Army Football Preview: Preseason Power Rankings

ESPN updated its Football Power Index (FPI) at the end of July, and the revised rankings hit Army right in the mouth.  The Black Knights are now the lowest-ranked of the service academies with an FPI of -9.7.  That puts them 103rd out of 130 total FBS teams.  ESPN projects that they would lose to an “average” team on a neutral field by something like 10 points.  The Four Letter Network projects that would-be “average” team as Fresno State out of the Mountain West, with an FPI of +0.3 points above the mean.
So Army’s doomed, right?  I mean, why even play the games when we have preseason computer projections?
Preseason Rankings: Army vs. the 2018 Schedule
For better or worse, this is about what we’ve seen the last several years as far as preseason power rankings are concerned.  I’m not sure what exactly has changed within the FPI model to drop the Black Knights a whole point in the last month—and five spots against the rest of the nation—but ESPN is heavily invested in both the Power 5 bias and its own proprietary quarterback ranking system (QBR).  FPI’s own explainer page notes the network’s emphasis on returning quarterback production, so the drop may very well coincide with Army’s naming a new starting QB who’s not Ahmad Bradshaw.  But maybe not.  ESPN also claims that a returning quarterback ought to be worth something like +3.3 points overall.  Army’s lost just a single point, so it might be something else entirely.
Honestly, it’s not that hard to see why the model is negative about the Black Knights.  First, because FPI is not so much trying to rank teams as it is trying to set lines for potential matchups and—crucially—to predict season outcomes.  This is why the Power 5 bias has become so pronounced in FPI modeling.  If it’s functionally impossible for a Group of 5 school to make the College Football Playoff, then the model cannot rank G5 schools at the same level as schools from the P5.  That would incorrectly predict outcomes, and self-fulfilling prophecies must fulfill themselves by definition.  This is accomplished primarily through strength-of-schedule differences between the P5 and G5, and that hurts Army because Army has by no means played a murderer’s row these past few years.  Not compared to, say, UCF or Duke at any rate.
Beyond that, FPI is grading predicted offense, predicted defense, and predicted special teams using a regression that goes back through four years of data.  It also factors in recruiting, returning starters (especially QBs), and coaching tenure.  All of which is essentially negative for Army save coaching.  In particular, the four year trend analysis hurts because the program only started seeing improved results in 2016, and even then, special teams still hasn’t quite turned the corner.  Moreover, it’s not like Army has recruited a ton of 3-, 4-, and 5-star recruits.  Instead, the staff is bringing in a few 3-star guys alongside a bunch of players that other schools essentially can’t recruit because of scholarship limits, and West Point’s coaching staff is hoping it candevelop starters through the system.  That’s worked reasonably well, but it’s not the kind of thing we can easily model in a simulation.
SB*Nation’s S&P+ system is similar, but it replaces strength of schedule calculations with various performance-based success rates against the mean.  Essentially, S&P+ uses actual plays as if they were Monte Carlo simulation runs, adjusted for down and distance.  It’s is a fascinating approach mathematically, but it’s intentionally weighted towards explosiveness, which penalizes triple-option teams rather spectacularly.  The model’s basic design skews to make Army’s greatest strength look like a weakness.  Still, it’s probably a better overall model than is FPI.
S&P+ puts Army 92nd overall with a composite score of -5.3 points against their theoretically “average” team, Northern Illinois out of the MAC.  But that by itself is a little misleading.  S&P+ is generally positive about Army’s returning players, but it ranks the Black Knights’ recruiting class 110th, and its use a 5-year rolling average to model past performance.  Considering how the last five years have gone, I don’t think that’s overly predictive.  Army is at -5.3 while Navy is at -3.3, and the difference is almost entirely down to total team performance between the years 2013 and 2015.  Yeah, Navy was better in those years.  No, you probably shouldn’t let that influence your sports betting going forward.  
Nevertheless, if all you care about is upper-tier recruiting rankings and returning starters, the argument against Army is pretty clear.  
Let’s start by looking at Army’s new starting quarterback, cow Kelvin Hopkins.  At 5’10”, 205 lbs, he doesn’t exactly fit the prototypical D1 ideal.  In fact, those measurements are way more common among D1 running backs, and yeah, that’s not surprising given what Army’s trying to do with its scheme.  But that is exactly the kind of thing that fucks up these compute algorithms.  Moreover, though Hopkins saw spot duty last season as Army’s “throwing” quarterback, he’s not exactly a known quantity.  Despite leading a dynamic, game-tying drive against Temple, his raw numbers were not great in limited action.  He went just 6/18 passing (33.3%) on the season for a whopping 76 yards with 1 admittedly crucial touchdown, 1 interception, and a sack.  Taken out of context, that sounds terrible.  Hopkins also carried 7 times for 40 yards (5.7 yards/carry) with a long of 22 yards but no touchdowns.  That’s not bad, but it’s fully 2 yards/carry below what Bradshaw put up, and this season, our new QB will be running behind an offensive line with just one returning starter.
You can see where this doesn’t look so hot on paper.
But analyzing quarterback play in a vacuum is the stupidest, most reductive kind of football analysis.  Let’s dig a little deeper.
Alas, it’s impossible to adjust for is sample size.  Seven carries is not a lot, and even eighteen passing attempts isn’t overly representative when we consider that thirteen of those came in a single, manic last-minute drive. Even those limited rushing stats include negative yardage from the 5-yard sack Hopkins took in the final minute against Temple.  Take out that sack, and even without adding back the lost yardage, he’s back up to almost 6.5 yards/carry.  Add in the lost yardage, too, and we’re back up to nearly 7.5 yards/carry, which is right in line with what we saw from the quarterback position all last season.  Granted, that’s not much to go on, but it is at least indicative of functional ability.  There’s not necessarily a reason to expect a drop off just because we have a new QB.  Which is good as far as it goes.  It still tells us nothing about Hopkins’ decision making within the triple-option offense.


