I love Fantasy Baseball. I love playing, watching, and arguing about anything related to sports, but I don't think anything can beat my passion in managing my virtual baseball team.
The reason I said that is because while I won't win my league every season (very far from it), I believe I spend more time in research than anybody out there. I want to show my friends that I actually do know more about the baseball, and I will do anything to prove that.
Up until this season, I always had been very confident about my dedication and ability find the right daily matchups for my starting pitchers. My thoughts were, "I don't need to pay for Clayton Kershaw if I can properly bench my pitchers, and pick up the right one from the waiver each day." Therefore, last year, I filled up my roster with all those flawed, but strong K-BB guys like Collin McHugh, Mike Fiers, Matt Shoemaker, and Shane Greene. I knew these guys had issues, but as long as I could avoid their occasional blowouts on long balls, I should be able to manage my ratios and to take advantage of their strong Ks.
In truth, for some reasons, my plan didn't pan out well. Every time I somehow find ways to bench them when they threw 7-inning shutout gems but to start them for their 2-inning meltdowns. At the end of the season, everybody's season ERA was lower than what they put on during the starts on my team. You may be laughing at me, but I'm sure you too had some similar experiences if you ever tried to study matchups.
Over the winter, there were some changes in my personal life, and I started building a baseball win-prediction model for living to beat the Vegas Sportsbook's odd. I did back-test my quantitative model under different scenarios to improve its accuracy, and the most striking discovery to me in the process was that including the matchup parameters didn't help me to predict the outcomes better. For example, you would think a lefty pitcher would perform far worse against a lineup filled with right handed hitters. Therefore, it makes perfect sense to lower the pitcher's win contribution for that specific game than his season average, right? Not really. In every test, my model performed way better when I didn't care about the specific matchups.
It got me to think. Is it possible that the way we consider matchups could actually be overrated? My conclusion was "Yes."
Last year, the Blue Jays' offense had a historical season and scored 891 runs, which was 127 more runs than the second place Yankees. On the other hand, the miserable Braves ended up only scoring 573 runs. According to these numbers, it's a no-brainer to bench your pitchers against the Blue Jays but to start against the Braves.
Let's divide up the numbers little bit, however, before we jump to the conclusions. The Blue Jays scored 318 runs more than the Braves during the entire season, which means that they produced 1.96 more runs in each 9-innings (let's ignore the extra innings). So if your starting pitcher goes 7 innings against the Blue Jays, his run-allowed average should be 1.53 higher than against the Braves. In short, the pitcher in average will allow 0.76 more runs against league's best offense (or 0.76 lower against the worst offense) than average (Remember, if we exclude the unearned runs and include extra innings into the equation, the figure would be even lower than 0.76).
Now let's regroup. So now we know that we are really talking about less than a run difference by studying the matchup. The real problem is that our season isn't as long as we think. Each starting pitcher gets only about 30 starts a season, and he will face the league's best offense probably once or twice, and 2 games are not big enough sample to average out all the other unpredictable factors that influences the runs. The pitcher may allow groundball, walk, walk, and HR to allow 4 runs, but he also can net only one run if the sequence goes HR, walk, walk, and groundball. A single swing of a batter at the bases loaded situation can range the run from 0 to 4 instantly, and the umpire's single strike call can make a huge impact too. There are also non-game factors like wind or injuries. Therefore, we may know that we are adding about a run for each start against the Blue Jays, but we don't know if we are adding it to 0 ER 7 IP game or 5 ER 3 IP game.
The season ERA does mean a lot, but individual game's outcome depends too much on the outside factors other than the pitcher's skill set and the matchup. I don't want to discourage you if you love playing matchups, and I believe you if you claim to have a knack for picking the right games to bench players. You might know something to study more beyond the opposing hitters. If you ever wondered why you can't seem to get the matchup right just like me, however, then here is your answer.
My solution to this whole issue? I would keep start my players if I trust them. Indeed I'm just buying into their season averages rather than the individual game outcomes.
I have written about a similar topic in DFS basketball. Read more, if you are interested.