Week 1 Penalty Trends

The 2020 NFL season is an outlier. With the ongoing global pandemic, no preseason games were played leading up to week 1, training camp schedules were modified, mini-camps and OTAs were cancelled, and the roster rules were changed to accommodate these (tell me if you have heard this before) extraordinary circumstances. The changes to how teams could practice I assumed would be felt even more by teams with new coaching staffs or personnel changes at important spots (mainly quarterback, receiver, and defensive back). Also, despite everyone's craving for football while we are all stuck mostly confined in our homes, I thought the play may be a bit disappointing. Without the proper time to development the necessary cohesion and synergy between teammates and the coaching staff, I was expecting a lot of sloppy play. To my eye (and my surprise), this was not the case in week one. Some may point to a bevy of missed tackles that would not have occurred without a more rigorous official training schedule and preseason, but without the data to back up such a claim I cannot say for sure if there were more missed tackles than normal. Another way to evaluate the quality of play is by looking at penalties across the league and fortunately, with access to play-by-play data, this is something I could investigate.

I compared the total number of penalties that I subjectively thought of as indicative of crisp play for every week one since the 2010 season. Before I dig into those, I first aggregated the total number of penalties in each season during week one.

Week one this season had the lowest number of penalties since 2010. Given my prior, I was shocked. The narrative that teams would not be as sharp, at least in regards to penalties, does not hold up to scrutiny. As I alluded to above, I broke it down my penalty type to see if I could find any aberrations in the data from this season.  

False starts by the offense in 2020 was on the low end of outcomes in the sample, similar to the levels in 2010. Defensive offside was way down, but neutral zone infractions were at an all-time high, tied with the number in 2019 (14 total penalties). Given the small sample of neutral zone infractions in a given week, I was not sure whether I should chalk this phenomenon up to defenders being lackadaisical when lining up pre-snap or just random variation.  If you were being more aggressive in parsing the data or more attached to your prior, I am sure you would subscribe to the former idea. Given the fact that there was literally no change from 2019, however, and I would say this is noise.

Delay of game penalties are a function of either the quarterback and play-caller not being on the same page, players not knowing where to line up, or just inattentiveness by the quarterback with regards to the play clock. We saw the most delay of game penalties since the 2012 season, another possible indicator of teams suffering from a lack of practice. Still, there was only one more delay of game penalty compared to 2019 and only two more than the 2015 and 2016 seasons. One delay of game penalty across 16 games is only 0.06 more penalties per game. Again, nothing here.

 Holding penalties were way down compared to prior seasons. The degree to which they were down makes me think that the referees were making an effort to call less of these in the 2020. This is something I will keep my eye on as the season progresses. Still, using holding penalties as evidence of sloppy plays is without merits.

Roughing the passer penalties are a good gauge of how disciplined pass-rushers are when attacking opposing quarterbacks. After a spike in 2018 when the NFL changed what constituted a roughing the passer penalty, this type of penalty was called at a similar level to 2019 and seasons prior to 2018. Given the set of rules, it seems defenders were just as disciplined as they would be with more practice and preseason games.

After looking at all these penalties, the idea that the unusual off-season schedule would affect the quality of play is not true when you dig into the data. Using penalties as a proxy for sloppiness is not the only way to evaluate the play in week one. We can look at offensive performance, special team performance, or turnovers to find any reason to think that the play was objectively worse this year than in prior years. EPA per play was actually higher in 2020 than in 2019 (0.0466 in 2020 versus 0.0459 in 2019). 71.6 percent of all field goals were made, compared to the previous rates of 84.2, 82.5, 88.1, and 82.1 percent in the prior four seasons. Overall, the rate in 2020 was the lowest since 2010 so if you want to use that as justification that preseason or extra practices were missed I guess that is warranted. But when we are enjoying football on Sundays, we care more about how the offenses perform. In 2020, there were 34 total fumbles and 20 total interceptions. These were both the lowest totals since 2016. So, pointing to turnovers is not grounds for disparaging the quality of play. With all of this being said, I think we can conclude that we saw characteristically great football in week one even without normal camps and preseason. Going forward, this may provide evidence that we do not need the preseason nor the extensive practice schedules characteristic of NFL teams. Coaches may hate this, but I am sure the players would welcome the rest. It will be interesting to see what this means in the future. Given the success of week one, I wonder if the league will dangle less preseason games and practices to get concessions from the players side in the next round of CBA negotiations with the knowledge that losing the preseason and some practices will not hurt the product on the field. The CBA expires after this season, so starting in March, keep an eye on this as the two sides negotiate and battle through the media.

