≡ Menu

Adjusted Interceptions: Career Totals

Yesterday, I did a fairly simple analysis to compare interception numbers across eras. Because I covered the methodology in the previous post, I am not going to regurgitate that information here. Instead, I want to just get right into it. When I did the career adjusted sacks post, I went step-by-step in the same manner I did in the single season article. This time, however, I think we can skip past all that and look at the end results.

Career Adjusted Interception Totals

The first table is sorted by the last column, but you can re-sort by any header you like. Using Rod Woodson as an example, here’s how you read the table: Woodson intercepted 71 passes against 8401 attempts at a 0.85% rate. His passing environment modifier (Mod) is worth 142%, and the softened version of that (Soft) is worth 121%. Taking the average of his actual interceptions and interceptions per 500 attempts in order to account for volume gives us the Mid adjustment, which is 98% in Woodson’s case. Applying my homebrewed league strength multiplier (LSM) gives him a 99% adjustment.

If we multiply Woodson’s 71 interceptions by Mod, Mid, and LSM, we get a whopping 97.7 adjusted interceptions for his career. If we dampen it by multiplying those 71 picks by Soft, Mid, and LSM, Woodson’s career adjusted interceptions come to 83.2, good for the highest mark ever. [1]For the ModTot, that’s 71 * 142% * 98% * 99%. For the SoftTot, that’s 71 * 121% * 98% * 99%.

Using the actual historical average as a baseline appears to be a bit much, going by the numbers it produces. I think having Charles Woodson, Ed Reed, Rod Woodson, a serial rapist, and Aeneas Williams as the top five (by ModTot) is fine; giving Chuck credit for 101 interceptions is a bit much for me. Moving the all time leader in picks, Paul Krause, down to ninth also feels a tad harsh as well. Sure, I think he tends to be overrated by people who look at one number and base their entire evaluation on that single data point, but I also think having such a commanding lead over any modern player should count for something. For this reason, I think the SoftTot column produces results with greater face validity.

The last column gives us a top ten of seven Hall of Famers, one senior candidate who will likely get the necessary votes soon, possibly the best safety of the 1990s who would be in Canton already if he played for Dallas or San Francisco, and a a vile monster who was good at picking off passes and not really much else.

Let’s Be Reasonable

The wacky looking career totals form the table above convinced me to try using a new baseline. I decided to use the last 40 years of football, which incorporates nearly the entire period of open offense football. [2]I refer to football in the wake of the Mel Blount Rule and rules enabling offensive linemen to extend their hands to block in 1978, as well as the subsequent offensive revolution heralded by the … Continue reading When I looked at that timeframe, the historical baseline moved from 4.80% to 3.16%. because of that, I have dubbed the new baseline the Austin Percentage. Having a lower baseline means that fewer players will see their totals go up, and only the most recent players will their totals go up significantly.

The table below is sorted by the last column, but you can sort by any of the headings. Using Krause as our example, read the table thus: Krause picked off 81 passes against 5623 attempts at a 1.44% rate. His volume adjustment is worth 117%, and his league strength multiplier is worth 88%. His Austin figure is 60%, which comes to 80% when the effect is halved. [3]Recall from the first table that his Mod and Soft were 91% and 96% because of the highest baseline. If we apply the Mid, LSM, and Austin modifiers to Krause’s 81 actual interceptions, his total plummets to 49.9, which ranks tenth on the career list. If we replace the Austin modifier with the softened version, Krause’s number falls to just 66.4, which allows him to maintain his place atop the interception mountain. [4]To arrive at the numbers in the Austin column, we use: 81 * 117% * 88% * 60%. To find the results for the HalfTot column, we use: 81 * 117% * 88% * 80%. These figures are rounded and will produce … Continue reading

If you earnestly believe older players relied too much on archaic passing to glean their big interception totals, the Austin column might be for you. Before we find Krause at number ten, only Rod Woodson and Eugene Robinson had any action prior to 1990. Recent ball hawk Richard Sherman is in a fourteen-way tie for 104th place in career interceptions, with 37. However, when Austin 3.16 comes around, Sherman jumps to 18th, which does feel more appropriate for one of the premier turnover artists of recent vintage. In fact, his 8.4 interception boost is the highest number of any player, just beating out the bonuses of 8.3 and 8.1 for fellow playmakers Xavien Howard and Marcus Peters. Wandering mercenary Aqib Talib finds himself pretty high on the career list when looking at the Austin total.

