Nylon Calculus Discusses Youth and Fit

Sometimes, the best stuff the Nylon Calculus staff comes up with is in our e-mail discussion threads. We previously took a long and occasionally contentious look at the utility of “one-number” metrics and devoted much of last week to talking about the idea of replacement level. With the NBA Draft on the immediate horizon, our staff spent the past few days talking about the ways in which youth and fit should enter into a team’s draft evaluations.

Ian Levy (@HickoryHigh): Age is now creeping up on wingspan as one of the most frequently cited characteristics for evaluating draft prospects. The appeal is obvious–the younger player the more room they theoretically have to grow. However, it seems to me that the younger a player is, the more variance there is in projecting their growth. On the other hand, while a four-year college player may imply someone who’s ready to contribute immediately, should it impact their evaluations that their performance came primarily against younger player?

How are we to weigh age as a variable in projecting draft prospects? How does age work with a team’s timeline of improvement and work with the larger concept of fit?

Layne Vashro (@VJL_Bball): I’d have to spend more time to give a better answer, but I’m not sure that there actually is less variance in older draft prospects.  For example, there were two seniors taken in the lottery last season.  They were both terrible.  Payne was probably the worst player in the NBA, and McDermott not only failed overall, but struggled hugely to do the one thing he was projected to have a near lock on (shooting).  If older players are both more predictable and more ready, these guys both should have been near locks to succeed.

Regarding the importance of age generally… it seems clear to be the development with age runs on a diminishing curve.  The difference between 17 and 18 is huge, that between 18 and 19 is a bit smaller …. once you get to the difference between say 23 and 24 it starts to be much less interesting.  However, most college guys are well within the range were a one year difference is very very meaningful.

Seth Partnow (@SethPartnow): My main question is, how much selection bias is there for age? If the “best” prospects to begin with (however defined) tend to enter the draft asap, does that skew the age effect?

Andrew Johnson (@CountingBaskets): I think the increased variance intuition would be correct if the more talented prospects did not tend to enter the draft earlier. If it was a random entry process having more data and moving further along the development curve would cut a little bit of the variance. Part of the appeal on the team side for adding another year before players become eligible for the draft, no doubt.

As it is high school recruiting rank is the best correlated variable to earlier entry into the draft. My current model uses the consensus recruiting rank, which lowers the influence of the age variable.

One way to look at Seth’s question on selection skew is to compare the strength of the age variable with expected aging curve. Before I added recruiting rank the effect was in the high side of my age curve estimates, with that variable it is on the low side. Though I think both are in a reasonable range.

Vashro: I initially assumed this was an issue, but having looked at it over the years I am increasingly convinced that it is not.  For a simple test, if we assume players are getting an artificial bump for being young due to the association with star early-entrants, we should expect a sharp drop in year-to-year projections with age among guys who stick around.  This doesn’t seem to be the pattern.  Here are some examples of my EWP model’s projections from the current class:

Sam Dekker is the only case that fits the expected pattern.  Kaminsky is a relatively extreme case in the other direction. He did not look like a prospect as a freshman, in spite of benefiting from the age bump in the model.  He then upped his game at a much faster rate than is typically expected over the next three seasons.  The rest of the likely-1st-round upper-classmen moved right along their expected curves and thus didn’t see their scores move much at all.  I should be clear, this is obviously a biased set since it focuses on guys who are being hyped as juniors or seniors, in many cases more so than they were in previous years.  That said, I have trouble coming up with examples of guys who projected as great prospects young then fell off hard as they stuck around (I’m sure there are examples though).

Overall…  most of the top prospects do leave early.  This means that older players tend to project as weaker pros largely just because the good guys have already left.  However, this doesn’t seem to pervert the age effect.  I’m confident that if the Anthony Davis’ and John Walls and Kevin Durants did stick around, the perceived “potential” gap between them and the rest of the NCAA population would be the same as it was in their freshman seasons.  They would just be putting up more and more ridiculous lines as they progressed through their collegiate career.

Johnson: The way I tested it was a just a little different than Layne. My model uses an age curve built from testing the expected improvement of NCAA prospects who don’t any year after their freshman year. I have alternated between that and a simple log linear adjustment, but the age curve tends to do slightly better in cross validation. That way I can test the model results to expected relative strength of the age curve.

Essentially not including any scouting ranks, either out of high school or draft makes the age curve a little stronger than expected, but not overwhelmingly.  In either case if a prospects improves year to year along the expected age curve they will produces the basically the same model rating.

Another interesting tidbit I found was that the draft entry year tends to have noticeably higher than expected improvement. A sophomore prospect who has an unexpected jump in production, hello Victor Oladipo, is much more likely to jump into the draft than with typical progression. (Or interpret that as staying in as freshman with a very disappointing year). In other words, college kids might be rational.

To turn things around a bit, I wanted ask what people think we can confidently say about both adjusting for opportunities and situations in college ball that prospects face and what we can say about development in the NBA?

