Each season has a wunderkind—a Jacob Wilson or Paulo Vitor Damo da Rosa that tears up GPs and PTs while juggling college exams, and the combination of their youth and good performance is a common touchstone. Every once in a while there’s also a notably older player, one who left the game and returned. But the vast majority of pros fall between the ages of 21 and 35, and while they’re in that range, their age rarely earns even a mention.
If you look at other professional competitive pursuits, physical or mental, you’ll see that this makes Magic an outlier. This is a well-understood topic in many sports, and for obvious reasons—the effects of aging on a baseball player’s reflexes or a basketball player’s jumping ability are obvious. The impact of aging on mental contests might be a little less clear, but it’s not an undiscussed topic like it is for Magic. A Google search for “’peak age’ chess” gives over 6,000 results. For “’peak age’ poker” almost 5,000. For “’peak age’ ‘Magic the Gathering’,” not even 100, and none of them actually relevant to the topic.
This struck me as an area ripe for analysis, but I quickly ran into two barriers: One, unlike baseball or chess, there are no established metrics of player performance for Magic pros, or at least none publicly available. There is no database of win rates at pro events, or even a comprehensive database of Pro Points. Some progress is definitely being made in this area. Chris Mascioli ran this fundraiser for a database of all matches played on the Pro Tour since 1999, which will hopefully be coming soon, and a couple websites with PT history have cropped up in the last month or so, though judging by player reaction, there seems to be some uncertainty about their accuracy. Regardless, for the time being, there’s no comprehensive performance measure, which makes determining how performance varies with age much more difficult. Secondly, finding the actual birthdate of a given player isn’t easy either. If I want to know when a baseball player was born, it’s trivial. The same cannot be said about professional Magic players.
Still, this is a topic I find really fascinating, so I decided to tackle it anyway, and try to work around those barriers with a combination of shortcuts and brute force. For the basis of my performance measure, I used the Pro Point standings at the end of each season.
Now, there are a few issues with this. One, Pro Points are not a perfect measure of skill or performance. Players travel to different numbers of events, and with the advent of the new World Championship, even players with identical travel willingness might have different opportunities to earn points. That said, it’s the best data available. It’s just important to keep in mind exactly what’s being measured.
Secondly, as I mentioned above, these aren’t fully reported for each year. Most seasons can be found on some prior iteration of the official Magic website, but the majority of them only go down to players with at least 10 points. If you’re only concerned with the Player of the Year race, this might make sense, but for our purposes, this is just incomplete data. Again, there’s nothing to be done about this that wouldn’t require months of work, and so we just have to acknowledge that shortcoming. Standings for 1997 aren’t available, and the points system wasn’t fully worked out in 1996, so I decided to start in 1998. In sum, this analysis starts with the Pro Point totals from 1998 through 2014 for every player with at least 10 points, which I think is the best that can be done at this time.
Finding ages is where brute force came in. Once I had the Pro Point data, I pulled out every name that showed up in at least two years (for reasons that will be clear later). I took those 695 names, and I did a whole lot of Googling. By far, the most common source of player ages were Top 8 Profiles, so thanks to whatever coverage reporter decided to ask that question first. Other sources included the short-lived pro player cards, tournament reports, and even the occasional local newspaper article. Eventually, I had a birth year for 576 of the 695 names, or over 80%.
Unfortunately, the players with available birth years were not chosen at random. As a rule, the more successful the player, the easier it was to find their birth year, as they were much more likely to have a Top 8 profile or pro player card with that information. Similarly, information was easier to find for recently active players than for those active in the early days of the Pro Tour, and for American players than non-Americans. These are definite biases, and again, should be kept in mind when looking at the conclusions of this piece. However, I don’t see any reason for the impact of age to have changed substantially from 1998 to 2014, or for American players to age differently than non-Americans. The conclusions of this analysis should be valid, recognizing that they’re drawn from the history of successful pros.
