Backpicks GOAT: the value of longevity and defense

The first time I ranked players using this method, the effect of longevity surprised me.

I think we learn a lot by comparing players through a career value lens like this, but it’s not how most of us typically think of “greatness.” We aren’t wired to rigidly add up the value of just about anything, and there are cognitive forces at play that minimize longevity in these kinds of comparisons. We don’t feel warm and fuzzy about ranking a guy who played 25 good years over someone who played eight great ones.

Consider the case of 2011 James, LeBron v. 1994 Jordan, Michael. Many of us “penalize” LeBron for his lackluster Finals series, calling into question his “reliability.” There’s some social hardwiring involved here; in a way, we “asked” LeBron to perform, and he let us down. Even though Michael didn’t play in 1994 (thereby providing no value), we didn’t “ask” him to do anything. If we did, we assume that he would have delivered based on his other performances. Averaging provides cognitive security; those who regularly deliver make us feel more confident they will come through again.

On the other hand, two or three seasons can feel too brief for validation. For many of us, players need to book a certain number of years to solidify their greatness. But once that happens, longevity becomes a wash — the average supersedes the sum — even though players keep adding value. We focus on top speed when total distance traveled is the objective; career-based lists often prioritize by peak.

So, are we overvaluing peak seasons?

Is longevity underrated?

We can test this with a simple thought experiment. Imagine three separate championship-over replacement player (CORP) curves:

  1. A linear one — this provides the same advantage an All-Star has when compared to a sub-All-Star that an all-time season has when compared to an MVP season.
  2. A superlinear one — essentially the “actual” CORP values — this provides a larger and larger advantage for seasons as they move from weaker seasons (like sub-all-star) to stronger years (like MVP seasons).
  3. A “steep” exponential one — this provides huge chunks of value for great seasons when compared to all-star or even all-league level ones.

The linear model is totally unrealistic based on the data we have about player impact and the historical odds of winning a series, while the steepest curve is likely too extreme and slanted toward top-end seasons in order to make a point.1 Still, if we used these three curves and plugged in the same seasonal valuations I used for the Backpicks GOAT, they would spit out the following rankings (not adjusted for era):

Notice that a low-peak, high-longevity career cracks the top-20 in a linear model (Stockton), and degrades in value as higher-peak seasons are given more and more weight. But, even in the steepest model, both Stockton and Reggie Miller are top-35! This is largely a function of scarcity — only 31 players since 1955 have been named to at least eight All-NBA teams. The antithesis of these players — a short-career, high-peak star like Steph Curry — moves from 44th to 34th to 26th as rarer seasons are given more value. However, it takes an extreme treatment for a player like Curry, who already has seven healthy NBA seasons under his belt at the time of writing this, to pass the Miller type.

Now look at Jordan (and most of the top players). Despite holding the highest peak ever in my estimation, it takes the steepest curve for Jordan to (barely) pass Kareem, and that’s before era-specific longevity would tip the scales back to Jabbar. So, it seems that beyond a simple difference in rating criteria, we intuitively undervalue longevity in these kinds of rankings. What’s going on here?

“First options” and the longevity threshold

When stacking up the multi-season MVP giants, peak play is often given preference. For most of us, it doesn’t matter that Kareem provided more career value — that thought isn’t even in the equation sometimes — it’s just that Jordan had a better peak, and Jordan didn’t have a short career, and thus Jordan goes first. Once all the multi-year MVPs are off the board, it might even be OK to slot in Bill Walton, because, well, he was better than all the longevity monsters when he played. (Walton was 38th on Elliot Kalb’s list and 27th on Bill Simmons from only one great healthy season.) Ironically, the more someone plays, the less their longevity seems to matter.

Ranking a sort of peak-prime combo does answer the question “who was best, at his best, for at least longer than just a few seasons?” And maybe that’s what most of us mean when we ask ourselves who was “greater.” I don’t think any GM would ever draft like this, but it’s an approach that seems to align more closely with our concepts of “greatness.”

There’s certainly nothing wrong with ranking this way. I’m not fully on board with it because of how challenging it can be to deliver value for an extended period of time. For instance, Larry Bird’s retirement nearly crippled the Celtics franchise in his 13th season, but Kareem was a key cog on four more Laker title teams after that point in his career. Kevin Garnett’s 13th season was his DPOY and near-MVP double in Boston. Stockton’s 13th and 14th seasons helped Utah to two close Finals losses. Replacing these players is really, really hard; All-Stars don’t grow on trees.

Which leads to the issue of “first options.” Most of us intuitively dismiss seasons in which a player can’t be the lead figure on a championship team — we can’t rely on them to carry the load. But the second, third and fourth-best players make it possible for a first-option to win titles in the first place! Without secondary All-Stars and role players, winning is nearly impossible, even for the highest-peak stars. So when totaling career value, these secondary seasons are important for both lower peak players like Miller and for pre/post-prime stars. As I put it when discussing CORP for Nylon Calculus:

“First options” aren’t quite as earth-shattering as we might think. While many people rightfully believe “there’s no way Klay Thompson could lead a team to a title in 2015,” they overlook the equally important counterbalance: It’s unlikely Steph Curry could have led that team to the title without Klay Thompson…You want Klay Thompson on your team, even if you wouldn’t want Klay Thompson to be the best player on your team.”

When we list out all of the non-lead players that made dynasties possible, it becomes clear how these secondary stars can tally up CORP value that surpasses alpha dogs: Sam Jones and Havlicek, Pippen and Rodman, Ginobili and Parker, Parish and McHaleWorthy (with Kareem) and even Bosh with the Heatles. In the first 60 years of the shot clock, those 10 secondary stars played on 32 title teams. They made most (or all) of those titles possible for the megastars.

