The Optimal Draft Strategy: BPA or Fit?

November 2022: (Attempting to) settle the debate

This was written and published in Swish Theory’s Roundtable: Best Player or Best Fit? (November 2022). It can be found here. I’ve included the opening paragraphs from the article, as well as my own thoughts.

NBA Draft analysis is a major focus of many Swish Theory writers. But we aim to not evaluate players and their skills in a vacuum, but also assess attributes with an eye to how the NBA game is played and team-building works in practice.

Our second roundtable will be focusing on just that: if you ran a front office, how would you establish your draft philosophy? To keep things simple for now, we’re placing at one end of the spectrum drafting for fit and the other drafting purely for best player. Our participants picked their philosophy from a menu of five options:

1. Best Player Available Only

2. Preference for Best Player

3. Equal Consideration for Best Player and Best Fit

4. Preference for Best Fit

5. Best Fit Only

1 – BPA Only

My issue with the BPA versus best fit debate: when drafting for best player available, we need to account for a requisite level of organizational trust in the ranking capacity of each team. The framework of analysis needs to be robust, in that a successful draft scouting mechanism can be replicated every season. Of course, this is highly dependent on the organization, and may not be possible as industry standard.

But on the contrary, proponents of the “best fit” proposition suggest a baseline of roster continuity – the foundation of the team will remain for the foreseeable future. While roster continuity has improved over the last few years, this is still quite a relevant confounding factor. “Best fit” also requires a requisite level of organizational trust in its ranking capacity: you need to trust that your projections are pristine, and then also compare prospect projections to determine which prospect would most effectively mesh with your team’s current talent.

So, we’ve established the following: BPA relies on

1. Trust in the team’s eval.

Best fit, meanwhile, relies on all of:

1. Trust in the team’s eval

2. Deciding which of the projected evals would mesh best with the core

3. Relative faith in the roster for the near-future

This is the basis for my fundamental pitch of BPA: it’s the probabilistically correct decision. For the sake of the year-to-year variability in the league, give me the option with fewer variables.

Fit is a nebulous concept, but I’m defining fit here as team fit – considering how well the team can integrate the drafted player into their lineup. Considering the developmental strengths and training bandwidth of the organization seems relatively intuitive: for example, the Raptors and Spurs have demonstrated a propensity for developing shooting in wings (let’s ignore that shooting dev is perhaps the most “elastic” skill acquisition endeavor). Still, I don’t consider developmental strengths a true function of fit because it’s personnel-dependent and thus not organizationally intrinsic.

Additionally, I personally believe that feasible skill dev is casually deterministic. In other words, players showcase varying propensities to develop skills such as shooting, and it is up to the organization to find those players and tap into their developmental strengths to allow those players to become the best (shooting/skilled) versions of themselves. As an example, Kawhi Leonard, often considered one of the great skill development cases, had high FT% + a ton of self-created long 2’s, indicating a propensity to develop off-the-dribble shooting later. Judging the skill acquisition and refinement potential of prospects seems far more BPA-esque, especially since this process is similarly not intrinsic to each team. 

Two more concepts that make me such a firm believer in BPA:

First, I think it’s hard to truly manufacture poor team fits in the draft. From the CP3-Harden rockets to ‘21 Nets, the idea that “there’s only one ball” is incredibly ignorant of the lineup versatility enabled by the sheer concentration of talent. I think you can probably pair any combination of players who can consistently draw defensive rotations. There are reasonable limits (don’t construct the 2022 Lakers), but oftentimes the well-roundedness of a team to meet the minimum offensive and defensive thresholds of goodness can be attained in free agency. Also, while there are some extreme cases (taking a traditionally low Expected-Value archetype high in the draft while already employing a high-level player of that archetype, like taking an undersized guard top 5 on a team with Trae Young), even those cases tend to work out. Darius Garland was drafted a year after Collin Sexton, and Anfernee Simons was drafted to a team with Dame and CJ. Five years later, they are both the lead guards for their respective teams. The concept of constantly drafting the best player available and seeing what sticks may not be the most asset-productive move in the short-term, but ultimately the players that adjust the best to the NBA ecosystem end up returning far over expected value. Note that I am not supporting a Darwinian-aligned view of player dev, but instead optimistically believe in the culmination of the intrinsic “dawg” + existing bundles of potential skill. Another side note: dev-locking (ex. Josh Christopher right now for the Rockets) happens too but that’s usually a byproduct of drafting too many players in a single cycle.

Finally, predicting skill dev is extremely hard. Especially in this era of ball-handling wings and position-less schema, there are increasing cases of outlier development and miscalculations of potential. That does not mean you conflate “take the best player available” with “shoot for the moon and find the next Giannis!!”; Instead, scouting departments need to research the underlying skill, neuropsychological, and athletic traits of outlier development, and find players who rank highest based on a holistic evaluation of their potential. Taking the highest ceiling is not the same thing as taking the best player – and it’s incredibly reductive to make that assumption. BPA can be an innately nuanced process without considering team-context. Consistently drafting players who have the highest probabilities of commanding defensive rotations may be a personal philosophy at its core, but I truly believe BPA is more probabilistically viable than conducting the same exact eval process and then strictly trusting your projections to conduct an additional decision-making process that determines which of the players would best fit the current roster. That doesn’t even include the overlying “best-fit” assumption of roster continuity.