Anatomy of a Loss: Philadelphia and Detroit

The Eagles and Lions suffered brutal losses on Sunday at the hands of Washington and Chicago, respectively. The Lions, with a healthy Matthew Stafford, had been receiving some buzz as a playoff sleeper going into the year by the more analytical types in the football universe. That is not to say that the Lions are expected to be world-beaters this season; but taking care of the Bears in week one was supposed to be the first step in realizing these playoff aspirations and the market indicated such; they were giving three points in their game Sunday afternoon. The Eagles, despite their rash of injuries in camp along the offensive line, were expected to easily dispatch this Washington squad as they looked to keep pace with the Cowboys in the NFC East. It looked like they were going to do such, racing out to a 17-0 lead in the early going. After that point, however, Philadelphia seemingly lost its ability to effectively move the ball against a young Washington defense. I wanted to look at these two losses, figure out why they happened, and what we should think moving forward for these two playoff hopefuls.

I will start with Eagles. The Eagles took their commanding three score lead with 6:54 left in the second quarter. According to the win probability model provided by the NFL data package nflfastR, which takes into account the pregame spread, the Eagles had about a 93 percent chance of closing the game out with a win. From there, it got ugly for Carson Wentz and co.

The collapse column groups the offensive plays for each team before and after point at which the Eagles scored to bring their lead to 17. Throughout the game, the Eagles were extremely aggressive throwing the ball, possibly due to the absence of Miles Sanders. As they built their lead, they threw the ball on 71.4 percent of their offensive plays and the offense was humming to the tune of a 6.07 yards per play and a 0.384 EPA per play, both elite figures. Washington, unsurprisingly, was having trouble on offense. While the 4.06 yards per play mark is certainly less than impressive, the EPA per play mark of -0.345 in the early going was disastrous. After Goedert's touchdown in the second quarter, the Eagles did not just regress, they looked like a totally different team. They were only able to muster a paltry 2.38 yards per play and gave back more than three quarters of a point per play despite becoming more aggressive throwing the ball. Wentz threw two picks and took 8 hits as the Washington defensive front had its way with the hurting Philadelphia offensive line. It is difficult to give Washington much credit here: it was still not able to move the ball effectively after the Goedert touchdown, but with the Wentz interceptions giving them some favorable field position they were able to capitalize on the Eagles ineptitude.  One issue that was consistent throughout the game for Philadelphia was getting to manageable third downs. 

On either side of the collapse the Eagles were, on average, facing a shade over 10 yards to go per third down play. At that point, it is asking a lot for the offense to continue to move the chains. The big difference in EPA before and after the infamous point in the second quarter was the Jalen Reagor 55 yard catch off of play-action. That play itself was worth almost six points (5.88 EPA) or a touchdown on the expected scoreboard ledger. With such a small sample of plays, if you were to pull that out of the sample or replace it with an average third down play, the picture is much more grim. The best way to fix these third down woes is to avoid third and long situations all together. Wentz was not able to sequence completions in the short and intermediate areas of the field throughout the game, thus Philadelphia's backs were up against the wall throughout the afternoon. 

You want to give some credit to Washington for fighting back in this game, but this loss falls squarely on the shoulders of the Eagles offense. Wentz was awful for a lot of the game. Posting a -0.758 EPA per play on those dreaded 40 plays is a recipe for disaster. By the end of the game, the Eagles had posted the worst offensive EPA figure in the entire league in week 1. While Washington has invested a ton of resources into its front seven, the back-end of their defense is still among the worst in the league. The fact that the Eagles had so much trouble exploiting this weakness is troubling. If you are an Eagles fan looking for a silver lining, I would say two things: first, we are talking about one game out of 16. The Cowboys lost this weekend too so they are not chasing a game after one week. Also the offensive line hopes to get healthier and develop some cohesion as the new faces are fully-integrated. 

