I think I messed up my analysis of Justin Gary in my Hall of Fame Ballot article last week.
This is really a story about "median finish" and its role in judging Pro Tour careers. In my previous article, I explained that I thought Justin was the kind of guy who would be made to look good by this stat, and when it didn't really distinguish him, I decided I probably wasn't voting for him. I then tried to find a different stat which could back up my intuition that Justin had consistent success as a player but was a couple of unlucky breaks away from having a lot more Top 8s on his resume. The other potential stat I came up with was Player of the Year finishes, and Justin looks consistent and good when analyzed in this light, but he only cracked the Top 10 once so I decided he wasn't truly great and moved on.
Before I get any farther, I feel obliged to point out my own potential biases. I have always said that the best players in the game are the ones who can consistently finish near the top, not necessarily the ones who actually win. I'm more impressed by two Top 16s than I am by a Pro Tour win, for example. There's a chance that I only think this way because I personally put up a bunch of Top 16s while only breaking through to the Top 8 once. I don't think that's what's going on, but it's obviously really hard for me to know for sure.
Similarly, when someone first proposed using median finish as a way to try to suss out who the good players were, I remember thinking "Yeah! That's exactly the kind of stat we need!" This was in the late 90s and the analysis was all done using a webpage of top finishes that was maintained by Elf (a.k.a. Michael Feuell). In my memory, I decided I liked the stat before I learned that I have the best median finish in the game's history. When I suggested that median finish be added to the spreadsheet of stats that we worked up to analyze Hall of Famers I was almost certainly aware that this would make my own eventual candidacy look good. That said, I really don't think I was motivated by self-interest – I really do think we need a stat that points out who is consistently doing well, as opposed to occasionally doing great. Median finish was the best one I knew of so I suggested it. Anyway, you should all take everything I say here with a grain of salt, think through my logic yourself, and make up your own mind about my potential biases.
While I initially thought median finish was an awesome stat, the more I look at it the less useful it appears. The problem is that it clearly goes down as your career gets longer. The Hall-eligible players with the most impressive median finishes also all have the shortest careers. (Mark Justice has the best median finish on the ballot and also the shortest career. I have the best median finish if you add in all future classes and I also have the shortest career amongst that group. And so on...) We all know that Finkel and Kai are the two best Magic players of all time, so if this stat was really a useful tool in picking out the best players, why aren't they up at the top of the list?
It seems clear that the way to have a high median finish is to quit playing while still in your prime. Finkel kept coming to Pro Tours long after he stopped practicing 24/7. He was still plenty good enough to keep himself qualified, but he was not good enough to make, say, Top 32 in half his events (much less put up 8 consecutive Top 23s like he did from 1997 to 1999). Similarly, if Justin Gary had retired when he started law school in fall 2003, the record would currently show that his median finish was an incredibly impressive 26.5 (the best across both Hall of Fame classes). Instead, he continued to come to Pro Tours mostly to hang out with his friends and those two twilight years resulted in an actual median finish of 58.5.
The idea that Justin somehow became less Hall-worthy because he played those extra dozen Pro Tours is absurd. Clearly I messed up my logic somewhere and median finish just cannot be used to prove anything without a detailed understanding of context. Maybe we should just throw away the whole idea that median finish might have something interesting to tell us, but I'm not sure I'm ready to go that far because that 26.5 that Justin put up in his prime is exactly the kind of stat I was fishing for to be able to defend a vote for him. Justin played 33 Pro Tours between 1997 and 2003 and he finished in the Top 32 at 19 of them. That's damn impressive and I don't care that he only made the Top 8 of three. Now the question for me is how do we develop a statistic that measures what median finish is trying to show but does not punish people for having long careers?
For that, I turn to baseball. The distinction between peak value and career value is common when it comes to analyzing baseball players, and it's not that one is more important than the other – they just measure different things. Who was a better pitcher, Sandy Koufax or Nolan Ryan? While Ryan's career totals are off-the-charts ridiculously awesome, everyone agrees that Koufax in his prime was significantly better than Ryan in his prime. (Koufax's career was then cut short by injury.) It's not really possible to say who was "better" in absolute terms so baseball pundits settle for saying than one had a higher career value while the other had a higher peak value.
I think we have some nice ways of measuring career value in Magic – lifetime Pro Points and lifetime winnings both speak to that. Lifetime Top 8s is also interesting, and it has clearly become a very glamorous statistic to quote, but I feel like it doesn't do justice to the Justin Gary's of the world. Starting with Worlds '97 (he made his PT debut at that event as a young "came out of nowhere" US National Champion) he put up the following finishes: 23, 28, 242, 30, 258, 6, 14, 19, 16, 9, 26, 18, 27, 56. So he only missed Day Two in two of his first 14 Pro Tours and he rattled off nine money finishes in a row, including 8 consecutive Top 32s. Wow. But if you go by Top 8s as the ultimate measure of accomplishment, he has just one to show for that run.
Top 32s or money finishes could be a stat to track more carefully, as we do with Top 8s. But that doesn't address the concerns of how good a player was at his peak. Maybe "percent of Pro Tours finishing in the money" is the answer, but that stat is once again going to punish players who keep playing past their prime.
The attempt to develop a peak-value statistic forced me to look for a canonical length designed to span a player's top performances. How long does a Pro Tour player's prime need to be to make him worthy of a discussion as one of the game's best players? Ten events? Fifteen? Twenty? Twenty seems like too many – baseball uses five years as the usual unit of time for discussing peak value and typical career lengths for baseball stars range from 15 to 20 years. Magic players' careers are a good bit shorter, so four years (with a typical season having five events) is probably too long. Two years (10 events) is probably too short, but I'm not positive of that – for a player like Olle Råde that actually covers more than half of his career. Fifteen events feels right to me. For players with short careers (like Råde at 18) this would cover almost their entire career, but for players who have played 40-50 Pro Tours, it lets us focus on just a third.
Taking all that into consideration, I propose "median finish for their best run of 15 consecutive Pro Tours" as a measure of peak value. Applying this metric to last year's inductees shows us the following breakdown:
I'm not sure that this proves anything, but it does at least pass the sanity check of having Finkel in the top spot and with a nice lead. Comer has seven finishes in the top 27, then a 50th, and then seven finishes at 63rd or worse during his best 15-event stretch so his resume takes a hit according to this stat; but I'm not sure that's actually a problem. This stat isn't supposed to measure overall value, just peak value. Comer's candidacy was much more about career value (and deck building), with Top 8s scattered across seasons 2, 3, 5, 5, and 6.
Justin Gary, meanwhile, has a peak-median finish of 25. During his prime, you could find him knocking on the door of the Top 20 in over half his Pro Tours. On top of all that, his career totals are also solid (4th on this year's ballot in both career winnings and lifetime pro points).
I'm not sure what exactly to think of peak-median as a statistic at this point – it seems to me to have some decent potential but I'm quite curious to hear what other people think of it. One thing I am sure of, now anyway, is that Justin Gary is an excellent candidate for the Hall of Fame and that I should have voted for him when I submitted my ballot last week. Sorry Justin ... hopefully this article can help other people see your career a little more clearly.