As every Millennial knows, Backyard Baseball is the children's computer game in which neighborhood boys and girls play pickup ball with kid-sized incarnations of everyone's favorite major leaguers. The 2001 edition, for example, featured the likes of Mike Piazza and Barry Bonds, traveling across distance and time for the honor of running around a suburban lawn with such characters as Jorge "Bonkers" Garcia. To a kid of the early aughts—back when MLB could more innocently attach its license to children's fare (Hey, kids! You can play as Jose Canseco!)—the game was irresistible.
The best player in Backyard Baseball, pro or kid, has always been Pablo "The Secret Weapon" Sanchez. Pablo is a short Hispanic boy with a knack for pounding baseballs incredible distances, despite his slight stature. Born a Leo (Aug. 18), he spoke exclusively Spanish in all of the Backyard Sports games until Backyard Baseball 2003, when he revealed that he learned Spanish in school and was fluent in English the whole time. (Good thing John Rocker never shows up in this backyard.)
Because he is so much better than everyone else, and because he is competing against actual MLB stars, the game practically invites the question: What would Pablo Sanchez be like as a real-life professional baseball player?
We're going to try and answer that question here.
In the game, players are rated on only four categories (batting, running, pitching, fielding) on a precise scale ranging from 1-10 baseballs. Pablo's batting rates a 10, his running a 9, his pitching a 5, and fielding a 10. Curious as to how the creators of Backyard Baseball (specifically Backyard Baseball 2001) decided to rate the pros as children, I looked into how the players' video-game ratings matched up with actual on-field performance.
I started by tallying the numbers of the featured major leaguers in the three years prior to their appearance in the game, adjusted for playing time. Rather than using advanced stats like WAR and VORP, I looked at the type of measurements that I assumed the folks at Humongous Entertainment were using to judge players in the heuristic dark ages of the turn of the millennium: home runs, RBI, and batting average.
By regressing the Backyard Baseball ratings of the major leaguers against their real-life stats, I was able to figure out which ones mattered most to the programmers. (I go into more detail about my methods at the bottom of the post.) It turns out that only the player's batting rating was useful in the prediction. From there, we can extrapolate what an average season for Pablo would look like: 48 home runs, 128 RBIs, and a .649 slugging percentage. Based on the rest of his ratings, Pablo would also steal 21 bases out of 29 attempts per year, hit five triples or so, and win three Gold Gloves within the first 10 years of his career. I even figure he'd weigh about 180 pounds.
For the hell of it, I ran the numbers for five other Backyard kids, and projected their average seasons from 1998 to 2000:
Where would Pablo play?
Projecting his performance was pretty straightforward, so I went even further and tried to guess what position he would play. Pablo often played shortstop when picked in Backyard Baseball (I know I always put him there). I decided to narrow my first test to find his chances of being an infielder. The simple explanation is that I determined where in the field—outfield or infield?—an MLB player with Pablo's statistics would most likely play. The more complicated explanation is that I created a dummy variable on whether the pros were infielders or not, and then regressed it against all statistics that were significant predictors (stats indicating speed, as it turned out). I then substituted in Pablo's predicted numbers to find the probability of him being an infielder. The first test yielded only a 28 percent chance that he would be an infielder.
So even though he was a great shortstop in his youth, he would probably be an outfielder in the pros—I predict a center fielder due to his speed. An interesting side note: I poked around, looking to find some people who have put up similar numbers to those predicted for Pablo. The two closest comps I could locate were Albert Pujols and Alex Rodriguez. Sure enough, Pujols was a solid outfielder before making the switch to first base, and A-Rod was a premier shortstop before moving to the Yankees. Here are Pablo's numbers next to Pujols's and A-Rod's average seasons from 1998-2000.
|Pablo (projected)||Albert Pujols||Alex Rodriguez|
Was Pablo on steroids?
One thing I noticed when looking through the roster of 31 MLB players in the 2001 game was that 10 of them had been heavily implicated in the use of steroids—among them "No Way" Jose Canseco, Barry "Big Guy" Bonds, and Jason "The Slugger" Giambi. What does that mean for young, outwardly innocent Pablo? Would he have been a juicer, too?
Similar to my methods for finding his position probability, I created a dummy variable indicating the use of steroids. By regressing it against the relevant factors—HRs being the most significant indicator—I determined the probability that Pablo would have taken steroids: 90 percent, according to my calculations. (That means his numbers look an awful lot like those of the 10 juicers on the Backyard roster.) And given his proximity to Jose Canseco, that number should probably be higher.