However, Wentz looked bad and this is not a good Washington team. If the Eagles continue to struggle to protect Wentz, the Eagles will struggle to make the playoffs in a loaded NFC. It was odd that Doug Pederson continued to throw the ball so often with such a large lead, but with Sanders hopefully back in the fold for week 2, maybe he leans more on the run game to bleed out the clock with the lead in the future. I should note that their commanding lead happened early on in the game, so even if they did have more trust in the running game, you cannot expect to win a game while trying to churn clock for two and a half quarters. The Eagles have a big test next week when they take on the Rams, who looked good against the Cowboys on Sunday night and came away with a win despite some characteristically questionable fourth down decision making by Sean McVay (punting on fourth and 1 from the 50 with a small lead and kicking a field goal on the Dallas 15 with only two yards to go). If Philadelphia plays like they did on Sunday they will be 0-2, already on the outside looking in for the playoffs and the pressure, if it has not already, will be turned up to the nth degree for Doug Pederson, Carson Wentz, and the rest of the Eagles organization.  

Now, let's look at the debacle at Ford Field. The Lions went up 23-6 after a Matt Prater field goal with 3:19 left in the third quarter. The win probability for Detroit at that point was 97 percent. From there the Bears scored 21 unanswered points to put the game away. The difference in performance of each team before and after this point where the game seemed to be decided, similar to the Philadelphia game, was stark. 

The Lions were woeful after an excellent start to the game and Trubisky caught fire after putting up a typical performance through three quarters. The one sack after the 23-6 lead for Stafford does not look too bad without any context. When you add in the necessary context, however, this changes. Stafford took a sack with 5:39 left in the fourth quarter up 10 on a second and eight from the Chicago 33 yard line. Firmly in range to kick of field goal to make it two touchdown game even following an incomplete pass, the Lions were instead stuck at the Chicago 42 on third down where they presumably would have to punt, pinning the Bears deep into their own territory. Trubisky would be then required to lead two long drives for the Bears to be able to win the game. Matt Patricia had other plans. Stafford threw a five yard pass to DeAndre Swift to bring the Lions to the Bears 37. Patricia trotted out Matt Prater to attempt a 55 yard field goal. Since 2010, field goals within a yard of a 55 yard attempt have made successful on about 57.5 percent. Multiple 0.575 by 3 and you have 1.725 expected points on such an attempt. Add 1.725 to 10 and you get a 11.725 expected score differential after a 55 yard field goal attempt. There is not much of a functional difference between a 10 point margin and a 11.725 expected margin. In either case, Chicago would have to score two touchdowns to win the game. The downside is by missing the kick, you give Chicago a short field which makes it much easier to score two touchdowns in five minutes. Sure enough, Prater missed the field goal and Chicago scored touchdowns with 3:03 left and 2:00 left after Stafford threw an interception at Detroit's 42 yard line with 2:45 left. Detroit still could have won the game. The Lions got the ball back with a shade under two minutes left, drove the ball down the field, and DeAndre Swift dropped a pass in the end zone with 11 seconds left. Stafford did not do Patricia any favors down the stretch by taking a crucial sack and throwing a costly interception, but this field goal decision by Patricia was arguably the most costly decision of the day, given Detroit would not have been in such a precarious position to begin with. Patricia has been one of the worst decision-makers on fourth down during his tenure in Detroit: 

In this game alone he punted the ball on fourth and one from their own 34 yard line, kicked a field goal on fourth and three from Chicago's 26, and kicked a field goal in the aforementioned situation at the end of the game. Patricia needs to be successful to keep his job in Detroit this year and make a push for the playoffs. Decisions like these are counter-intuitive in reaching this goal. If Patricia keeps making these types of decisions in high leverage situations, the Lions may leave some more wins on the table and Patricia will be another victim of the vicious NFL coaching carousel.

All play-by-play data via nflfastR