Partnow: Shouldn’t that be under the subhead “Boeheim Draft Blunders” Andrew?

Johnson: Yes, probably.

The stats head in me wants to object that in a multiple comparison case, some coach is bound to have a bunch of busts. But, we do have good reason to believe the zone defense he plays inflates steals and blocks and probably isn’t the best prep for the next level. Plus, I don’t like the way he runs the program in general. So I mentally dock kids coming out of their a peg or two, though not in any actual model.

Nick Restifo (@itsastat): Unfortunately, my draft model does not account for scheme or defensive style, both of which undoubtedly have an effect on player performance, and thereby, player projection. Most models use pace adjusted statistics, and in this way account for the speed at which a player’s team plays at. The influence of defensive scheme on a player’s block/steals has been brought up before, but this concept is of course not limited to defensive scheme. Offenses too, play a role in inflating and deflating a player’s opportunities for assists, rebounds, field goal attempts, and three point rate. To use an NBA example, the implementation of the triangle offense took a (similar) Knicks team from iso heavy to pass heavy in a few games. Imagine what this could mean for something like a draft prospects’ assists, one of the more predictive player traits (at least in my own studies).

Levy: Please give me a nudge if this is an inappropriate metaphor, but when I was in college I worked evenings at REI. There was a lot of professional development done for the sales staff and we were steeped in the language of features and benefits. The features of a product are exactly what they sound like. For a tent, it would be the weight, the durability of the material, the head space, the floor space, the ease of set up, the LED zipper pulls, roomy mesh pockets and every other conceivable bell or whistle the manufacturer could attach. As a salesperson you were supposed to talk with the customer about and highlight the specific features that matched their needs — turning them into benefits.

Thinking about the draft, it strikes me that each player is a unique combination of features. Some of offer more, some offer less. But ultimately the draft evaluation process (and a player’s overall value) are primarily dependent on the conversion of features to benefits, matching a player’s abilities to a team’s needs. Even a product (player) with less features (skills) may offer move overall value (value) if those features (skills) match a customer’s (team’s) needs. It would seem to inject a heavy dose of context into the final result.

Circling back to the original idea of youth and my heavy-handed metaphor here, is youth a player aspect that is a universal benefit?

Johnson: I’ve never really understood the REI membership plan.

To back up here, I am a big best player available in the draft guy. Layne has laid the case out really well so I don’t want to steal his lines. But the key is not to think of draft picks as something that’s going to return value immediately, so immediate needs shouldn’t play into it. I am struggling for the right metaphor, like buying a tent for a trip you might take in three years at an undisclosed location with an undetermined group?

On to youth, to me it’s merely a variable to help determine expected value. I do get a lower age adjustment for models based on the first two years. Also, there are certain skills that translate over more quickly, like rebounding and shot blocking. So one could argue that a team with a very high discount rate could judge best player available differently. However, in most instances I think those teams should look at the free agent or trade markets to fill those immediate needs, and view the draft as a pipeline of future talent.

Restifo: How valuable is fit when your’e already in position to be drafting high in the NBA draft? Do you really want to pass on the best player available because he doesn’t fit in quite as well with your 25 win collection of players? While there is definitely value in quantifying fit, I don’t think (when all is said and done) that it will be found to have a significant influence on where early first round prospects should be taken.

Johannes Becker (@sportstribution):Intuitively I would treat fit like I tread wins in the MVP race. If two candidates are super close, I’d pick the one that I can more easily distribute minutes too.

Generally speaking: Is there an example where drafting a Top 5 pick for fit worked out well? I remember not drafting Michael Jordan because of Clyde Drexler didn’t, but one example is rather anecdotal.

Restifo: I do, however, think there is something to be said for drafting fit in the late first round and into the second. At that point, when the players being drafted already have such a small chance to succeed in the NBA, why not gamble on a player who serves a need for your winning team? But then again, who knows what your team looks like in four years, when that drafted player is beginning to peak, (if he is a meaningful player at all)?

In general, once we get past the first round, I think teams should be looking for variance. With such low chances that your drafted player can contribute or even play, why not gamble on those players that are the most intriguing from a risk/reward standpoint? Many teams already do this, but I always shake my head when I see a decent but not memorable prospect with limited physical tools taken in the second round over taller, more athletic, more exciting options. If you’re really looking for someone like that proven producer with limited upside, that kind of player can be signed after the draft or promoted up from the NBDL.

Partnow: The thing is, those kinds of players can’t be signed or promoted from the DLeague. I mean they can, but each individual guy is most likely not NBA caliber. Getting a serviceable backup PG or 4th big man on a second-round salary isn’t nothing either. The “potential” guys might just as often be those players with great measurables but no idea of how to play basketball, how to put those “tools” to work.

