clock menu more-arrow no yes

Filed under:

Recruiting Rankings and Player Impact, Part 2

New, 30 comments

Recruiting continues with the commitments Ryan Otton and TJ Hall, so welcome to the team gentlemen and Go Huskies! One is a 4-star legacy recruit, top 5 at his position; the other is a 3-star most of us had never heard of before he committed, not even in the top 40 as an “athlete.” Which one is most likely to have the better career? Otton certainly seems like the better bet, but that begs a couple of questions:

  1. What does “success” really mean when it comes to recruits.
  2. What can or should we reasonably expect from the players on the team?
  3. Does UW really do a better job than other teams in terms of either identifying underrated gems and/or of developing players to outperform their rankings?

So how do we figure out the answers? As discussed in Part 1, this study sprang out of some thoughts I had in the Coach Gregory’s LB Legacy article, revisited and refined, and most importantly broadened to include comparison groups. Namely, the two teams the UW sees as their most important rivals for conference supremacy, and both of which are generally out-recruiting UW: USC and Oregon. To beat those teams, our players need to perform better than their players on the field. How do we know if we’re doing well? The study looked at 261 total LB and DL recruits over an 11-year period between the three schools, including every player from 5-stars to walk-ons who suited up at those positions from 2010-2020.

As discussed in Part 1, not every one of those players got on the field - in fact, only 215 of 261 made it to an active roster (on the team, not redshirting) during the selected time period. When we talk about how a player did, they needed to have had a chance to do something. They might not have accomplished anything, but they needed to at least have the opportunity.

Measuring IMPACT

To measure how well players accomplished things on the field, I developed IMPACT as a composite stat earlier this year to look at player performance on the field. It combined routine tackles (including solos and assists) with a variety of big plays, using the following formula (BP = TFL + Sack + PD + FF + FR + INTx2 + TDx3). This started off as a study of LBs at UW, so it’s by no means a universal stat for all positions. I could’ve included DBs in the study too, but given how much crossover I found between LBs and DLs, that seemed like the best, most cohesive spot to start.

A few notes about IMPACT.

  • Sacks are counted in addition to tackles for loss even though sacks are a subset of TFLs, as the average yardage loss on a sack is substantially greater than on an average TFL.
  • Interceptions have the same net BP value as a forced fumble + a fumble recovery, as in each case the player is causing and then recovering the turnover.
  • These were combined to give a total IMPACT score, based on total tackles + 3x Big Plays. This intentionally weights big plays as being more important than tackles, which is probably self-evident but it also helps sort out players who are just on the field managing the basic necessities (someone has to make tackles at some point, so on almost every defensive play someone will get one) vs. the players who are making key game-wrecking plays that tilt the field. The best players, of course, will do both.
  • Impact could be measured in the aggregate, rewarding consistency and longevity, or as a factor of IMPACT/Game. Given the variations in how many seasons players might play (especially for the most successful players, whose very success might induce them to leave school early), IMPACT/Game is a useful tool. Still, there is no ability like availability, and compiling an extensive track record of success definitely counts for something too.

NOTE: When I say “per game,” I mean divided by the number of games that a given player was eligible to play (i.e., on the roster, in school, not redshirting). If you’re hurt and you’re redshirting, those don’t count as games to you. If you’re on the active roster and you’re not redshirting but you’re not playing for any reason (academic, disciplinary, injury, just not good enough to beat out the players in front of you), those still count as games when you’re not contributing on the field.

Means and Medians

These basic statistical principles are key to evaluating how a typical player does in a given situation. A mean is a statistical average: you add up the value of each thing in your sample and divide them by the number in your sample. If I have three boxes, one has 8 Husky hats, another with 1, and another with 0, that’s 9 hats, 3 boxes, giving us a mean of 3.

That’s obviously not a very representative sample, though, and that’s where the median value is helpful, as it reduces the impact of statistical outliers by simply defining the point where half of the values are above and half are below it without caring what those values are. With our Husky hat example, our values are 0, 1, and 8, so the median value is 1.

Yo Dawg, we heard you liked Husky hats, so we put a Husky hat on your Husky hat!

Put in football terms, having one Shaq Thompson, Leonard Williams, or Kayvon Thibodeaux makes your mean value look sexy, but if you’ve only got one of them and a dozen Jake Wambaughs (thanks for your service, Jake, we love you), that’s not so good for the total picture of the team. More to the point, it makes it much harder to pinpoint whether players are succeeding, and how you can measure or evaluate the success of players relative to expectations. One great player is just one great player, and allowing their stardom to make your overall average look better can distort what’s really happening with the rest of the gang.

