Saturday, October 8, 2011

The Great Unknowns

Penn is the epicenter of this debate.

Senior point guard Zack Rosen is on pace to become the second most productive Ivy player of the past 10 years. Flanking him are fifth-year senior Tyler Bernardini, a streaky but at times lights out shooter who can add value at both ends of the court, and sophomore Miles Cartwright, an uber-athletic slasher who could vault to All-Ivy status if he could hold onto the ball a little better.

Even if you like the Quakers' returning depth at the guard position behind those three stars, there are two gaping holes at the four and five left by the graduation of Jack Eggleston and Conor Turley. Internal candidates do exist - the ever fragile Mike Howlett and sophomores Fran Dougherty and Cameron Gunter - but the true hope rests on the shoulders of heralded freshmen forwards Greg Louis and Henry Brooks.

How reasonable is that expectation, though? Brooks is the 19th most highly rated recruit of the past 10 years. Louis is 43rd most highly rated over the same span. Both are in the top 10 percent of the over 400 recruits from that timespan. But do the ratings even matter?


Harrison Schaen is one of those names that rings the faintest bell but not much more for many Ivy enthusiasts who don't root for Princeton. But he's the answer to the question that one might implicitly ask if pondering Ivy League basketball recruiting.

With all the hubbub about Harvard's 2011 duo of Wesley Saunders and Kenyatta Smith and its recent wave of 2012ers Mike Hall, Siyani Chambers, Evan Cummins, along with Penn's 2011 and 2012 hauls of Patrick Lucas-Perry, Henry Brooks and Jamal Lewis, it might be a little surprising to know that Schaen remains the highest rated recruit to join the Ivy League over the past 10 years.

That Schaen's name is merely the terminal point for a matter of trivia and not revered to the same level as recent stars Ibby Jaaber, Ryan Wittman and the like is the appropriately sobering starting point for a discussion of the link between recruiting ratings and future output.

The lofty ratings Schaen garnered from the various recruiting agencies had more to do with projection than performance. His freshman season with the Tigers was infused with the same projectability and promise: he hit a clutch bucket with 22 seconds left against Harvard at Jadwin to force overtime in a game Princeton would ultimately win. In terms of production, however, it would end up as his only field goal attempt in eight minutes on the floor.

Once John Thompson III left for Georgetown after Schaen's freshman season, the 6'9 forward's noted ambivalence toward the Princeton offense surfaced, along with his puzzling decision to leave the team and the school for a year for personal reasons. Upon his return, Schaen couldn't shake off the rust that had accumulated on a game that hadn't even fully formed before he left Princeton 18 months prior. Veiled animosity between Schaen and new coach Joe Scott bubbled over into the press, and the once-heralded prospect found himself off the roster by the start of the 2006-2007 season.

Schaen isn't alone as a highly-rated flameout. Numbers 10 and 11 on the list (fourth and fifth excluding the 2011 and 2012 classes) were Quaker big men Justin Reilly and Ryan Pettinella. Neither could consistently crack the rotation, and Pettinella ultimately transferred to Virginia, where he was unable to find any greater success than he had with Penn. Add numbers 13 and 15 to that group, Sam Kaplan and Andrew Van Nest, and the evidence would seem to indicate that recruiting ratings might not just be useless predictors, but that they might be inversely related to success.

I've obviously cherry-picked the more disappointing failures thus far. So, how does the scene look with all of the data included?

The other members of the overall Top 20 (of which we have data for 11, since nine are from the classes of 2011 and 2012) met with varying levels of success, but all could be at the very least deemed good Ivy players. Columbia center Ben Nwachukwu (No. 3 overall) and Penn forward Steve Danley (No. 5 overall) had solid careers despite playing beside much lower rated bigs (John Baumann for the Lions and Mark Zoller for the Quakers) who put him much more impressive numbers. Yale's Ross Morin (No. 12 overall) continued the tradition of New Haven being a post factory. The final three are all still active, as Princeton's Ian Hummer (No. 14 overall) has established himself as a great interior scorer, while Harvard's Brandyn Curry (No. 18) and Penn's Zack Rosen (No. 20) are arguably the league's top two point guards.

