But even if we had perfectly estimated the fundamentals, Boswell has been outperforming them in a way that is unsustainable. Heading into this week, he had made 29 of his 30 kicks, which at 97% was the second-highest conversion rate in the league (behind Jake Bates, who had attempted half as many kicks). It was also significantly above Boswell's career rate (87% coming into the year) and league average (85%). My model doesn't include an adjustment for kicker accuracy because outside of Justin Tucker all NFL kickers are incredibly tightly clustered.
(In the article above, Boswell was the 3rd-best kicker in the league in terms of accuracy over wind- and distance-adjusted expectation-- in large part because of his ability to kick so well in a tough stadium like Hines Field! But the expected difference between Boswell and a league-average kicker was just 3%, and the error bars are huge. For most non-Tucker kickers, the difference between "good" and "bad" is about 3 points in expectation over the course of a season. Philosophically, I tend to be extremely wary of overtuning / overfitting my models, and when I've tested in the past, the predictive gains were far too small, IMO, to justify the added complexity. I would have felt justified in adding a Tucker-specific boost just because he was so far ahead of the rest of the pack, but doing so would be fairly pointless in a streaming kicker model because Tucker is never available to stream. And now it looks like time has caught up to him, anyway.)
Had Boswell converted field goals at his career rate (88%, which is 3% over league average), he'd have entered this week with 79.2 points from field goals instead of 87, a loss of about 8 points. That would drop him to 99 points in 10 games. He'd still be the #1 fantasy kicker, but it would leave him in a virtual tie with Seibert and Fairbairn rather than head-and-shoulders the top guy.
But the real driver behind Boswell's production has been an incredibly fluky scoring mix from the Steelers. Heading into the week, 58% of Pittsburgh's scoring plays had been field goals rather than touchdowns. League average is 41.3%. 58.8% of Pittsburgh's scoring
opportunities (touchdowns or field goal
attempts) were field goal attempts, which ranks second behind the Cowboys (who are a bit skewed given their willingness to let Aubrey kick from insane distances in non-end-of-half situations-- though Pittsburgh has also been doing this with Boswell this year). League average is 45.3%. As a result, Boswell had scored 45.9% of Pittsburgh's points coming into this week-- league average is 32.7%. All of these percentages are... crazy and wildly unsustainable. And they all tend to regress.
It's true that not every kicker should have the same expectation for these values. Historically, kickers on low-scoring teams account for a marginally higher percentage of points. Kickers on teams with conservative coaches (measured by willingness to go on 4th down) account for a marginally higher percentage of points. These differences are fairly marginal, though-- historically kickers score around 31% of points (we're up slightly this year) and these factors might push a kicker as high as 33% or as low as 29% in expectation.
Are defense-first teams significantly more likely to settle for field goals? Here's the % of points scored by kickers on some notable teams of the past.
2000 Ravens -- 40.5%
2002 Buccaneers -- 36.8%
2015 Broncos -- 35.2%
2006 Ravens -- 34.3%
2013 Seahawks -- 34.1%
2008 Steelers -- 33.7%
2006 Bears -- 33.5%
Total -- 35.3%
Yeah, that's slightly above league average, but we're talking about the best defenses of the past 25 years here, and Pittsburgh isn't on that level. (They're certainly nowhere near the 2000 Ravens and 2002 Bucs, arguably two of the top five defenses of all time.) And again, even if they *were* on this level, if Boswell had scored 35.3% of Pittsburgh's points through ten games, he'd have 82.25 points, which would rank him 7th overall, 6th in points per game, and tied with Cameron Dicker.
Do Mike Tomlin / Chris Boswell teams get a disproportionate share of their scoring from the kicker? Here's every Steelers season since they signed Boswell.
2015 -- 32.4%
2016 -- 27.3%
2017 -- 35.0%
2018 -- 21.4% (first season without Le'Veon Bell)
2019 -- 39.8% (first season without Antonio Brown)
2020 -- 26.4%
2021 -- 39.4%
2022 -- 41.6% (first year without Ben Roethlisberger)
2023 -- 37.5%
Total -- 32.6%
One could point to the 41.6% mark in 2022, but that was with scoring cratering under rookie Kenny Pickett. Even with that insane rate, Pittsburgh only ranked 9th in kicker scoring-- and most of that was not because of Boswell. (Boswell scored 6.5 points per game in his 12 games, while Matthew Wright and Nick Sciba averaged 10.0 points per game in the five games he missed.)