So now that we've indicted our friend Pablo, let's find the numbers he would have put up if he did in fact juice. Earlier we found his average output, but that was without taking steroid use into account. Now, when I factor steroid use into his output regressions. I get the following table:
|Weight||177.21 lbs.||197.94 lbs.|
(The "clean" totals are what you get with the dummy variable equal to zero, i.e. 0 percent probability he would've done steroids; the "juiced" totals are what you get when it's one.)
One note: Steroid use did not directly improve RBIs or slugging percentage. Instead, it significantly affected home run total, which then led to an increase in both of the other two stats. In fact, when holding home run total constant, steroid use tended to have a slightly negative effect in both of those categories.
So: The pro Pablo would be a very successful center fielder, producing All-Star numbers year after year. And had he done steroids, as he likely would have given the crowd he was running with—peer pressure can be strong at age 12—he would have gained 20 pounds and produced at a level that would have beaten out Carl Yastrzemski in his Triple Crown season. (Pablo might've won the Triple Crown last season, batting average being the only uncertainty). But then again, his success would've been tainted by steroid revelations, and his emeritus years in the game would've been given over to legal entanglements and angry recrimination. Perhaps it's better that we remember Pablo Sanchez in his glory years, reigning over the backyard with his counterparts on the Crazy Wombats and the Mighty Melonheads.
* * *
Running: When analyzing running rating, I used statistics that would directly indicate foot speed, such as steals, triples, GIDP (number of double plays grounded into), and even weight. Not surprisingly, high steals and triples and a low GIDP were the main indicators of a good running rating. Though I expected there to be a strong negative correlation between weight and running rating, the former was not a significant predictor.
Batting: For batting rating, I used traditional mainstays like home runs, RBIs, and slugging percentage (which takes into account batting average). My regressions said that home runs and RBIs were not significant in predicting batting when slugging was included. It appears that the programmers were fair to all players—they didn't put too much emphasis on power when determining batting rating and thus didn't punish contact hitters for their lack of home runs. That said, slugging percentage was consistently the best predictor of batting rating.
Pitching: Estimating pitching proved futile. Since only two MLB pitchers were featured in the game—and both were rated a perfect 10 in pitching—there was no way to make any inferences based on pitching statistics. (Remember: All players are assigned pitching ratings, even the MLB position players.) The only variable that was significant in estimating pitching rating was whether the player was actually a pitcher (obviously). No other variables significantly explained the rating, which leads me to believe that the ratings for position players were chosen rather arbitrarily, but there was somewhat of an inverse relationship between pitching rating and the other ratings given to the players. Players with relatively high pitching ratings tended to be rated poorly across the board elsewhere. This is probably because Randy Johnson and Curt Schilling naturally had lower hitting stats than everybody else. The only exception was Alex "Little Mitt" Gonzalez, the lowest-ranked pro player in the game, who was below average to terrible at pitching and batting. Yeah, and fielding, too. In fact, the programmers rated him worse than 20 of the neighborhood kids and better than only nine of them, including, to his immortal shame, seven girls and a wheelchair-bound boy. No wonder he wasn't featured in subsequent editions of the game.
Fielding: Fielding rating was another difficult one to predict. First off, there tended to be a positive relationship between running and fielding rating, which makes sense, since a faster player is likely to have a larger range. But there are no traditional measures that offer an accurate portrayal of a player's defensive contribution, especially comparing across positions. To account for this, I used a statistic called Total Zone Total Fielding Runs Above Average Per Year. In English, this is the number of runs per year, above or below the league average, that can be attributed to a player's defensive contribution. This more accurate statistic proved to be statistically significant in predicting fielding rating, presumably because it gives a better understanding of a player's defensive performance.
The main indicator by a long shot, however, was number of Gold Gloves won, showing that prestige held a fair amount of weight when assigning fielding ratings. Also, raw number of Gold Gloves was a better indicator of fielding rating than was the number of Gold Gloves won divided by number of years in the league, which means that younger players were discriminated against because they had less time to develop a reputation as a good fielder.
Regressing is a numbers-minded column by our clever friends at the Harvard College Sports Analysis Collective, a student club dedicated to quantitative analysis of sports strategy and business. Follow HSAC on Twitter, @Harvard_Sports. If you have any comments or ideas for future columns, email them to firstname.lastname@example.org.