Part of the exhilaration of managing a strat-o-matic baseball team comes from the freedom to make decisions that are unburdened by the constraints of tradition. Does your win expectancy increase by replacing Mike Piazza with a better throwing catcher in the ninth once the potential tying run reaches first base? Then send out a new catcher in the middle of the inning. Want to use your closer to escape a crisis in the seventh? Just do it. No idea is too crazy. Put a team together, implement your strategies, run everything through a 162-game season, and see what happens. Did you finish 101-61? Then you're a genius. Or was it 71-91? Then come up with something new and try again. It's a wonderful laboratory for experimentation.
As you attempt to construct an optimized pitching staff you might be tempted to ask all kinds of outside-the-box questions. What might happen, for example, if you keep a right-handed and left-handed 5th starter option and carefully choose one or the other depending on the day’s matchup? In the strat-o-matic world, you don't have to worry about keeping guys sharp and stretched out. Concerns about the happiness of the cardboard rectangles that represent the two hurlers will never enter your mind. But there’s still the question of whether the approach is worth a roster spot and, if so, how you go about choosing who makes the start on a given day.
To make the discussion more concrete, let's assume that the right-hander and left-hander under consideration are Rick Porcello and Drew Smyly. We'll further assume that their 2012 statistics are representative of their current talent level. Both pitchers had traditional splits in 2012. Porcello was better against right-handed batters with a .317 wOBA allowed compared to Smyly's .327. But Drew was much better versus lefties with a .293 opponent wOBA compared to Rick's elevated .380 which tied for worst among AL pitchers who faced at least 300 lefty batters. Using this data, Smyly should fare better against a lineup of all lefties while Porcello should do better against a lineup of all righties.
Of course, you'll never see a ML manager turn in a lineup of nine right-handed batters against a right-handed pitcher or nine lefties against a portsider. Look at the numbers and you'll see quite the opposite. The Angels had 1648 plate appearances last year against left-handed pitching and exactly 8 of those PAs were by batters hitting from the left side. The Athletics, under the guidance of platoon wizard Bob Melvin, sent up 61 percent left-handed batters against right-handed pitching and 73 percent right-handed batters against left-handed pitching. Overall, major league batters enjoyed the platoon advantage 55 percent of the time in 2012.
Trying to select the best starter against an opposing team involves trying to predict the lineups they’ll use against your candidate pitchers as well as predicting the distribution of outcomes for each batter/pitcher matchup. This can be done, but to keep things simple we’ll try a more global approach and see where it leads us. First, we’ll propose a way to divide the American League into Porcello teams (P-teams) and Smyly teams (S-teams) using 2012 numbers. Then we’ll see how things would have gone if each guy only pitched against the teams in his group. A more detailed game-by-game analysis would likely perform better, but this global method should allow us to assess if a 5th starter platoon concept has merit.
A club is a good candidate for classification as a P-team if it used a small fraction of left-handed batters against right-handed pitching in 2012. Let L represent this fraction for each team. Similarly, a team would make a good S-team if it used a small fraction of right-handed batters against left-handed pitching. Denote this fraction as R. We can now use the difference, D = L – R, as a simple measure of whether we should start a lefty or righty against a team just based on how often they achieved the platoon advantage against each kind of pitching. Using this approach, teams with the smallest values of D are the best choices as P-teams and teams with the largest values of D are the best choices as S-teams. If we rank American League teams using D we find that the Indians with their heavily left-handed lineup were the most obvious S-team in 2012 and the Angels with their heavily right-handed lineup were the most obvious P-team.
As a small-sample sanity check, let’s consider teams in the AL Central. Cleveland, Minnesota, and Kansas City had the 1st, 3rd, and 5th largest D values respectively in the AL and therefore are definite S-teams. Last year Mr. Smyly faced 151 batters on these three division foes and allowed an excellent .241/.300/.336 batting line while posting a 2.45 ERA. In his outings against these S-teams, Rick Porcello allowed a horrific .342/.389/.498 batting line to his 244 batters while posting a 4.73 ERA.
The other AL Central team, Chicago, ranked 10th of the 13 non-Detroit AL teams according to the D measure which renders the White Sox as a clear P-team. So how did our pitchers fare? Porcello faced 105 White Sox batters and held them to a miniscule .190/.229/.350 batting line while posting a 1.98 ERA. Smyly, on the other hand, faced 47 White Sox batters and was knocked around to the tune of a .286/.326/.500 line while posting a 5.56 ERA. So far, so good.
Now let’s extend the analysis to the entire American League. A threshold on D is required to define the partition, but for simplicity we’ll divide the league into seven P-teams and six S-teams with Rick getting an extra team due to his seniority. If you like details, Seattle, New York, and Oakland round out the S-team set beyond the previously mentioned AL Central entries. Against American League S-teams, Smyly allowed an outstanding .213/.281/.315 batting line over 197 batter’s faced with a 2.59 ERA. On the other hand, Porcello allowed a much worse .329/.374/.498 line over his 350 batters faced while posting a 4.76 ERA. Conclusion: Start Smyly against S-teams.
For the American League P-teams, Porcello allowed a respectable .274/.318/.398 line (356 PAs) with a 4.30 ERA while Smyly was somewhat worse, especially in the power department, with a .267/.320/.491 line allowed over 182 PAs with a 4.81 ERA. Conclusion: Start Porcello against P-teams.
So what might we expect from a two-headed monster in the 5th rotation spot if the 2012 numbers carry forward? If we average the batting lines that we get from letting Porcello pitch to AL P-teams and letting Smyly pitch to AL S-teams then we get .244/.300/.357. An appropriately weighted average would probably look even better since three of the six S-teams are located in the AL Central and Drew was better against S-teams than Rick was against P-teams. But let’s see what we can say about the simple average. Felix Hernandez recently signed a deal worth 175 million dollars after allowing .241/.296/.333 last year with a .249/.312/.374 split away from pitcher-friendly Safeco. Thus the Porcello/Smyly combination lands neatly between the King and the King without his court. That's royalty in the fifth starter spot, you could say, for much less than a King’s ransom.