Expanding to the Top 20 from the class of 2002 to 2010 subset, we pick up success stories in Patrick Saunders (No. 22 overall), Drew Housman (No. 26 overall), Tucker Halpern (No. 27 overall) and Oliver McNally (No. 28 overall) and disappointments in Khaliq Gant (No. 21), Dee Giger (No. 23), Zach Crimmins (No. 24), Mike Howlett (No. 25) and Garvin Hunt (No. 30). In the cases of Gant and Howlett, injuries played a large role in making each far less of a factor than they might have otherwise been, but injuries are part of the game and can't be ignored if attempting to place odds of future success.

Out of those Top 20, that's 10 successful players and 10 unsuccessful players for a hit rate of about 50 percent. Looking at the next 40 from the class of 2002 to 2010 subset, there are 11 success stories, four pushes and 25 failures (11 out of 36 is 30.6%). Notable players from this group include Austin Morgan (38th overall), Laurent Rivard (40th overall), Kyle Casey (42nd overall), Miles Cartwright (53rd overall) and Tyler Bernardini (58th overall).

The next 40 (61-100 from 2002 to 2010 and 75th to 133rd overall) have a hit rate of seven success stories, two pushes and 31 failures (7/38 = 18.4%). The next 100 fared slightly better at 21 percent, while the final 148 dropped to a 12.2 percent success rate.

THE 40/40/40/40 CLUB

Matt Stehle was an earlier version of Jeremy Lin in Cambridge.

Yes the two played different positions, and no, Stehle didn't have the hoop dreams of an entire ethnicity placed upon his back, but both were stat sheet stuffers - the only guys until Keith Wright in 2011 to light up as nationally rated in ten different Pomeroy metrics. Much like Lin, as a sophomore Stehle carried a terrible Harvard squad as far as he could, only to get a little more help as a junior to push the team back toward .500. The roads diverged as seniors, with Stehle's 2006 squad suffering a monumental Ivy collapse, while Lin's 2010 team posted Harvard's first 20-win season in school history.

While Lin managed mentions, but no rating, from Rivals and Scout (enough to push his rating off the floor of 40 in this analysis), Stehle went completely ignored by those aforementioned two as well as Hoop Scoop (ESPN ratings were incorporated into the analysis as well, starting with the class of 2007). He is joined by 83 other recruits in the group of Ivy players to have received zero recognition from the various scouting agencies.

Of those 84 potential diamonds in the rough, only six (Brown's Damon Huffman, Harvard's Matt Stehle and Brian Cusworth, Dartmouth's Leon Pattman and Princeton's Noah Savage and Kyle Koncz) ever became good to great Ivy players. Given that Mack Darrow might still be a push, that leaves six success stories and 77 failures (6/77 = 7.8%). The numbers grow even more stark when you consider that in the post-ESPN age of recruiting, only 17 recruits have managed to go completely ignored by all four ratings agencies, and the most successful player out of those 17 has been Darrow.

Now, 17 does not a large sample size make. But there's even more evidence that the small correlation between ratings and future success is growing over time.


The modern age of high school scouting started with the Class of 2002, as Scout and Rivals started filling and maintaining their databases with myriad prospects looking for Division I offers. Those two sites joined Hoop Scoop, which had been publishing lists of the top players by class for years prior.

The data set used for this analysis starts with the Class of 2002 and runs through the present, though the performance statistics are only available through 2010. There are a few metrics by which performance is judged, but the most common used in the regressions is the Value Added concept popularized by the Cracked Sidewalks blog.

Regressing the average composite ranking for each prospect from the Classes of 2002 to 2007, we see a very, very weak trend (R^2: 0.02). The difference between the lowest rating of 40 and the highest rating over the past 10 years (Schaen's 91.7) was a Career Value Add rise of 2.8 percentage points (1.9% vs. 4.7%) or about 0.7% per year, which is trivial at best. In other words, if you had a million chances, you'd bet on a bunch of 90s over a bunch of 40s and feel confident that you'd have the better team, but for a roster size of 15, it wasn't anywhere near guaranteed that the 90s would outperform the 40s.

