Home Field Advantage Is Dead. Long Live Home Field Advantage by Ben Clemens August 19, 2020 Empty stadiums are hardly the weirdest thing about baseball in 2020. There’s the shortened season, the universal DH, the runner on second base in extra innings; if you’re looking for ways the game has changed, there’s no shortage. Today, however, I’d like to talk about those empty stadiums, and their effect on home field advantage. A quick warning: this is going to be an article full of dry tables and plenty of math. I think it’ll be worth it, though. One question looms over everything else when it comes to home field advantage: what percentage of games does the home team win? Over a very long horizon, everything else is just noise. In 2019, for example, home teams won 52.9% of the games they played. In 2018, that number stood at 52.5%. Long-term home field advantage bounces around between 52% and 54%. It’s good to play at home. How about this year? To look at 2020 data, we need to do a little manual work. So far this year, four teams have played “home” games in opposing stadiums: the Marlins, Blue Jays, Yankees, and Cardinals. The Orioles also played part of a suspended home game in Washington against the Nationals. In all forthcoming analysis, I’ve removed those games from both the home and away datasets used in this article. It’s never exactly clear what home field advantage is measuring — rest, comfort, the crowd, umpiring, or some mixture — so games with nominal home teams playing in away stadiums are best ignored for these purposes. With that caveat out of the way and those games excluded, home teams have won 50.6% of their games through Monday, August 17. At the broadest possible resolution, home teams are winning a lower percentage of their games this year. Maybe the crowd really is king. That’s wildly insufficient for our purposes, however. One of the key tenets of baseball analysis is that merely looking at wins and losses is usually insufficient unless your sample size is enormous. Normally, I’d suggest using Pythagorean expectation here to guess a record. That doesn’t work when looking at only home games, however, because home teams skip the ninth inning when ahead. In 2019, for example, home teams were outscored on the year. This year, they’re scoring slightly more runs than their opponents. We’ll need something more granular than Pythagorean record to find a result. Fortunately, we can simply look closer. Take home and away strikeout rate, for example. From 2010 to 2019, home batters struck out in 20.3% of their plate appearances. Road batters struck out in 21.2% of their plate appearances. The talent level of both groups was nearly exactly identical, and yet they performed differently. Walk rate is the same — 8.4% for home batters, 7.8% for away batters. In nearly every offensive statistic, the home team has an edge: Event Frequency By Home/Away, 2010-2019 Split BB% K% HBP% HR% AVG OBP SLG Home 8.4% 20.3% 0.9% 2.9% .257 .325 .416 Away 7.8% 21.2% 0.9% 2.8% .249 .314 .401 Home Ratio 1.076 0.958 1.025 1.034 1.032 1.037 1.039 If you’re wondering why this is the case — why home batters get hit by pitches more often at a similar rate to how much more often they hit home runs, for example — this isn’t the article for you. These are merely useful baselines against which to compare 2020’s strange version of baseball. I wasn’t sure what to expect from 2020, but the data look surprisingly similar. Here’s that same table for 2020, with the aforementioned games removed: Event Frequency By Home/Away, 2020 State BB% K% HBP% HR% AVG OBP SLG Home 9.5% 22.6% 1.3% 3.7% .244 .323 .427 Away 8.8% 23.8% 1.2% 3.4% .238 .313 .405 Home Ratio 1.081 0.95 1.104 1.105 1.025 1.032 1.054 That’s right: nothing meaningful has changed. At a runs scored level, virtually nothing has changed. From 2010 to 2019, home teams scored 6.5% more runs per plate appearance than visiting teams. In 2020, that number stands at 6%. At the plate appearance level, 2020 looks like an unremarkable season for home field advantage, lack of crowds and all. We can still go deeper. Why do home teams perform better in virtually every offensive category? One theory is that umpires subconsciously favor home teams with unequal strike zones. If a borderline pitch gets called a strike at different frequencies for the home team and away team, that value will add up over time. Turning a strike into a ball means more hitters’ counts and thus better production on balls in play, more walks, and fewer strikeouts — sound familiar? To investigate this, I came up with a test. I took every single pitch thrown with two strikes between 2015 and 2019 where the ball touched the border of the regulation strike zone. In other words, a borderline pitch that would either be a strikeout or a ball if the batter didn’t swing. I further refined this down by removing all swings, then split the data into home and away. Over those five years, home batters faced 11,363 such pitches. 52.2% of the time, they were called strikes. Visiting batters faced 11,863 pitches, roughly the same amount. Those pitches were called strikes 54.3% of the time. That’s 250 extra strikeouts and 250 fewer continued plate appearances, and that’s only on pitches that actually touched the border of the zone on two-strike counts. The same pattern held true on the first pitch of an at-bat, though with a decreased magnitude; 71.2% of such pitches were called strikes for away batters, against only 70.5% for home batters. With three balls, those numbers were basically the same: 71.3% for away batters, 70.6% for home batters. Across all counts, umpires called borderline strikes at an increased rate on visiting batters. It’s still early to make sweeping conclusions about the same data in 2020. With two strikes, for example, home batters have only taken 332 pitches that clipped the boundary of the strike zone, while away batters have taken 320. The sample sizes are too small to be certain, but there’s interesting evidence of a change. Home and away batters have both seen those pitches called strikes at a 58.6% clip — no advantage for the home team here. On the first pitch of each plate appearance, away batters have taken just over 1,000 borderline pitches. 77.2% have been called strikes. Home batters have taken 1,077 pitches, of which 77.4% have been called strikes. With three balls, away batters see strikes called 77.1% of the time, while home batters see them 76.4% of the time, though both samples are tiny. Is this evidence that umpires are calling a truer strike zone without crowd noise? Perhaps! It’s too soon to say using this technique, because there simply haven’t been enough pitches. Painting with a broader brush introduces separate uncertainties, replacing sample size with sample representativeness. Absence of evidence is not evidence of absence. Still, there’s an interesting dichotomy at play here. As best as we can tell, umpires are calling the strike zone roughly evenly between home and visiting teams without crowd noise to sway them one way or the other. Despite that, however, measurable outcomes at the plate appearance level still favor the home team. What can we take away from this? Though it’s a bit of a lazy answer, it’s fair to say that more study is needed. There are other ways to measure marginal umpire calls, though all suffer from the same main issue: without a more robust sample of borderline pitches, there are simply too many factors at play that can confuse the signal leaking through. Framing? Pitcher, batter, and umpire identity? Those are each too large of an effect to be confident in the data at this sample size, which means more work on my part is necessary to control each of those factors. I simply don’t have a model that does so right now. Even without doing so, however, I’m comfortable saying this: the umpire advantage given to home batters might decline somewhat (or in total) without a home crowd around. Despite that, home teams still have the edge. Perhaps the other factors that often come up when the reasons for home field advantage are discussed — rest, comfort, familiarity, and so on — are bigger than we realized, or perhaps in a month one of these two opposing data points will give somewhat. Either way, I’ll be watching with great interest.