"Who's the best quarterback in the NFL" is a perpetually asked questions by fans, analysts, and commentators of the NFL. As this question can be subjective, there's always been a strong desire to use support QB stats and analytics as proof for a particular point of view.

In reality, there is no perfect metric for measuring QB play, as shown by our analysis - What are the best metrics for NFL quarterbacks?.

nfelo uses its own QB ranking model for generating predictions, which is how the list above is ordered, but the best approach for determing the best Quarterback is generally a holistic one that considers a wide variety of metrics ranging from traditional stats like yards, touchdowns, and interceptions, to more advanced ones like Expected Points Added (EPA), QBR, and Completion Percentage Over Expected (CPOE).

The table on this page hosts a wide variety of these metrics to allow fans to explore the data and make their own decisions.

nfelo's QB Elo

The table is sorted by nfelo's QB Elo model, which is a continuation of the work done by the now defunct 538. This model is meant to be less descriptive (ie how well did this QB perform in the past) and more predictive (how will the QB perform in the future).

Like Passer Rating, it uses traditional counting stats, and devides them by dropbacks to create a measure of efficiency. Unlike Passer Rating, it also takes into account a QB's rushing performance, which has become an increasingly important part of the game.

The stat is also represented relative to the league average, which makes it more "era-adjusted" than other nominal metrics, which suffer from passing stat inflation in the NFL.

Traditional Passing and Rushing Stats

This section includes traditional passing stats, like Passing Yards, Passing Touchdowns, Interceptions, and Yards per Attempt. It also includes rushing and sack information.

Why Traditional Stats Still Matter: While advanced metrics attempt to improve over traditional stats, traditional stats remain important for several reasons:

• They are well understood
• They are highly available, especially for historical data
• They still do reasonably well as measures of performance. It may surprise you to know that Passer Rating is just as stable and predictive as EPA when it comes to measuring QB performance.

Completion Percentage Over Expected (CPOE) - Advanced Accuracy

Included alongside the traditional stats is one advanced stat -- Completion Percentage Over Expected, or CPOE for shor. CPOE measures how much higher a quarterback's completion percentage is relative to what we'd expect based on the types of throws attempted by the QB. Deep throws and throws to the sideline are generally completed at a lower rate. CPOE rewards quarterbacks for completing these types of difficult throws, while penalizing them for missing easy throws like screens and check downs.

You can read more about CPOE and other "Over Expectated" metrics in our post Over Expected Metrics Explained -- What are CPOE, RYOE, and YACOE.

Rate Statistics - Efficiency Metrics

Rate stats measure the percent of attempts or dropbacks that resulted in a particular outcome. For instance, TD% represents the percent of pass attempts that resulted in a passing TD. Success rate is the percent of dropbacks that resulted in positive EPA, and Exp% is the percent of dropbacks that were explosive plays.

Key Rate Statistics:

• Touchdown Rate (TD%) - Percentage of pass attempts resulting in touchdowns
• Interception Rate (INT%) - Percentage of pass attempts resulting in interceptions
• Success Rate - Percentage of dropbacks that created positive EPA
• Explosive Play Rate (Exp%) - Percentage of dropbacks resulting in explosive gains (15+ yards)

Why Rates Can Be Preferred Over Totals: Passing stats can be biased by a QBs situation. A QB on a team with a bad defense may need to throw more, which inflates their totals. Rate statistics cut through some of this bias by measuring efficiency more directly

All-in-One QB Rating Systems

The ratings section is a collection of popular methodologies for quantifying the quarterback position. This includes traditional passer rating, Adjusted Net Yards per Attempt (ANY/A), nfelo's custom QB Elo metric, and ESPN's QBR where available. Two notable omissions from this section are Approximate Value from Pro Football Reference and DYAR from Football Outsiders. Since these models are not open source, or easily acquired, they unfortunately can't be included here.

There is no perfect science to determining which all-in-one rating metrics are best for NFL QB Rankings, as they all have various strengths and weaknesses. Generally speaking, all of these metrics tend to correlate with each other fairly closely, so you can't really go too wrong.

