While the initial results seemed skewed, the 2020 NBA Draft Combineās data has mostly turned out to be in line with prior years on statistical testing. Combine results are difficult data in general, but for 4 of the 5 measurements, itās safe to proceed as normal.
The NBA Combine, in general, is very difficult to get meaningful information out of. Higher is better, sure, but the things tested donāt exactly translate to what occurs in the game, so certain players like Grayson Allen or Jordan Bone, whose athleticism is less functional in-game, get an undue boost in perception.
This year, however, was even more exceptional than normal. Numbers came in in a slow trickle, and one of the first complete sets of numbers we got were those of Zeke Nnaji, who put up this line:

Those numbers are, across the board, amazing for a guy his size. The shuttle time is especially amazing, but the running jump and sprint are also in the top five percent for guys of comparable size. But it wasnāt just Nnaji. We also got an early report of Cassius Stanley putting up a 44-inch vertical, and while Tre Jones and Daniel Oturu werenāt as extreme of outliers, they were both above average across the board for their position group.
This slow trickle of results weeks after they were expected created some hesitance to accept the results, even from high profile sources.
Will just note that I have some general skepticism about some of the NBA Draft Combine numbers floating out there. Not that theyāre being reported wrong, but more about the methods of measurement and the lack of singularly consistent entity doing the measurement.
ā Sam Vecenie (@Sam_Vecenie) November 2, 2020
As more numbers trickled in, that general consensus sort of lingered around, which wasnāt without cause. We were still seeing seemingly high numbers across the board and from multiple different players. But with everything in, we can show that mathematically, with one key exception, the measurement this year was consistent on average with prior years and trends.
How does the NBA Draft Combine data from this season compare with previous years?
To show this, weāre going to start with the lane agility time, because the process will be similar for all the individual measurements. So first, weāre going to build a base model for lane agility using purely physical traits about the player. i.e. we know that as a general rule, taller players are slower, and heavier players are slower. Based on an initial pass, Iām also going to include variables for wingspan, to control for a shorthand for body type, and the square of height, since it appears that the effect of height is quadratic rather than strictly increasing, since really thereās not too much cost to height for players between 6-foot and 6-foot-6. That gives these results, which is a solid base to build into:

So now, we add a dummy variable thatās equal to 1 if itās this year and 0 elsewhere, and see what that does to the regression and get these results:

This shows that a player of the same general physical profile in 2020 is likely to be faster on the lane agility drill than those from other years. So, results in, combine untrustworthy! Except, pause for a second and consider what would happen if players had been getting faster over time in general. This wouldnāt be at all surprising, given that the league as a whole has been pushing out guys who arenāt fast enough for their position, and it bears out in the numbers.

So as we can see, there is a clear general trend that lane agility times have been decreasing in general. If, then, we capture that original trend and then add our dummy on top of it, as we do below, the dummy variable is not significant beyond what was already captured by the trend, which indicates that the numbers from this year are as trustworthy as they ever were.

So with a pattern for Lane Agility established, what about the other drills?
Well, the standing vertical jump is the same general pattern. There is an existing trend that players of the same build jump a little higher every year, and once you account for the trend, the dummy doesnāt matter, as seen below.

And with standing jump having been the same, itās unsurprising that the running jump also shows no impact beyond the existing trend.

The Ā¾ court sprint is a little different since it has no year-over-year trend in general (players sprint just as quickly today as they did in 2000), but the general reasoning is the same. First, weāll show that thereās no trend.

And since thereās no trend, itās no surprise that the dummy for 2020 doesnāt show up as significant. Measurements in this statistic for 2020 only were in line with no trend based on year, so even putting the dummy on its own shows no statistical significance either.

And that leaves us with the shuttle run, which Iāve intentionally left for last. This is because we actually knew ahead of time that the shuttle run was slightly different this year. To sum up the linked article, in a normal year youāre expected on the shuttle to react to a light that tells you both when to start and which direction to go. Since not every facility had access to this equipment, they did away with the lights for consistency same across the board. As a result, record after record was set. Twenty-four different players recorded times this year that wouldāve been the single best time ever prior to this year. Guys like Reggie Perry increased their time by over three-quarters of a second, despite this being a drill with times that normally range in the 2.5- to 3.5-second range. Which, in retrospect, should probably be a major red flag that Udoka Azubuikeās time was still 3.61 seconds, but nonetheless.
Walking through the same process with the shuttle run shows that, like with prior tests, there is an existing year-over-year trend. Unlike prior measures though, even once that trend is controlled for, thereās still information left in the year based on a one-year spike.

With that in mind, if you wanted to translate numbers from this year to other years, based on the coefficients I would add around .53 seconds, which leaves this class in much more standard territory. No one breaks Malcolm Brogdonās record, but Devon Dotson is still very quick and most players still fall right around the three-second mark.
In conclusion, while the early results did seem crazy, the combine results this year were actually trustworthy with the exception of the shuttle times, which is to be expected given the differences in the drill. This also tells us that at some level, NBA-level athletes are already getting a pretty significant percentage of their athletic potential out of their bodies, since even with months to train there wasnāt a large difference. Ultimately, the combine still has the same problems it always does, but outside of the shuttle run the data this year is just as trustworthy as it always has been.