Terry Bradshaw finished his career with 212 touchdowns, 210 interceptions and a 70.9 passer rating. Kurt Warner threw 208 touchdowns against only 128 interceptions, and his 93.7 passer rating ranks 8th in NFL history and 2nd among retired players. But Bradshaw played from 1970 to 1982, while Warner played from 1998 to 2009. As a result, comparing their raw statistics holds very little meaning. Comparing across eras is very challenging, but not impossible. And in this case, once you place the numbers in the proper context, Bradshaw’s numbers were arguably more impressive than Warner’s numbers.
Let’s start with Bradshaw and begin by looking at his Relative ANY/A for each year of his career. For new readers, ANY/A stands for Adjusted Net Yards per Attempt, defined as
(Gross Pass Yards + 20 * PTDs – 45 * INTs – Sack Yds)/(Attempts + Sacks)[/math]
Relative ANY/A simply compares a quarterback’s ANY/A average to league average, a necessary element when comparing quarterbacks across eras. In the graph below, the size of the bubble corresponds to how many attempts Bradshaw had in each season, while the Y-Axis shows Bradshaw’s Relative ANY/A (by definition, 0 is equal to league average). The graph shows a clear story: for the first five years of his career, Bradshaw was a below-average quarterback, but over the rest of his career, he was one of the best in football. His best year came in 1978 when Bradshaw finished with a RANY/A of +2.0, which was the third best mark in football (only a hair behind Roger Staubach and Dan Fouts). Those stats, combined with a 14-2 record, led to Bradshaw being named the AP’s MVP that season.
Now, let’s do the same for Warner. We can see that he had several years with low pass attempts numbers — mainly 1998, 2003, and 2006 — but what really stands out is the odd shape of his career. His best RANY/A years were his first, as the chief executive officer of the Greatest Show on Turf. But from 2002 to 2006, Warner never started more than 10 games in a season (due to injuries, Eli Manning, and Matt Leinart), although his RANY/A was good the last three of those seasons. Later on, like Bradshaw, Warner retired while still being a solidly above-average passer.
The reason to use RANY/A and not ANY/A is that by adjusting for era, we can compare apples to apples. So let’s combine our two bubble graphs and look at Warner’s and Bradshaw’s careers together, with the X-Axis now showing “year of career” instead of simply year.
Early on, Warner was the much better passer, but that edge doesn’t hold up for very long. Once we move past Warner’s scorched-earth GSOT days — where, admittedly, Bradshaw simply can’t compare — the old Steelers quarterback holds the edge for the remainder of their two careers.
After the era adjustment, Warner still holds an edge due to his run from 1999-2001 but there are other adjustments to be made. Consider that Warner played nearly his entire career in the NFC West when that division was one of the worst in football. As we’ll soon see, a strength-of-schedule adjustment will narrow the gap between these two quarterbacks.
But first, I’m going to make another adjustment that will put Bradshaw in a more positive light. I’m going to exclude his horrendous rookie season, which drags down his career averages considerably. Is that fair? I leave that up to the reader to decide. Bradshaw was the number one overall pick in the draft, while Warner went undrafted in 1994, was cut from Packers camp that same year, played in the Arena Football League for four seasons, and made it on to the 1998 Rams as the third stringer. In that context, I think it’s fair to give Bradshaw a pass for miserable stats at age 22 when he was handed the starting job even though he clearly wasn’t ready to play, since Warner’s career stats don’t reflect his level of play at a young age.Now, on to the strength of schedule adjustment. We can compare schedules, but the easier method is actually to combine the era and schedule adjustment in one step by simply looking at the average defense each passer faced. In Warner’s 12 seasons, his average opponent (weighted for the number of pass attempts thrown by Warner) allowed 6.54 AY/A1, while Warner averaged 7.55 AY/A; this means Warner was 1.02 AY/A better than average over the course of his career. For Bradshaw, he averaged 6.06 AY/A in his final twelve seasons while facing opponents that allowed, on average, 5.43 AY/A; this gives Bradshaw a grade of +0.64 AY/A. (For those curious, the “league average” AY/A during the Bradshaw years (weighted by the number of attempts he threw in each season) was 5.37, meaning Bradshaw faced a slightly easier than average schedule; for Warner, it was 6.26, meaning he faced a much easier schedule.)