While some recent players saw modest gains, older players saw their totals fall off a cliff with the lower baseline. Emlen Tunnell, a real life hero who picked off 79 passes—but did most of his damage in the 1950s—suffers a reduction of 43.6 from his total. He goes from ranking second on the official list to 54th on the Austin list. That seems a little steep, even to a noted old school player hater like I am. Night Train Lane and Johnny Robinson join Krause and Tunnell as the only other players to lose at least 30 from their totals. Turn-of-the-century players like Sam Madison and Patrick Surtain see almost no change in their career numbers.

I think the last column makes the most sense at first glace. Tunnell, Robinson, and Jim Norton all lose more than 20 from their real numbers, and no one gains more than 3.6. Krause loses 14.6, but because Tunnell lost 22 and his lead over anyone else was huge, he remains in first place. Rod Woodson loses 4.7 from his total, while Charles Woodson and Ed Reed each lose about half a pick, resulting in the three ending pretty clustered, and all close to Krause at the top. While Tunnell has a large reduction, his actual number of interceptions was so high to begin with that he still ranks sixth here.

I am often interested to see where Ken Riley and Dave Brown will fall, relative to one another. Riley has 65 interceptions to Brown’s 62. The Austin adjustment puts Brown ahead, while my preferred adjustment leaves the Bengals legend with a 54.6 to 53.4 lead. Riley never made a Pro Bowl, but he earned first team all pro honors once and second team honors twice. Brown made one Pro Bowl and one all pro second team. Given how close together these two are in terms of actual production, the gap in their public perception is pretty interesting to me. When you consider the fact that Brown was his team’s top corner, while Lemar Parrish was the top corner in Cincinnati until 1977, the issue is further muddled.

I will leave further commentary to the FP faithful, if any remain.

 

References

References
1 For the ModTot, that’s 71 * 142% * 98% * 99%. For the SoftTot, that’s 71 * 121% * 98% * 99%.
2 I refer to football in the wake of the Mel Blount Rule and rules enabling offensive linemen to extend their hands to block in 1978, as well as the subsequent offensive revolution heralded by the likes of Bill Walsh, Don Coryell, and Joe Gibbs.
3 Recall from the first table that his Mod and Soft were 91% and 96% because of the highest baseline.
4 To arrive at the numbers in the Austin column, we use: 81 * 117% * 88% * 60%. To find the results for the HalfTot column, we use: 81 * 117% * 88% * 80%. These figures are rounded and will produce slightly different results if you copy and paste to work with them yourself.
{ 2 comments }

Adjusted Interceptions: Single Seasons

I recently reopened a discussion about sacks Chase started years ago. Today, I’m going to rehash another topic our prolific host has covered time and again (and again): adjusting interceptions for era. Unlike sacks, which have official number dating back to 1982 and unofficial ones published as far back as 1960, interceptions have official records as far back as 1940. This gives us much more data to work with, but it also provides similar challenges that Sid Luckman presents when adjusting passing stats: the game is so different now from what it was in the 1940s that trying to compare the numbers side by side ends up killing newer players when adjusting for dropbacks or hurting older players when adjusting for passing environment. But getting it 85% right is better than not doing anything at all, so I’m going to do it anyway. As I did with the sack posts, I will go through my progressions of adjustments one step at a time, so that you can see how we arrived at the final numbers.

Normalizing for Volume

It stands to reason that intercepting ten passes against 300 attempts is more impressive than intercepting the same number of passes against 600 attempts—at least, as far as getting interceptions is at all impressive when divorced from other aspects of play. Because of this, it is necessary to put players on a more even playing field. Using Bill Dudley‘s frankly ridiculous 1946 season as an example, follow the table like so: Dudley, playing in 1946 for the Steelers, appeared in 11 games and snagged 10 interceptions against 162 opponent pass attempts. That comes to an outlandish 6.17% interception rate. If we adjust for volume by giving all players credit for their interception rate multiplied by 500 attempts, Dudley’s 1946 comes to 30.9 picks per 500 passes. If you look at that number and mutter “well, that’s just too high,” then we are in agreement. Thus, I took the average of their actual picks and their attempts per 500 passes to find the number in the last column, which I have labeled Mid. Using that adjustment instead, Dudley’s season was only good for 20.4 volume-adjusted interceptions.

If we stop here, it’s easy to see the glaring issue: the much higher interception rate in the days of yore leaves us with a list that doesn’t feature it’s first player after the year 2000 until the 210th spot. Even using the mid number, Ty Law‘s 2005 doesn’t show up until 115. Clearly, we need to keep going.

