A Manifesto for the “Best Player Available” Approach

Apr 4, 2015; Indianapolis, IN, USA; Kentucky Wildcats forward Karl-Anthony Towns (12) and forward Willie Cauley-Stein (15) celebrate a basket against the Wisconsin Badgers in the first half of the 2015 NCAA Men
Apr 4, 2015; Indianapolis, IN, USA; Kentucky Wildcats forward Karl-Anthony Towns (12) and forward Willie Cauley-Stein (15) celebrate a basket against the Wisconsin Badgers in the first half of the 2015 NCAA Men /
Apr 4, 2015; Indianapolis, IN, USA; Kentucky Wildcats forward Karl-Anthony Towns (12) and forward Willie Cauley-Stein (15) celebrate a basket against the Wisconsin Badgers in the first half of the 2015 NCAA Men
Apr 4, 2015; Indianapolis, IN, USA; Kentucky Wildcats forward Karl-Anthony Towns (12) and forward Willie Cauley-Stein (15) celebrate a basket against the Wisconsin Badgers in the first half of the 2015 NCAA Men /

“Draft the `best player available’ (BPA)” is not exactly a controversial position. Many coaches, general managers, and fans advocate “BPA” over “Fit”. Still, when you listen to draft discussions, or read through mock drafts in the weeks leading up to the main event, you are met with countless references to how different prospects mesh with the rosters of drafting teams. I assume some of this is just a function of it being tough to construct an interesting narrative without emphasizing roster synergies and antergies. However, teams do often make draft day decisions that are clearly motivated by need or fit, and I would argue that this is to their detriment.

When pressed, most folks defend a “tiered” system. Tiering involves ordering players into hierarchical groups exclusively based on talent/potential, but then making decisions based on fit within those tiers. There is nothing wrong with this concept, but it doesn’t actually do anything to solve the problem of trying to optimize between BPA and Fit. The useful debate is over where you should stand along a continuum of tier sensitivity. Strict “BPA” is nothing more than a subset of the tiered approach with all tiers of size = 1, and strict “fit” is nothing more than a subset of the tiered approach where a single tier encompasses all players. Here I am arguing that you should shrink those tiers to the point where fit only comes into play when all other features fail to discriminate.

Talent always wins:

I mentioned above that mock drafts consistently make reference to current roster construction and how a player fits into each team. Interestingly, in the common “redraft” articles that discuss what the order should have been in retrospect, any mention of fit is conspicuously missing. Individual tastes may vary, but if you are making a redraft list for 2012, your order will look something like: Davis, Drummond, Lillard, Green, Beal… The order of the teams making these selections is completely irrelevant. Try to think of a case where in hindsight you would advocate selecting a player X who you now view as inferior (even slightly) to player Y, with player Y still on the board because of team fit. At the extreme end, nobody thinks it was a good idea to pass on Michael Jordan because of redundancy with Clyde Drexler, but it is a difficult task even with a much tighter talent gap than Jordan – Bowie.

To be fair… Conjuring cases for the thought-experiment above might get a lot easier when you move away from star players into the supporting cast. Locking in a star player through his rookie and second contracts is the key goal of the draft, but most players do not become stars. Rooking-scale contracts can still be a great value even in these cases and here considering something like selecting a point-guard over a slightly superior big-man based on fit seems much more appropriate. That said… I am still skeptical.

Rookies are not good:

Useful rookies are few and far between. Even many players who ultimately become good pros, or even stars, hurt their team whenever they are in the game as rookies. Nikola Mirotic is the only rookie in the 2015 season who played a meaningful role on a contending team. Some rookies on lottery teams played heavy minutes and put up numbers, but even the best of these were roughly neutral value due to flaws they likely will overcome in time.

