This was post number one as part of the Harvard Sports Analysis Collective’s Linsanity Day. You can read the other posts here, here, here, or here.
by Andrew Mooney
On February 9th, John Hollingerās Player Efficiency Ratings on espn.com displayed one very curious result. There, sandwiched at number two between Lebron James and Kevin Durant wasā¦Jeremy Lin?
Itās been two weeks since the basketball world was first caught up in the throes of āLinsanity,ā as Jeremy came off the bench to torch the Nets for 25 points and seven assists. In his next four games (all starts), Lin scored 109 points and effectively revitalized the floundering, injury-ridden Knicksā season. Predictably, a host of terrible puns, nicknames, and general Internet absurdity soon followed. For a moment, I almost missed Tebow.
For my part, Iāve regarded Linsanity with a more tempered eye (despite his excellent choice of undergraduate institution). This sports year has already seen one Harvard grad burst suddenly into prominence after an otherwise nondescript career, and as weāre now aware, Mr. Ryan Fitzpatrick found it quite difficult to sustain his lightning start. Will Lin have a similar regression? Letās dig into his past to find any clues that might tip us off to his true ability.
Linās first NBA experience came with the Golden State Warriors, signing as an undrafted free agent after a stellar senior year with the Crimson. At Golden State, Lin was an entirely unremarkable bench player. His primary problem was getting on the court. Lin appeared in only 29 contests, averaging just 9.8 minutes per game. Yet even that number was inflated by extended spells of garbage time; he played over ten minutes in only 13 games, and the average margin of victory in those games was 15.9 points. As such, itās hard to draw many conclusions from this period. He rarely played, and when he did, it was often when the gameās result was no longer in doubt and, thus, not really representative of true NBA competition.
Lin spent the remainder of ā10-ā11 with the NBDLās Reno Bighorns, and it was here that he showed flashes of the skills weāve seen over the past week. Over 20 games for the Bighorns, Lin averaged 18.0 points, 5.8 rebounds, 4.4 assists, and 2.1 steals, while playing 31.8 minutes per game. However, as no rigorous translations exist for projecting D-League performance to the NBA, itās tough to discern whether these numbers actually represent an improvement in his game.
As Matt Kamalsky of Draft Express noted during his time with the Bighorns, Lin was āextremely tough around the basket and show[ed] a very good understanding ofā¦how to use screens and subtle changes of direction to turn the corner off the dribble. Lin won’t land on a highlight reel any time soon, but he gets the job done in the D-League.ā
While a generally positive assessment, it certainly doesnāt scream of potential NBA stardom. So what changed.
First, the toughness around the rim alluded to by Kamalsky has only increased. Lin has consistently displayed an ability to finish at the basket throughout his career. With the Warriors, Lin converted 58.1 percent of his attempts around the rim. This year, with the Knicks, he has boosted that number to 63.3 percent.
The main weakness in Linās game was his inability to knock down midrange jumpers. This was something about which Kamalsky expressed concern in his assessment of Linās D-League stint: āhe’ll need to become a reliable set shooter to give himself more staying power in the NBA.ā And for good reason; in Golden State, he was just 1-4 on attempts from 10-15 feet and 4-15 from 16-23 feet. As anyone who is familiar with Rajon Rondoās career can tell you, this is particularly damaging for a pick-and-roll point guard who prefers to get into the lane. Defenses are perfectly content to allow those shot attempts if the guard canāt knock them down.
Yet, with the Knicks, knock them down he has. From 10-15 feet, Lin is 5-9, and from 16-23 feet, heās 12-19. Overall, the improvement in his shooting from 38.9 percent (28-72) with the Warriors to 49.7 percent (71-143) this year is highly significant, with a p-value of 0.004.
What accounts for this progression? Maybe, to quote Glenda the Good Witch, heās āhad it all along.ā For one, he simply didnāt (and still doesnāt, really) have enough shot attempts in his first NBA go-around from which to draw any firm conclusions about his game. His poor shooting with Golden State could have been a fluke, and the improved numbers weāve seen from Lin this season might be more indicative of his true ability.
Or perhaps he worked tirelessly on his game during the offseason and lockout, priming himself for one more shot at the pros. Personally, Iām more inclined toward this second explanation. In responding to his recent success, the one sentiment expressed universally by his former teammates at Harvard has been his extraordinary commitment to hard work. The unconventional form on his jumper also suggests that heās never been a natural shooter, but with enough practice, he may have found a method thatās effective, if not pretty.
This week promises a couple new wrinkles in the Lin saga: notably, the reinsertion of Carmelo Anthony into the lineup. How will Lin coexist with the Knicksā biggest star? Can he prevent the offense from devolving into the dual black hole attack it was a few weeks ago, with Anthony and Stoudemire jacking up shot after shot? Iāll be interested to see what happens to Linās own production. He wonāt remain the focus of the offense, but heāll get his looks. Can he continue to knock down open shots, as defenses rotate toward the Knicksā two elite offensive players? Stay tuned; or, if youād rather not, avoid the Internet for a few days.
I have yet to read an article on the stability of basketball stats from game-to-game, but a basic two-proportion Z test may not be entirely appropriate. In a hypothetical situation, if his true probability of making a shot changes from shot to shot, and from game to game, then the binomial distribution can greatly underestimate the variance of his sample proportion (p*(1-p)/n). Now, that may be an extreme case, and that p-value of .004 suggests a highly significant difference even with some potential increased variance. But something to think about…