I’ll close this out by talking a bit about recruiting.  Army’s recruiting might not look “good” when compared to P5 schools that field multiple 3- and 4-star athletes in the underclasses, but nevertheless, it has improved dramatically in the Monken era.  Standout players like LBs Andrew King and Jeremy Timpf developed within the system, but both were essentially unheralded recruits who succeeded through the force of their personal commitment to excellence.  Succeeding classes have been physically larger than were the Classes of 2016 and 2017, and what’s more, as the Black Knights’ fortunes have improved, so too has the competitiveness of the kids choosing to play at West Point.  The Class of 2019 brought in a lot of players but not a superabundance of guys with competing D1 offers.  Since then, though, successive classes have gotten both bigger and faster.  Today, most of the newest recruits are choosing Army over other D1 schools.  These are still 2- and 3-star players, but at least in theory, they ought to have a higher ceiling than the guys who played at Army as recently as 2016.  After all, you can help a skinny kid get stronger, but you can’t teach somebody to be tall.
We’re only now starting to see how that will shake itself out.  This year’s offensive and defensive lines, with multiple new starters on both sides of the ball, ought to tell us a lot about how well Army has succeeded in recruiting and developing its players.  That’s not necessarily easy to quantify in terms of preseason rankings, but all things considered, it shouldn’t necessarily be cause for alarm, either.


* * *

Reading the Coverage

Yards & Stripes: Navy Head Coach Ken Niumatalolo, Army Football Preview and Air Force Unveils Alternate Uniforms
The episode features a very good interview with Navy Coach Ken Niumatalolo, especially in reference to Navy's early game at Hawaii (if you care about that sort of thing).  Later, there's some talk about Army's evolving offense that essentially confirms a lot of what I speculated about last week.  It's well worth a listen.



Scrimmage blog: Plenty to work on (Army Football Insider)
Monken was asked what he was most happy about from the scrimmage. His answer: “The effort of the scouts. I was most pleased with their enthusiasm. We hit some pass plays. We protected the quarterback. We had some good runs. They were some guys that stood out when you watched them. There were some good things.”

Sounds to me like Army's got a bunch of really good young players on their scout team, and that some of those guys are already pushing for playing time.  Though Coach Monken wasn't super-pleased with the results, that's not exactly a disaster long-term.


Comments

  1. In regards to S&P+, Bill Connelly admits that it underrated triple option teams. I think you can expect Army to outplay those rankings (and same goes with Navy and Air Force).

    ReplyDelete
    Replies
    1. I saw that & probably should have linked to it. Really, tho, it’s only important in terms of betting lines. The current line is Duke (-13.5), basically the FPI line.

      Not to spoil next week’s preview, but I’d definitely take Army & the points.

      Delete
    2. Heck yeah. I’d take the points all day long. I’m not sure Army will win, but I expect a close one.

      Delete

Post a Comment

Popular posts from this blog

The Mystery of Malvern Manor

D&D: Guinea Pigs & Were-Guinea Pigs