Here is the impact past rookies have had on their team:

rookz /

The red dots in the plot above show the expected value of prospects taken in each draft slot. The first pick is expected to produce like a real NBA starter (~5 WS). No other pick is likely to offer even that, and expectations quickly drop below average rotation value (~2.5 WS) before we move out of the top five. These numbers likely overstate the real impact of rookies. Top picks are often force-fed minutes, which artificially inflates an aggregate statistic like Win Shares. Looking at rate metrics the value of rookies is even bleaker.

Rookies simply are not good enough to matter for roster construction. Not only is the baseline expectation for a rookie low, but the variance is far too high for a team to wisely view prospects from a “fit” perspective. The pink dots on the above plot show one standard deviation on either side of the mean expectation at each pick. Based on where the lower points fall, not only can most rookies be expected to play only a marginal role, but there are high odds that they will be completely useless. This means that drafting a player in a needed role does little to confidently fill that spot in the roster. The team still needs to find a more reliable veteran through free-agency or trade to secure that position.

I doubt these numbers are a shock to anyone’s world-view. The ineptitude of rookies is common knowledge. The obvious response to my argument here is “who cares, I’m not talking about fit next year, I’m interested in the long-term fit with the team.” This counter makes sense… but only because our confidence in what the future will look like is unrealistic.

You cannot fit a player into a roster you cannot predict:

Below is a plot showing the average proportion of a team’s minutes accounted for in each successive season by players who were on that team in the initial season:

churn /

You should not factor initial fit into draft decisions because rookies are unlikely to matter, but accounting for fit further down the line is equally problematic due to roster churn. On average, half of the team has been overhauled by a player’s second season, and by that third and fourth season when rookie-scale players are typically most valuable, the team has almost completely changed. It makes little sense to select players in the 2015 class based on fit, because it is impossible to know what rosters will look like in 2018 when the current class becomes meaningful contributors.

Assumed core pieces fail to develop as hoped, unexpected players suddenly become central to the team, and previous stars leave in free-agency, lose their luster, or fall prey to injury. The Spurs, Clippers and Blazers were the only 2015 teams running out the same core envisioned in 2012 (both the Clippers and Blazers may soon be losing at least one piece, and the Spurs are the exception to every rule). In contrast, four teams are without a single player from the 2012 season (HOU, PHI, ORL, MIL) others are completely inconsistent with whatever their imagined core was at that time (BOS, NOP, PHO, DET) and still others are down to one significant hanger-on who may or may not be a part of the future (NYK, LAL, MIN , WAS, DAL, SAC, CHO). Even looking at the recent championship contestants, three year ago today the Warriors had just made a tough call on Steph Curry over Monte Ellis and were cautiously optimistic about Klay Thompson’s solid rookie season, while the thought of LeBron going back to Cleveland was a sad little fantasy held by only the most deluded of Cavaliers fans.

In all likelihood, you do not know what your team is going to look like throughout most of the relevant part of current prospects’ careers. This makes fit-based decisions awkward, since you cannot fit for an unknown. Instead, fit decisions are modeled on a fantasy where all young pieces develop optimally and currently established players neither unexpectedly deteriorate nor leave.


One important caveat in all this is that nobody actually knows who the best player available is at the time of the draft. I can only advocate taking who you think is the best and accepting the noise inherent in that process. This is again a place where someone is likely to come back with a fit argument. “If the error around your evaluation is larger than the difference between two players, maybe you should consider fit?”… maybe, but as I noted above, the errors around your assessment of fit are likely just as large. Not only that, but errors in player assessment not only apply to overall quality but also the actual skills that player will develop and role he is going to fill. Players consistently surprise in ways that impact their fit as much as their net production.

Team-design is important, but that design is a structure you construct around talent, not a defined box with placeholders you look for the appropriate talent to fill. There are better ways than the draft to deal with gaps and redundancies in the roster. Free-agency offers the opportunity to patch holes with known quantities, and even in the absolute worst-case scenarios of talent blocking talent (e.g. Chris Paul blocking Eric Bledsoe) there is always a strong trade market for good young players. Hitting in the draft if simply too important to allow eliminating options for any reason other than future potential. Take the best player available.