Some work has been done on this idea before, including this post on a Blue Jays blog that concludes the opposite: that R.A. Dickey (a notable knuckleball pitcher) is actually MORE effective with the dome closed in Toronto. The theory goes that because "higher humidity inside the dome makes Dickey’s knuckleball rise and become harder to hit, in much the same way that wind at his back does".
I sought to redo a similar analysis involving more than one pitcher and more than one stadium over years of pitches, so I took scores of descriptive pitch data from FanGraphs for:
- R.A. Dickey (Starter)
- 2008-2016 with Seattle Mariners, Minnesota Twins, New York Mets, and Toronto Blue Jays
- Steven Wright (Reliever)
- 2013-2016 with Boston Red Sox
- Tim Wakefield (Starter)
- 2002-2011 with Boston Red Sox
I then took box score data from Baseball-Reference to cross-reference which games were played in dome conditions vs open air. This wasn't as simple as simply labeling some stadiums as dome and others as open air, since some MLB stadiums are open or closed only part of the time. Chase Field in Arizona, Miller Park in Milwaukee, Safeco Field in Seattle, Minute Maid Park in Houston, and Marlins Park in Miami are all stadiums that have retractable roofs that can be opened or closed game-to-game.
This ultimately gave me 136 games in dome conditions and 5,831 games in open air for these three pitchers over 15 seasons.
Baseline
| # of Games | Total Batters Faced | Pitches | ||||
| Pitcher | Dome | No Dome | Dome | No Dome | Dome | No Dome |
| Dickey | 93 | 892 | 1,990 | 4,788 | 7,432 | 17,724 |
| Wright | 5 | 993 | 108 | 1,003 | 461 | 3,803 |
| Wakefield | 38 | 3,946 | 956 | 6,370 | 3,470 | 23,056 |
| Total | 136 | 5,831 | 3,054 | 12,161 | 11,363 | 44,583 |
Only 20% of total batters faced were faced in a dome, so we'll have to standardize any comparisons.
| Knuckleball % | ||
| Pitcher | Dome | No Dome |
| Dickey | 79.63% | 79.51% |
| Wright | 84.88% | 85.00% |
| Wakefield | 84.14% | 84.45% |
| Average | 82.88% | 82.99% |
The above table shows the percent of overall pitches thrown that were knuckeballs, and the difference between a dome and no dome is marginal.
Conclusion: Way more data for open air stadiums, but no significant difference in knuckeball use
Control
deGrasse Tyson's original question relates to the effectiveness of the pitch in each stadium type, but I'm going to go deeper and analyze whether the pitcher's control is affected, as well as whether the pitch type is less predictable to the batter.
| Wild Pitches / Pitches Thrown | Hit by Pitch / Batters Faced | % Pitches in Zone | ||||
| Pitcher | Dome | No Dome | Dome | No Dome | Dome | No Dome |
| Dickey | 0.35% | 0.22% | 1.11% | 0.98% | 48.57% | 51.23% |
| Wright | 0.65% | 0.32% | 0.93% | 0.90% | 47.62% | 46.41% |
| Wakefield | 0.52% | 0.30% | 1.67% | 1.29% | 51.06% | 54.84% |
| Average | 0.51% | 0.28% | 1.24% | 1.06% | 49.08% | 50.82% |
Wild pitches are almost twice as common in a dome, and batters are hit more often as well. The most telling comparison, however, is what percentage of pitches are in the strike zone. Over that many pitches (40,780 combined between the two starters), the increased accuracy in an open air stadium is highly significant for both Dickey (z-score of 7.08) and Wakefield (z-score of 11.48).
Conclusion: More control for the pitcher in an open air stadium
Predictability
Having less control on the pitch isn't necessarily a bad thing: due to the unique movement of the knuckleball, this may actually be an advantage for the pitcher. So I next looked at how predictable the pitch is to the batter:
| Swing % | Contact % | |||
| Pitcher | Dome | No Dome | Dome | No Dome |
| Dickey | 45.72% | 47.83% | 80.05% | 81.14% |
| Wright | 42.00% | 44.28% | 72.44% | 77.97% |
| Wakefield | 45.82% | 47.62% | 76.06% | 80.28% |
| Average | 44.52% | 46.58% | 76.18% | 79.80% |
Combined with the lower percentage of pitches in the zone, the lower swing percentage in a dome seems to indicate that batters are simply laying off bad pitches (that occur more frequently in a dome). The contact percentage is a better comparison: all three pitchers have a significant increase in the batters they face making contact in open air (z-score of 2.91 for Dickey, 6.82 for Wright, and 12.81 for Wakefield).
Conclusion: More predictability for the batter in an open air stadium
Outcome
These two slices don't support a uniform conclusion: pitchers have less control in a dome, but batters also can't predict the ball's movement as well in a dome. So it comes down to the outcomes of the plate appearances between dome and open air situations.
| Walks + Hits / Batters Faced | Strikeouts / Batters Faced | |||
| Pitcher | Dome | No Dome | Dome | No Dome |
| Dickey | 30.45% | 29.51% | 17.54% | 17.15% |
| Wright | 33.33% | 29.81% | 17.59% | 19.14% |
| Wakefield | 30.86% | 30.14% | 15.69% | 15.48% |
| Average | 31.55% | 29.82% | 16.94% | 17.26% |
These three pitchers both give up more walks and hits in a dome, as well as have less strikeouts. This would seem to imply that the knuckleball is more effective in open air, as deGrasse Tyson suggests. For the kicker though I went to advanced stats:
| Field Independent Pitching | ||
| Pitcher | Dome | No Dome |
| Dickey | 4.65 | 4.38 |
| Wright | 6.00 | 4.62 |
| Wakefield | 4.88 | 4.87 |
| Average | 4.77 | 4.65 |
The above table shows a side-by-side of xFIP, courtesy of FanGraphs:
xFIP (Expected Fielding Independent Pitching): An estimate of a pitcher’s ERA based on strikeouts, walks/HBP, and fly balls allowed, assuming league average results on balls in play and home run to fly ball ratio.
The overall weighted average difference is 0.12 xFIP, which translates to 0.12 runs per 9 innings (since this is an equivalent to ERA). This is a fairly small difference, and isn't statistically significant over all three pitchers (z-score of 0.16, p-value of 56.34%).
Conclusion: Slightly more effective in open air, but not to a statistically significant level
