Comparing athletes across eras - an exercise in futility
Often in sports commentating (professional and amateur) the topic of cross era comparisons comes up. Either over beers at the pub or online, the arguments can be fun or even heated (my eras best player is better than yours, yada yada yada). There are, however, hardly ever any winners in discussions like these. Both sides of the proverbial equation tend to stick to their guns, both sides have statistics to back their arguments (although cherry picking these can be rife) and both sides prove very difficult to be convinced of the merits of opposing views. Sounds futile doesn't it? But rest assured, there are ways in which analytics and research have made important developments which hardly get discussed and the future capabilities of technology could potentially settle these arguments going forward.
My favourite sport of all-time is basketball having played it throughout my youth and continuing to watch it in my spare time. In many a basketball discussion group online, I find the topic of era by era comparisons draw fiery and impassioned comments which, for the most part, are lacking of in-depth analysis but are full of emotion. I'll be one of the first people to agree that cross-era comparisons are difficult especially as the environment changes but perhaps there are better way to analyse than listening to subjective ramblings.
So what are some of these subjective thoughts I see? Some of the main ones include athletes not being as athletic in the past as they are now, the level of skill being quite a lot less, the best players in the past standing out because they played against untalented individuals and whatever else can be used to disparage what was basically greatness across different eras. Here was an interesting example which did the rounds on Twitter as it called into question the greatness of Michael Jordan (full link HERE)
Whilst you can argue back and forth across the era divide, the problem with all of these discussions is the lack of scientific method or more in-depth analysis that would remove the subjectivity from the equation. With that said, is there a way though to really settle the score in debates like this? I believe there are at least two.
One way I've seen this done (impressively) was in the David Epstein TED talk from 2014 (Are athletes really getting faster better stronger?) where his studies showed that had the 1936 Olympic 100 metre champion Jesse Owens had the same environmental conditions as a runner like Usain Bolt in the same race of 2008, Jesse would only be a stride behind the more modern athlete. That is quite insane if you think about it. We are shown record after record across athletics and other sports tumble and fall every year yet bio-mechanical analysis shows that in some sports, had the environmental conditions at least been the same, past athletes would still stand out in today's crowd. This is counter-intuitive at first glance and makes you rethink historical events (almost in the same vein as Malcolm Gladwell's Revisionist History does). But when you think about it some more it makes perfect sense. The thought that athletes are automatically superior today than 30, 50, 100 years ago assumes human beings have evolved. Evolution takes millions of years yet we see today's athletes heralded as gods and yesterdays tossed aside. Time to rethink that.
Are athletes really getting faster better stronger?
The other approach came from researchers from MIT in a 2011 paper titled "Methods for detrending success metrics to account for inflationary and deflationary factors*". At the time this was a new statistical approach was aimed at finding intrinsic talent in athletes regardless of era. The TLDR (too long didn't read) here being that discounting or inflationary factors need to be applied to different eras of a sport. For example, in the steroid-infused National Baseball leagues of the 1990's, achievements of players like Barry Bonds and Mark McGwire need discounting to compare them to a normalised standard. Likewise, in eras of basketball where competition was lower and the ability to score 30, 40, 50 points was easier, they too would need adjustments. The results of re-ranking player statistical achievements in baseball and basketball are available from ResearchGate (HERE) and it is well worth a read if you've the time but the detrending or inflating of statistical achievements would be welcome in various cross-era comparisons, at least for me.
Based on the above I would think it appropriate to rethink our subjective views on our favourite players. Not to say that comparisons aren't fun, but that there are much better ways to try to settle these debates and the results might not be what you expect.
Having said all this, I am interested in how data science and technology can play a part in providing a potentially more encompassing answer. In a future world of sports analytics, I imagine an environment that takes advantage of the advances in Artificial Intelligence and Machine Learning. Imagine a world where the body movements of past players are mapped from various archival footage and accurate depictions that inherit their tendencies, strengths and weaknesses are created. Imagine the accurate simulations that could be created when you do this. You could train and test the created simulacrums with how accurately they are able to simulate historical outcomes. The model/system is rewarded for more accurate results that play out as close to what happened as possible and unleashing this across the spectrum of players both past and present would create the ultimate comparison tool. I know critics would be quick to attack this as still having flaws but it would be very hard to argue with the data science (unless you actually had a time machine and taxied players across the time horizons).
In any case, the future of sports analytics is certainly going to be interesting from a variety of angles such as sports broadcasting, electronic gaming or even the world of sports gambling. The future of analytics (namely AI and ML) is going to make it possible for us to answer questions like this. The problem will be with getting the word out there to the casual fan or professional broadcaster. Teams across various sports are embracing advanced analytics, so why shouldn't we?