An argument in favor of “fit” is players as assets don’t have static value. No matter how highly a guy is thought of when he was picked, if he never plays, he’s severely depreciated in value. Why take a guy in the second round who by virtue of being positionally blocked has next to zero chance of making your roster? At that point you might as well light the pick on fire (or sell it for cash, which is functionally the same thing.)

Positive Residual (@presidual): Layne’s post really prompted a lot of thinking on my part. Perhaps the most critical reflection is that the draft might be best suited for a long-range planning strategy, yet teams too often (whether by design or by accident) bias themselves toward the short term by overvaluing “fit.”

An alternative framing of the issue might be one of “mismatch.” Maybe teams really are striving to build for the long haul, but their current roster needs get weighted too heavily — like some form of hyperbolic discounting.

Overall, I can only imagine how difficult this stuff is for teams, especially with restless fan bases that are tired of losing. Given the complexity, perhaps there’s something to be said for an approach of simplification: it’s hard to forecast players; it’s hard to predict the future; therefore, just go with the talent, unless there’s a really compelling reason to deviate. But put the burden of proof squarely on “fit.”

Partnow: A lot of what PR is concerned about should be completely unsurprising once we stop viewing teams as monolithic. Draft selections are made by general managers, and while GMs aren’t held to quite the same demands for instant gratification as are coaches, how many execs (aside from Sam Hinkie) are really allowed multiple years of “wait and see” before selections are expected to produce. Especially if notions of BPA can only be vindicated by oncourt production, picking a guy who is “blocked” from getting on the court is akin to setting oneself up for failure for a rational, self-interested GM. Take Jusuf Nurkic. What if JaVale McGee doesn’t get hurt and the Cavs don’t offer two first for Timo Mozgov? He maybe never gets on the floor and for an already-under-fire GM, “missing” like that could have you out on the street.

Of course, truly functional organizations work hard to align the incentives of individual employees with that of the franchise as a whole. But that’s the exception rather than the rule, it seems.

Johnson:I am always a little skeptical of the variance arguments, simply because I haven’t seen any good models that separate that from expected value. We can intuit about athleticism, length and youth, but a versatile player like Draymond Green who could have easily fallen into a tweener trap had a pretty high variance.

In terms of the blocked position issue, I think that is an issue teams need to take into some consideration. There should be at least a development plan, and right now there is enough of a stigma on the D-League, any player only playing down there certainly by their second year is bound to lose value.

Restifo: These are all good points. While not really a “fit” consideration, any team drafting should keep in mind their relative standing in the league. I think a team like the 76ers should have much higher risk tolerance while someone like the Warriors should take a more conservative draft approach. It is important to note, however, that fit and relative league standing should be considered minute when compared to the differences between how good players are. If one player is demonstratively better than the other, that is the player that should be drafted, and these things should really only be considered when there are severe differences between player choices and when the best player available notion is in question.

At every level of basketball (and other sports), I think it is important to not fall into the trap of drafting  “a backup PG”. Draft a player and let player performance dictate what’s what.

Success is very noisy indeed. No player can be a draft success without opportunity, and the opportunities go to those players with reputations. The further into the draft we go, the more the reputations of player’s dwindle, and the shorter the window of opportunity for each player gets. There are undoubtedly plenty of players who would have been at least been “good-enough” in the NBA, but never got the opportunity they perhaps deserved. Many of the draft’s best finds were simply players that seized an opportunity(like an injury to a player above them on the depth chart) to increase their value league wide. Without the Knicks, does Langston Galloway have an NBA career?

In the second round especially, player value depreciation is a factor that must be kept in mind. The moment a player is drafted, (It’s hard to refer to player’s as assets but), a team’s asset moves from that of a draft pick to a player. And unlike draft picks, the longer the league has seen not seen a player play, the less valuable (fairly or unfairly) that player is. You better believe the 76ers could get more for Dario Saric on draft night last year then they could right now. Swinging for the fences on the flawed 6’9 Euro SG with a much smaller chance of ever playing a minute in the NBA does indeed come with a much greater chance of “lighting the pick on fire”, as Seth says.

Johnson: Ugh, another definition issue! In terms of the draft fit is commonly used as positional roster fit need, such as, “The Celtics need a power forward to replace free agent Brandon Bass.” Or a specific skill, “Charlotte needs shooting.”

The first, and most frequently used definition, is the least defendable. But, the second one is not much better, in my opinion, since drafting an overall inferior player too high for one skill represents an overpay. Lastly, there are style of play issues, such as, “we like players who can switch on the pick and roll.” But I think that is more of an issue of weighting that skill in your best player available calculation than “fit,” as it’s not dependent on who is currently on your roster.

Partnow: Considering how we just talked (at excruciating length) the difficulty in universalizing baseline player values. how can we address BPA without reference to fit? How “good” would Tyus Jones be in Atlanta? It’s a trick question because shots from 28 feet on the baseline are tough from a seated position, plus they don’t count.