For most, that regression analysis was unnecessary. Penn fans could tell you about Mark Zoller, Ibby Jaaber and Brian Grandieri. Brown partisans could point to Damon Huffman and Mark McAndrew. Dartmouth folks could wax poetic about Leon Pattman and Alex Barnett. And the other five Ivies could join with their own applicable examples.

Things began to change with the Class of 2008, though. It could be that with the Worldwide Leader on the scene, that competition became stiffer and just getting the Top 100 right was no longer the prize - now it was about how far down the list you could rank kids correctly.

Whatever the reason, the results for the 111 recruits to enter the league over the past three classes have improved dramatically. The explanatory power rose from very, very weak to, well, just weak (R^2: 0.16). Now, your bottom rung 40s had an expected Value Add of -0.2% (0% is replacement level, so they would be expected to be slightly below) and the recruits hitting the 90 mark would have a career expectation of 8%, a much more dramatic and meaningful spread than the first six classes in the analysis.

While the predictive abilities are growing, it's important to keep the results in context. The question of whether Zack Rosen's 84 means he'll be a much more productive player than Kyle Casey (74) is still far too fine. The expected difference is just 1.6 percentage points over four years or 0.4 percentage points per year, basically negligible from a value added perspective. But Rosen's 84 does distinguish him from a recruit coming in with an average rating of, say, 50 (5.3 percentage point expected career difference; 1.3 percentage points per year).

Also, not all recruiting services are created equal. Running the same regressions by recruiting service for just the classes of 2007 to 2010 (i.e. the "ESPN-era"), Hoop Scoop graded out by far the worst (R^2: 0.02; statistically insignificant, but positive result). ESPN finished third, but just behind the top two and well ahead of Hoop Scoop. Rivals narrowly lost out to Scout but both possessed about the same predictive ability (R^2: ~0.1; 0.1 percentage point rise in Career Value Add for every point rise in average recruiting rating).


Nine of the 11 best players recruited into the league since the Class of 2007 will be members of the 2011 and 2012 entering cohorts.

Six of these (Wesley Saunders, Kenyatta Smith, Mike Hall, Siyani Chambers, Evan Cummins and Jonah Travis) belong to Harvard, while three (Patrick Lucas-Perry, Jamal Lewis and Henry Brooks) belong to Penn. In fact, the first non-Harvard or Penn 2011 or 2012 recruit on the list is Cornell's Shonn Miller (33rd overall since 2007 and 14th among the 2011 and 2012 entering classes). Miller is also the only non-Crimson or Quaker recruit to merit an average rating above 70.

Does this signal a rising Harvard and Penn duopoly at the top of the league?

Once again, even the strengthening of the recruiting rankings' predictive abilities over the past three classes has left us with a decently weak regression. But, if correct, the difference between nabbing a kid in the upper 80s, where Harvard and Penn have been hunting, and the mid-60s, where most of the rest of the teams have been falling, is about 4.0 percentage points of Value Add for a career and 1.0 percentage points per season. With a couple of kids, it would be hard to be confident that this relationship would truly manifest itself. If the Crimson and Quakers keep replicating this gap, though, the likelihood of the ratings differential leading to on-court performance differential will increase.

The recruiting rankings are nowhere near good enough, nor will they likely ever be, to allow us to make a definitive judgment about whether a 74 like Greg Louis will be better than a 72 like Shonn Miller. Or even whether an 84 like Henry Brooks will certainly outperform a 66 like Steve Moundou-Missi.

Unlike in the advent of these ratings, though, there is enough of a relationship to determine that a group of five 80s or 90s will almost certainly outplay a group of five 40s or 50s. While such a statement should seem obvious, that result wasn't as certain five years ago. More importantly, with Harvard and Penn hunting in that 80 and 90 range and other teams struggling to get out of the 60s, there's a possibility that we already have a window into future results today.

No comments:

Post a Comment