The Boston Orphans have a four-game lead in the AL East, but it's not going to hold up. They're a baseball team that wasn't built to go the distance. The bullpen's too thin, and not even Todd Helton's shoulders are broad enough to keep carrying this offense. I know this because I'm the Boston Orphans' manager, GM, owner, and only fan. They're going to fall apart, and it's going to be all my fault.
This is life in the WhatIfSports universe, an online sports simulator that lets its users build teams using any single season of professional baseball player from 1885 through the most recent real-world season. (The other major sports are offered as well, but baseball's by far the most satisfying, thanks to its century-plus of precise and meticulously kept statistics.) Does the thought of 1980 Rickey Henderson leading off while 1927 Lou Gehrig bats cleanup flood you with nerdjoy? Welcome home.
WhatIfSports has been one of my favorite things on the internet for nearly a decade. But it didn't turn into an obsession until I realized that it wasn't actually about baseball at all. Or more specifically, until I realized it was about forcing yourself to forget most what you love about baseball, and filling that void with numbers and strategies you only barely understand. That is, if you want to win.
Building a team is easy: just fill out your position players and pitching rotation, add in a few bench and bullpen players, and tinker with your lineup until opening day. Players are assigned salaries based on a combination of historical performance and popularity, and salary caps are imposed to prevent teams packed with too many Walter Johnsons and Babe Ruths.
If you're familiar with fantasy baseball, this is its obsessive, omniscient cousin. WhatIfSports doesn't just compare historical lineups and say which is better. It simulates the outcome of every individual at bat, weighing the statistical strengths of opposing pitchers and batters. I saw it as a chance to settle, in some small way, the kind of hypothetical-steeped bar fights about who history's greatest players really were, and to cheer on some personal favorites one more time. It seemed like fun.
I'd been handed my own digital Field of Dreams; instead of clearing out a cornfield, all I had to do was click my dumb trackpad.
That's about as much context as I had when I built my first teams seven years ago. I thought it was all I needed. I was and remain a staunch Baltimore Orioles fan, which meant stocking my squad with not just franchise luminaries like Cal Ripken, Jr., but also the genial scrubs of my youth, the Jeff Reboulets and Randy "Moose" Milligans. I remembered that my father had been a Duke Snider fan, so I tossed him in as tribute, along with Golden Age mainstays like Mel Ott and Mickey Mantle. I'd been handed my own digital Field of Dreams; instead of clearing out a cornfield, all I had to do was click my dumb trackpad.
And then I lost. And lost, and lost, and lost some more.
WhatIfSports games are simulated three times a day, and I began to feel lucky if I won even one of those. I was steadily and consistently dominated by teams stuffed with players I'd never heard of or considered—alien names like Addie Joss, more familiar but equally unexpected ones like Howard Johnson. A surprising number of my opponents started 1993 Tony Phillips at second base, even though he had been an outfielder that year.
It became clear pretty quickly that they weren't just playing the game better, they were playing a different game entirely. I was there to live out a giddy fantasy, they were there to methodically shred dreamers like me to pieces. Poor Moose never had a chance.
I had made the exact team I wanted, spent hours machining the perfect nostalgia engine, only to be met with the obvious reality that the simulated world of WhatIfSports is at its essence an algorithm, and that algorithms can be exploited. The way to win wasn't just to find your favorite players, or even the best ones; it was to find the most undervalued assets, the breakdowns in the system. Which is how I found myself at the WhatIfSports forums, a community devoted, at least in part, to doing just that.
There are 3,273 threads in the WhatIfSports SimLeague Baseball forums, with 47,226 individual responses to sift through. The topics range from evidently helpful ("How much does speed matter?") to arcane ("seems like pitchers HR % stats are not normalized") to, well, message-boardish ("Where are you from????"). It's here that I learned the intricacies of how the simulation engine works, how the outcome of every single plate appearance is adjudicated. I tried to learn about something called Log5 normalization, until I ran into passages like:
Here's the formula before the platoon adjustment:
H/AB = ((1.066AVG .934*OAV) / LgAVG*) /
((*1.066AVG .934OAV) / LgAVG + (1.066- 1.066*AVG )(.934- .934*OAV)/(1-LgAvg*))
Where, LgAVG = (.934*PLgAVG + 1.066*BLgAVG)/2
The 1.066 and .934 reflect the 53.3-46.7 weight in favor of the batter. The output of this formula is a .2502 chance of a hit, which WIS increases by 4.5% to .2614 since Babe has the platoon advantage. Park factors would also increase or decrease the final result.
Even though some (all?) of the math was over my head, I slowly shed the mistakes of teams past. I learned how to use the ballpark I played in to my advantage, saying goodbye to home-run-friendly Camden Yards in the process. I saw that Addie Joss was a deadball-era pitcher, and as such was considered by many to be a bit of a cheat, and felt better about losing to him so often. I resigned myself to the fact pretty much the only Baltimore Oriole worth drafting was 1988 Bob Milacki, who only pitched 26 innings but was virtually unhittable in them. He represented a solid return on investment.
I haven't looked back. I regularly draft players I'd never heard of, or more likely never liked in real life. I find them in the WhatIfSports draft center, zeroing in on the stats I want at the salaries I need before even looking at the names attached. I've never seen Duke Snider come up in these searches.
Every so often a favorite player will align with my team's needs, but it's usually a happy coincidence. I traded genuine affection for my teams for a pursuit of winning. Or at least, of not getting my ass kicked: Apparently even my lizard brain is only good for slightly better than break-even baseball. After seven years and 3,604 simulated games, I've only scratched out a lifetime win percentage of .534. That's the problem with perfection. Figuring out the right way to attain it is still no guarantee you'll actually get there.
The Boston Orphans are likely going to wind up one of those .500ish teams. And I can't help but think, when I look at AJ Burnett and Ryan Church and all the other players I couldn't care less about in real life, that I might be better off as a wistful loser than robotically mediocre.
It's a feeling that always lasts right up until the next draft.
Perfect Worlds is a series on Motherboard about simulations, imitations, and models. Follow along here.