Incorporating League Environment

The next step is to incorporate the league average interception rate for each season. To do this, I used all seasons from 1940-2021 and found the three-year rolling average, with each given year in the middle (So 2017 would include the average of seasons 2016-2018). Then, I found three numbers: the cumulative interception rate from 1940-2021 (4.02%), the average of averages for each year in the sample (5.18%), and the median rate from the sample (5.21%). Then I took the average of those three numbers (4.80%) and used it as the historical baseline.

The next two tables use this step. The first of the two displays adjusted interception rates, while the latter of the two displays adjusted totals. Using Xavien Howard‘s 2020 as an example, read the table thus: Howard played 16 games and had 10 interceptions against 545 attempts, good for a 1.83% pick rate. The rolling average for 2020 is 2.28%, so Howard gets a boost of 210.5% (4.80/2.28) in the column labeled Mod. If you think that’s too higher, I included a softened version, which is the average of Mod and 100% (in this case, the Soft number is 155.2%). When using the Mod figure to adjust his interception rate, Howard gets credit for a rate of 3.86% (that’s 1.83% * 210.5%), the highest number on record. Using the softened version gives him 2.85% (1.83% * 155.2%), which ranks 18th.

This one is interesting to me, because the modified version seems too skewed in favor of modern players, while the softened version doesn’t feel harsh enough toward the old guys. We’ll go to the table below to see what that looks like in terms of interceptions rather just the more abstract percentages.

Incorporating League Environment (Again)

Let’s use J.C. Jackson as our example this time. In 2020, he played 16 games and hauled in 9 interceptions. We know his adjusted rates from the table above. Using the full modifier on his actual interceptions gives him 18.9 adjusted interceptions, while using the soft modifier gives him 14.0. Jackson is the rare current player who actually gets a boost from using per 500 attempt numbers, albeit a small one. Using the full modifier multiplied by his interceptions per 500 attempts (9.1 from the first table) leaves him with 19.2, while using the softened version gives him credit for 14.1. Note, I did not use the Mid figure from the first table, because too many columns makes these things unwieldy, in my opinion. Instead, I saved that for the last table.

Looking at the Mod and Soft multipliers applied to interceptions, without accounting for volume, just leaves us with a huge list of recent players. While I believe modern defenders to be both superior and in a more difficult position because of rules and schemes, I don’t think it makes sense to give them this much of a boost. Especially when the point of this whole exercise is not to measure the quality of a player, but rather use a variety of factors to more appropriately compare his interception totals to those of other defenders. One need only look at the career of Darrelle Revis to know that having a relatively low turnover total doesn’t preclude a player from greatness. And Ken Riley‘s career makes it evident that a player can find himself quite high on the career pick list without having been the best cornerback on his own team during his prime.

Putting it All Together

Below is the final table for today. Here, I have tried to strike a balance between adjusting for volume and adjusting for environment, but I kept battling with myself over whether I preferred full rate modifiers or soft ones. So I decided to just present both and let the reader decide. Using the controversial 2021 Trevon Diggs season, read the table thus: in 16 games, Diggs had 11 interceptions against 612 passes, good for a 1.80% rate. His Mid volume adjustment (from the first table) is worth 91%. That, combined with his 211.2%environmental modifier (Mod from the second table) gives him 21.1 adjusted interceptions in the Mod-Mid column. Using the Soft modifier instead gives him 15.5.

Instead of using the 4.80% historical baseline that I found, Chase most recently used 3.5%. Doing so doesn’t do much to the orders of the lists any, but it does have a significant impact on the totals by degree. So Diggs would still rank first in the Mod-Mid column and second in the Soft-Mid column, but he would have something closer to 15.4 and 12.7 as his adjusted interception total. While these numbers are more or less abstract and don’t really matter, I do think having the lower baseline Chase used produces results that look more realistic, even if the 3.5% figure was chosen at random (and I don’t know if it was or was not chosen at random). In fact, when I looked at career totals, I actually preferred to use an even lower baseline of 3.16%, which represents the last 40 years of football and covers basically the entire period of post-1978 rules changes that help permanently drop leaguewide interception rates below five percent.

When looking at the results above, the last column seems to produce the most even mix of old and new players. Oddly, however, I may prefer the Mod-Mid column when looking at career totals, which we will see later.  [1]How much later, I simply cannot say. Regardless, I think accounting for both volume and passing environment, in some form or fashion, helps put the numbers into more proper context. Even if it does take a little shine off my man Dick Lane.

 

References

References
1 How much later, I simply cannot say.
{ 2 comments }