So after adjusting for strength of schedule and era (and removing Bradshaw’s rookie year), we get a much different picture. Remember, the raw stats show that Warner had a career 7.55 AY/A average, while Bradshaw was at just 5.83, which results in a large difference of 1.72 Adjusted Yards per Attempt. But after these adjustments, Warner comes out as just 0.38 AY/A ahead of Bradshaw. We can quantify exactly how much of that gap was closed by each adjustment: most of it comes from the era adjustment (+0.89), with the rest split between eliminating 1970 from Bradshaw’s line (+0.22), and the SOS adjustment (+0.23).
But we’re not done yet. We have acknowledged that Warner played in a much more pass-friendly environment, that he wasn’t put on the field as an unprepared 22-year-old, and that he had an easier schedule. But Warner also benefited from playing the majority of his games in a dome or in Arizona’s retractable-roof facility. I looked at Warner’s and Bradshaw’s statistics in dome games, in outdoor games, and in games with half-domes or retractable roofs (for Bradshaw, this consists of two games in Dallas; for Warner, it’s his games in University of Phoenix Stadium, and one game each in Dallas and Houston). Here’s how to read the table below. For Warner, he had 1,246 regular season pass attempts in domes. His Expected AY/A based on the defenses he faced was 6.84 (this is the combined SOS/era adjustment), while he actually averaged 8.98 AY/A. Therefore, in dome games, Warner averaged 2.14 AY/A over expectation.
Warner was much, much better in dome games than he was outside, and that must be part of the discussion when comparing him to a player like Bradshaw. Now remember — and this is something the Peyton Manning detractors often forget — the numbers here overstate the advantage Warner gained from playing in a dome. That’s because nearly all of his outdoor games were road games (and most of his dome games were home games), where we would expect his AY/A be lower. And we must remember that most of Warner’s dome games came when he had the best supporting cast of his career, so the high average in dome games is largely a result of having Marshall Faulk, Orlando Pace, and the rest of the one of the most talent-rich offenses ever. So it would be wrong to look at this table and say playing in a dome made Warner. For Bradshaw, his performance in dome games all came on the road, of course, with the vast majority coming in the Astrodome against the Oilers. But in outdoors games, he provided +0.85 AY/A better than expectation.
So where does this leave us? It’s up to the reader to decide which is more impressive: being +1.02 AY/A better than average with 47% of your passes coming outdoors, or +0.64 AY/A better than average with 90% of your passes coming outdoors. Certainly some penalty must be given to Warner since his numbers came in friendly environments and he was far from dominant in outdoor games. One must also remember that Bradshaw wasn’t just playing outdoors: a fair number of his games were in cities like Pittsburgh, Cleveland, and Cincinnati in November and December, which probably provided worse conditions than the average Warner outdoors game, too.
For me, the journey here is more important than the destination. On the surface, Warner’s career stats look significantly better than Bradshaw’s. But the entire point of the analytics movement is to put statistics in proper perspective. After adjusting for era, strength of schedule, different career arcs, and weather, we can see that their numbers look very similar. And honestly, that’s how it should be, since we’re talking about two Hall of Fame caliber quarterbacks. Much of football analytics can be shortened to “putting everyone on the same playing field.” The idea that analytics can’t solve everything is true, but this is still a much better method of comparing quarterbacks than simply throwing out the stats because you can’t judge quarterbacks from different eras.
We must also remember that determining which quarterback was better involves a different level of discussion than what is in today’s post, where I simply compared their statistics in the proper light. In a broader debate, you’d also want to include playoff performances. In this regard, both were outstanding: I ranked Bradshaw number two and Warner number four on my Super Bowl era list of best playoff passers. You’d also want to take into account the supporting cast of both players, although again, both have similar arguments here. Bradshaw had Hall of Fame caliber receivers in Lynn Swann and John Stallworth; Warner had that in St. Louis (Isaac Bruce and Torry Holt) and in Arizona (Anquan Boldin and Larry Fitzgerald). I encourage you to discuss in the comments which quarterback you think was better, along with your thoughts on this method of analysis.
- Note that I am now using Adjusted Yards per Attempt instead of ANY/A, as we don’t have reliable game-by-game sack data going back to 1970. This is not a big issue, in my view, since the players had similar (era-adjusted) sack numbers. To the extent it is an issue, it’s to Bradshaw’s detriment, as he was slightly better at avoiding sacks. [↩]