In fact, there's a strong correlation between Pittsburgh's scoring offense and the kicker's share (this is a common trend that I mentioned above, but it's especially pronounced in Pittsburgh during the Boswell era). Pittsburgh averaged at least 24.9 points per game in every season from 2015-2018 and again in 2020; during these years, kickers scored 28.4% of all points. On the other end, Pittsburgh averaged 20.2 points per less in the other four years, and kickers accounted for 39.5% of all of their points during those years (which is a truly insane percentage-- and still more than 6% lower than Boswell has this year).
Overall, the correlation between points per game and kicker share is quite robust in the Tomlin/Boswell era-- 0.844, which means "how good the offense is" explains more than 70% of the variation in "what share of the scoring goes to kickers". But Pittsburgh's offense hasn't been terrible this year (it's been downright terrific compared to e.g. the Kenny Pickett era!) They were averaging 23.3 points per game heading into Week 12; based on the 2015-2023 data, that should result in an "expected kicker share" of 32.5% (basically... right on the long-run league averages). If Boswell was getting that share, he'd have 75.7 points and rank 11th among kickers.
So Boswell to this point had been something of a perfect storm of factors-- everyone underrated Pittsburgh, Boswell was putting up a career-best year, and the Steelers had a truly demented rate of scoring drives resulting in field goals. All of which mattered a great deal in terms of how much Boswell has been worth through 11 weeks. If you had him, congratulations, he's made a massive difference for your fantasy squad. Rent-a-Kicker's recommendations regarding him have been wrong. Flat wrong. Period, full stop.
But none of those factors are *predictively* useful. Setting aside the question of whether I've been wrong about Boswell to this point (I have been, full stop), what matters now is how likely it is that I'm wrong about Boswell
going forward. And... obviously I don't think I am. I've tested in the past whether these factors are stable and predictive, and they're not. I've tried incorporating "points scored to date" in my model before and every time it added complexity while producing worse recommendations.
I also think back to previous kickers the model has been most consistently wrong about. In 2020, I even
devoted one of my other columns, Regression Alert, to analyzing all of my failed Rent-a-Kicker recommendations of Jason Myers (who at the time played on an offense that had scored 23 touchdowns against just 2 field goals-- a brutal 8% FG ratio).
At the time I wrote that, I had recommended Myers to kicker streamers in 4 of the 5 weeks he played. Despite this, he ranked 31st in points and 28th in points per game. Readers who followed my recommendations undoubtedly lost games as a result. That sucks and is something I take very seriously. But I wrote that when I'd investigated in the past, FG mix was not predictively useful, so despite the fact that I had been wrong and had cost my readers wins, I would continue recommending Myers going forward. (I also predicted that the kickers who were most hurt by FG mix would outscore the kickers who were most helped by FG mix to that point; they did.)
From the time I published, Seattle scored 21 FGs against 32 touchdowns, a 39.6% FG rate that was basically right on league average. Myers scored about 32% of Seattle's remaining points, a rate that was basically right on league average. Myers ranked 3rd in points and 7th in points per game. The fact that I *had been* wrong did not predict that I would *continue to be* wrong.
There have been other examples. I was consistently too low on Nick Folk to open 2021; that was entirely because I was (or rather Vegas was) underestimating the Mac Jones-led Patriots offense; that was corrected over the second half of the season. I was consistently too high on Jason Sanders over the first half of last year, again due to scoring mix-- he had 37 extra point attempts vs. just 9 field goal attempts through his Week 10 bye. That FG mix reverted just as strongly as Myers, and Sanders was the #2 fantasy kicker after his bye, one point behind first.
I'm not saying my process is perfect or above reproach. Remember-- the very best models are going to be wrong 45% of the time. Through 11 weeks, I was very wrong about Chris Boswell and people undoubtedly lost games as a result. Some of them will miss the fantasy playoffs as a result. Some of them will lose out on championships they otherwise would have won if not for my advice. That's something that I am constantly aware of.
All of this is just to explain what my process is and why it is the way it is. I don't think my model is above questioning or criticism. But in seven years of searching, I have been unable to find anything that made it perform better than it currently does.