After all that preparation, I was finally able to do the actual math, using a technique lifted from baseball analysis known as the delta method. I took every player who had both an age listed and Pro Points for two consecutive years, and calculated the difference in performance between the first year and the second. Then, those players were sorted into buckets by age pair, so that, for example, the change in Kai Budde’s Pro Points between his age-21 and -22 seasons (2000 and 2001) was lumped in with the change in Christian Calcano’s pro points between his age-21 and -22 seasons (2010 and 2011). Using that, the average change for each age pairing could be calculated. The following is a table with the number of valid pairings in each bucket.
Unsurprisingly, the samples are most robust in the 20s, but there are at least 10 paired seasons starting at 16-17 and running all the way through 34-35. I’m going to use those as the endpoints of the curve, but the conclusions will probably be strongest for the 18-30 range, where there are at least 50 season-pairs in each bucket.
The other nice thing about the delta method in this case is that it limits the impact of changing point structures and the like. Since a player-season from 2000 will only ever be compared to 1999 or 2001, the differences in point “meaning,” so to speak, will be relatively slight, whereas comparing 2000 to 2010 or 2014 would be a lot more methodologically iffy. In an ideal world, there would be an available performance metric that didn’t change from year to year, but since there isn’t, this is the best that can be done.
Enough talk! Using all the above information, what does the aging curve look like?
This chart starts with the peak performance year (age 22) set at 0, and the value for each other age is the estimated distance from 22. For example, the average change from a player’s age-22 to age-23 season is -0.9, so 23’s value is -0.9. The average change from 23 to 24 is -2.1, so 24’s value is -3.0.
The first thing I notice is that this looks essentially like one would expect it to, and while this gets into confirmation bias territory somewhat, this is very reassuring. If the result was a line of random-looking fluctuations, I’d be much more worried that this analysis wasn’t picking up any real information. Obviously, given the errors and data problems discussed above, picking apart the exact shifts isn’t worthwhile (e.g., don’t conclude that players get much worse from 17 to 18), but the general shape seems very informative.
What I see is a fairly quick increase from 16 to 20, between 9 and 10 pro points in 4 years, followed by a plateau from 20 to 23. The values for those 4 years are all within 1 point of each other, suggesting that there is very little age-related change in performance for players in their young 20s. Following 23, there is a slow and moderate decline to 32 (between 10 and 11 points in 10 years). After 32, the decline appears to steepen, but that is getting into the age pairs with smaller sample sizes, so I’m not comfortable concluding too much from that.
I think the most interesting thing to take from this chart is the size of the overall effect. Each year’s value from 16 through 32 is within 10 Pro Points of each other, and while that is a lot of Pro Points, these changes are happening over several years, suggesting that aging is not a huge factor in player performance. This might not be a huge surprise—Magic is a game of the mind, not the body.
Also interesting is the long peak from 20 through 23, but I think that might be misleading. Pro Points are not merely an indication of skill, but of opportunity. The likely single largest determinant of Pro Point totals is player skill, but close behind that is the ability or inclination of a given player to travel for tournaments—the best player in the world can’t get Pro Points without attending tournaments. I would guess the ability to travel increases sharply as a player turns 18 and becomes an adult, and the inclination to travel decreases as a player leaves their early 20s and enters a phase of their life in which stability is valued more. Separating performance from willingness just isn’t possible at this time, so keep in mind exactly what’s being measured here.
Finally, I have one last note. I spent a lot of time in this article talking about the caveats that need to be made around this analysis, and the potential sources of error, and the reasons this might not be answering exactly the question you want it to. I do that because I think it’s very important to be up front with all that, and to be clear that this is not the end-all, be-all on this topic. This is, however, the best we can do with this data. It might not be perfect, but it’s better than nothing, and my hope is that seeing what can be done with detailed data will push both Wizards and other individuals to record that data more frequently.
As always, reach out to me on Twitter @henrydruschel if you have any questions or comments, or if you are doing a project for which could use the birth years of almost 600 pros, because I’ve got you covered on that front.
The incompleteness of the dataset I’m using means you really shouldn’t take this chart too seriously. For each year and category, this is the number of Pro Points awarded to players with an age in that category divided by the total number of Pro Points awarded to players with a listed age, so there are lots of missing Pro Points and potential biases. That said, I thought this was really interesting—pro Magic looks like it’s becoming an old person’s game.