Rating System Breakdown:

• Passer Rating - Traditional metric emphasizing completion percentage, yards per attempt, TD%, and INT%
• ANY/A (Adjusted Net Yards per Attempt) - Yards per attempt adjusted for TDs, INTs, and sacks. Unlike Passer Rating, it also takes into account rushing yards
• QB Elo - An predictive system meant to support an elo based team strength model. While it incorporates similar stats to other measures, it is tuned for prediction, and therefore values them in different ratios. As an example, QB Elo has a strong bias towards rushing quarterbacks relative to other measures.
• ESPN QBR - Proprietary metric emphasizing clutch performance and EPA-based concepts

Expected Points Added (EPA)

EPA stands for Expected Points Added. It measures the quarterback's contribution towards offensive production by taking game context (down, distance to go, score, etc) into consideration. This is expressed as both a rate (QB EPA per Dropback) and a total.

Why EPA Matters: EPA is, perhaps, the most commonly used advanced metric for measuring quarterback performance. The analytics community likes quarterback EPA because it accounts for contextual information. For instance, completing a low depth of target pass short of the sticks on third down is not materially worse than throwing a deep interception as the opponent will start their next drive in approximately the same part of the field. While EPA can account for this sort of context, metrics like Passer Rating or ANY/A will penalize the interception and reward the short pass.

Total vs Per-Dropback EPA: Just like any other stat, EPA can be measured in total or on a per dropback basis. Total QB EPA is good for measuring the aggregate impact of a quarterback, while QB EPA Per Dropback is a better measure of efficiency.

For instance, Josh Allen might lead in total EPA due to high volume, while Aaron Rodgers leads in EPA per dropback due to exceptional efficiency. Both metrics provide valuable insights depending on your analysis goals.

EPA Components - Situational Breakdown

EPA Components look at where a QB's EPA production came from by comparing their component averages to the league average. For instance, a QB with a +0.02 Sack EPA vs League Average lost 0.02 points less per dropback due to sacks than the average quarterback. A QB with a positive Sack EPA component would be considered better at sack avoidance.

Average Depth of Target (aDOT) Analysis

Depth stats look at how far down the field the QB threw on average. aDOT stands for Average Depth of Target, while "vs Sticks" looks at target depth relative to the first down marker.

aDOT Interpretation: Typically playmaking QBs with a lot of athletic ability like Russell Wilson, Lamar Jackson, and Josh Allen can create more time to throw down field, while pocket QBs who rely on timing and precision like Tom Brady, Matt Ryan, and Mac Jones throw shorter passes. The offensive system implemented by the head coach or offensive coordinator can also impact aDOT. For instance, under Matt LaFleur, Aaron Rodgers went from being a high aDOT QB to low aDOT QB.

Coaching Impact on aDOT: The drop in Aaron Rodgers's depth of target came with an increase in yards after the catch (YAC) and completion percentage. Despite throwing shorter passes on average, Rodgers had more passing yards per attempt, and won back-to-back MVPs. Patrick Mahomes and Justin Herbert are two additional examples of quarterbacks with exceptional arm talent who actually play an efficient style of football predicated on YAC over aDOT.

aDOT vs Sticks Analysis: "Vs Sticks" measures how many yards short of the line to gain the QB threw on average. While this is usually used in negative ways to highlight the conservativeness of QBs like Alex Smith, nearly all QBs in the NFL have negative aDOT vs the Sticks on average.

For these reasons, it's important to view QB aDOT and aDOT vs Sticks in the context of a quarterback's broader performance. High or low aDOT can simply be the result of play style or scheme, and it's a common misconception that pushing the ball down the field is the mark of good quarterback play.

Winning and Clutch Performance

Winning encompasses high level stats like wins, losses, and ties, in addition to advanced metrics like Win Probability Added. QBs that make big plays in critical moments (think Joe Burrow in his 2021 Super Bowl run) tend to produce the most WPA.

Win Probability Added (WPA): WPA measures how much each play increases or decreases a team's chances of winning. Fourth-quarter comebacks and late-game touchdown drives generate significant positive WPA, while costly turnovers in crucial moments create large negative WPA.

In addition to WPA, the table also includes combacks and comback opportunities, as defined by instances where a QB had possession down one score or less in the 4th quarter.

It may not be terribly surprising, but perhaps the two most impressive QBs from a comback and comback conversion perspective are Tom Brady and Patrick Mahomes. They didn't often trail in the 4th quarter, but when they did, they had the best chances of coming back and winning.