Big Picture Numbers

The following tables illustrate the kind of per-game impact the 215 active players in this study had on the field, divided by recruiting ranking. Table 1 shows the mean value for players at each recruiting rank, Table 2 their median value. The four columns on the left show the mean or median values for only the 1st, 2nd, 3rd, or 4th seasons of players in those categories; usually that number goes up but not always. The four columns on the right indicate their cumulative IMPACT/Game for their entire career, which illustrates the progression of players over time.

Table 1: Player Impact/Game (Mean)

Recruit Rankings 1st Year I/G 2nd Year I/G 3rd Year I/G 4th Year I/G 1-year career I/G 2-year career I/G 3-year career I/G 4-year career I/G
Recruit Rankings 1st Year I/G 2nd Year I/G 3rd Year I/G 4th Year I/G 1-year career I/G 2-year career I/G 3-year career I/G 4-year career I/G
All Players 1.58 2.87 3.83 5.15 1.58 2.23 2.76 3.36
Walk-Ons/2* 0 0.19 0.16 0.28 0.22 0.66 0.8 0.99
Low 3* 0.87 0.87 1.11 2.39 2.1 2.74 2.8 3.56
Mid 3* 0.63 0.84 1.81 2.34 1.26 1.69 2.58 3.51
High 3* 0.66 1.58 2 2.48 1.46 2.43 3 3.54
Low 4* 0.45 1.07 2.07 2.49 1.29 2.29 3 3.74
Mid 4* 1.44 1.86 2.6 3.24 2.16 2.59 3.24 3.72
High 4* 1.67 2.24 2.49 3.18 3 3.37 3.63 4.12
5* 3.3 3.92 4.31 4.71 3.81 4.26 4.98 5.21
Mean season length: 13.2 games

Table 2: Player Impact/Game (Median)

Ranking 1st Year I/G 2nd Year I/G 3rd Year I/G 4th Year I/G 1-year career I/G 2-year career I/G 3-year career I/G 4-year career I/G
Ranking 1st Year I/G 2nd Year I/G 3rd Year I/G 4th Year I/G 1-year career I/G 2-year career I/G 3-year career I/G 4-year career I/G
All Players 0.53 1.78 2.80 3.86 0.53 1.16 1.70 2.24
2*/WO 0.00 0.34 0.11 0.64 0.00 0.17 0.15 0.27
low 3* 0.36 0.31 0.86 1.29 0.36 0.34 0.51 0.71
med 3* 0.51 1.34 3.41 3.79 0.51 0.93 1.75 2.26
high 3* 0.76 2.31 2.80 4.39 0.76 1.54 1.96 2.57
low 4* 0.45 1.65 3.48 3.71 0.45 1.05 1.86 2.32
med 4* 1.67 2.12 4.20 5.34 1.67 1.90 2.66 3.33
high 4* 0.72 3.56 3.37 5.32 0.72 2.14 2.55 3.24
5* 3.03 4.55 4.73 5.23 3.03 3.79 4.10 4.39
Mean Season Length: 13.2 games


The above tables give us a set of general benchmarks, based on a fairly substantial data set for three relatively comparable schools. Without knowing much more at this point, you can make some observations like:

  • Assuming they stay in school and play a full 4 seasons, the career mean values for low, medium, and high 3-star players are basically identical. Based on median values, however, low 3-stars produce less than a third the career value of a medium 3-star (and even less compared to high 3-stars, who have a noticeable gap above mediums). This suggests that there are some star success stories in the low 3-stars inflating the mean but that overall low-3* players are much more likely to produce like walk-ons than like scholarship players.
  • Mean values also suggest nearly identical 4-year career performance for low and medium 4-stars, with a significant gap between them and high-4* players. Median values tell a different story, however, with medium-4* players far outclassing low-4* and, surprisingly, also outproducing high-4* players. One guess at the cause of this might be players who were good enough to keep getting some playing time but not good enough to go pro early, resulting in the high-achievers in 4* land leaving the washouts who never developed behind to drag down the category overall. The data don’t seem to support this, however, as we’ll see in future installments.

While numbers they provide a notional relative scale, they’re also fairly abstract, so in future installments we’ll be “humanizing” these numbers by breaking down recruiting and performance for UW, USC, and UO’s LB and DL crews to compare them, and see what patterns, if any, emerge from the data to answer the questions at the beginning of this article. First up in Part 3, the Hometown Huskies!

Go Huskies!