Since we typically devalue these kinds of secondary seasons, we dole out more credit to “first options” and thus overestimate their worth. I certainly did before going through this exercise.

Defense, defense, defense

For me, the other huge revelation after quantifying hundreds of seasons against each other like this was the importance of defense. Seven of the first nine players on the list are two-way big men, and while only Shaq and Kareem approached transcendence on offense, the defensive value of these players explains why all-around bigs were the most coveted asset in the sport’s history before they all wanted to play like guards. Height is super important in basketball, largely because it’s harder for small guys to impact the game defensively. (Of the top-20, only Nash, West and Robertson were under 6-foot-6.) It aids longevity too; as athleticism fades, height remains.

Adjusted plus-minus (APM) data suggests that the best defenders might carry 5 points of impact per game and the best offensive players can top 6 points per game in a given season. Before diving into historical data, my impression was that offensive players were way more valuable than top defenders. Even with a clear offense-defense asymmetry (shown below), that’s a claim I cannot reasonable defend anymore. Defense matters. A lot.

Most GOAT-listers out there evaluate offense with a fine-toothed comb but then slot players into really rough buckets on defense because of a lack of historical information. Yet guards racking up all-defense selections are rarely in the same league as dominant bigs. A significant chunk of the great defensive seasons on record come from big men; among four-year scaled APM peaks, 31 of the top 40 results (78 percent) were from bigs, including the top 11 players. This jibes with historical roles around paint protection, rebounding, pick-and-roll containment, etc.

Furthermore, it’s more likely for elite bigs to maintain defensive value from team-to-team when compared to offensive stars because high-end defense essentially fits everywhere. Lesser defenders can lose some value depending on scheme and their teammates.2 This is a huge deal on a scale like this, and it’s a reason why so many defensive studs (e.g. Draymond Green or Dikembe Mutombo) post giant plus-minus numbers over and over. Jordan was an elite wing defender, but did his defense really move the needle while carrying such a huge offensive load? Two decades of plus-minus data (and four decades of game-level plus-minus) make it unlikely that he was close to someone like Olajuwon on D.

Most historical analysis has been offensive-centric — biased by the box score and ball-watching — and most rankings seem to underestimate both the high-end impact of defense and the wide range of defensive impact between most players. The difference on defense between an all-league guard and an all-time level interior defender can be the same difference on offense between an MVP-level attacker and a fringe All-Star.

For Jordan, adding an additional point to his defensive valuations would give him a CORP of 42 percent! (Versus his current estimated GOAT peak of 31 percent.) But a quick thought experiment shows how dangerous this is. This would place Jordan near Patrick Ewing as a defender, who posts huge impact-metrics that are defensively oriented, so in theory I’d then have to bump Ewing up too, but then that brings him up next to Hakeem, so I’d have to bump Hakeem up and then Bill Russell shoots further off the charts. Until I sat down and stacked seasons up next to each other, these issues didn’t reveal themselves.

So about the GOAT list…

Using traditional position designations, this top-40 contained eight point guards, small forwards and power forwards, seven two-guards and nine centers. So it’s top-heavy with bigs — perhaps because size ages well — but otherwise positionally balanced. I’m also comfortable with the distribution of players across eras, as recent generations were slightly over-represented. If anything, I’m concerned about valuing the first 10 years of the shot clock too much; the game rapidly developed in the ensuing decade as more money poured into the sport and the talent pool expanded.3 Here’s a breakdown of the list based on the year each player entered the league:

From that last decade of draft classes, James Harden has the best chance to crack this list first. Russell Westbrook has a shot too, although the potential of Anthony Davis and Giannis Antetokounmpo make them likelier candidates to shoot up the list for me. And there’s likely a longevity giant or two lurking in the wings, but it’s harder to foresee which all-league performer will crank out another decade of value.

Looking back, this exercise highlighted both underrated careers (like Miller’s) and how absurdly valuable marathoners like Abdul-Jabbar and Malone were. It also demonstrated how painfully close many of these careers were; it’s more fun (and satisfying) to draw a clear line in the sand, but the differences in impact rarely seem to warrant it. Emphasizing order — who is 12th versus 14th — now feels hollow to me, and I’ve come away largely GOAT-agnostic. Plus, different criteria will produce radically different lists.

I give Jordan the best peak, but it’s not by a lot. LeBron has the best consecutive eight-season stretch ever. Kareem has the most valuable career relative to his era. And even Russell has a backdoor claim as the most valuable player of all time. Any of those four are great GOAT choices. I think (hope) that the career ranges that players fell into on this list are a good starting point for any all-time discussion or best player list moving forward.

So, some 80,000 words, 500 videos and 250 graphics later, I hope you’ve enjoyed this as much as I have. Thanks so much for reading and being open to something a bit different.

  1. For instance, it minimizes role players and All-Star seasons, which flies in the face of the value of the supporting cast. The actual CORP curve could still be refined slightly with more accurate modeling.
  2. From that same APM set, the variability of players with a season in the top offensive decile is 4 percent greater than top decile defenders, just short of statistically significant (with p at .05). Yet, there is slightly more variability among mid-level defensive seasons (30th to 70th percentile) than offensive ones, which is statistically significant, suggesting that the impact of lesser defenders varies more based on circumstance.
  3. For the record, the game was already filled with giant men by the ’60s, and by the beginning of the ’70s, it was the highest paying of the major American leagues. Even in 1947, the typical salary was comparable to the national median, and by the end of the ’60s, the minimum rookie salary ($10,000) was well ahead of median income.

Comments are closed.