clock menu more-arrow no yes

Filed under:

Skill vs. Luck

New, 88 comments

How much of success, or lack thereof, in close games can we chalk up to plain old luck?

Photo by David Banks/Getty Images

Baseball is random and weird. Take Max Muncy, who burst onto the scene after being released in 2017 to become one of the game’s best hitters. Or Erik Kratz, a longtime backup catcher who tallied five hits in the Brewers’ NLDS sweep over the Rockies despite failing to amass double-digit hit totals in the majority of his nine MLB seasons. Even entire teams can be subject to seemingly unexplainable fluctuation in performance. Since 2012, the Red Sox have alternated from worst in the AL East to first place and world champions the very next year, immediately followed by back-to-back dead last finishes and three straight division titles in the seasons directly after that. Of course, some of these otherwise bizarre occurrences can be attributed to fairly recognizable improvements such as Muncy refining his swing or Boston reshaping their roster through the acquisition of stars like Chris Sale and J.D. Martinez.

I’m not here to discuss those types of oscillations in success, however. Rather, we’re going to hone in on a sometimes charitable, other times unforgiving, but constantly present factor in baseball: luck. Luck is the 2016 Rangers setting an all-time MLB record with their unheard of 36-11 record in one-run games, only to flounder to an MLB worst 13-24 record in those same nailbiters the next season. I wanted to determine if there is any way to control ones performance in these tight games, or if they are merely the coin flips they so often appear to be.

I compiled data from all 450 individual team seasons over the past 15 years (2004 to 2018) and tested to find if any consistencies in one-run game performance existed. I compared each club’s winning percentage in one-run games from one season (Year X) to the very next (Year X+1) and plotted the results on the graph below.

As the messily scattered dots and virtually flat trendline indicate, there is evidently no method to the baseball gods’ madness when it comes to wins and losses in one-run games. In fact, the coefficient of determination (R2 value) between year one and year two shows that just 0.5% of all variation can be explained by the previous season’s record in close games, mathematical lingo for what essentially amounts to zero correlation.

Perhaps teams with an elite bullpen were able to escape with a disproportionately favorable record in these games? The Brewers led baseball in 2018 with 33 one-run wins and just 19 such losses. To what extent did a relief corps headlined by Josh Hader, Jeremy Jeffress, and Corey Knebel allow the Brewers to outperform their underlying metrics? I then compared each team’s record in one-run games with its bullpen stats, including park adjusted ERA, FIP, and xFIP in addition to FIP-based and runs allowed-based WAR per inning pitched. Remember that R2 values can range from 0 to 1, with a 1 signifying 100% correlation.

Bullpen Performance vs. Win% in One-Run Games

Bullpen Metric r^2 value
Bullpen Metric r^2 value
RA9-WAR/IP 0.066
fWAR/IP 0.065
FIP- 0.053
xFIP- 0.051
ERA- 0.048

None of these metrics even crack an R2 value of 7%. While bullpen talent is a marginally better predictor of win-loss record in one-run games than less scientific alternatives, there is yet to be evidence suggesting that teams can exercise great control over this performance. 12% of all teams I studied experienced a shift of at least 170 points in one-run winning percentage–or approximately four wins, assuming the average club plays roughly 23.2 such games per season–from one year to the next. In any given season, an estimated 40% of the league will deviate from its expected (calculated using a team’s runs scored and allowed figures) Pythagorean record by four games or more, in no small part due to random chance.

Over the past three seasons, the Cardinals have been a few extra one-run games breaking their way from securing what potentially could have been eight consecutive playoff berths in 2018. Just a little less good fortune in some past seasons, however, and St. Louis enters 2019 without any division titles this decade. From players to teams to entire eras, statistical variance in baseball is inevitable. The result can be skewed-by-luck win-loss standings that contribute to brash narratives surrounding teams and seasons. With a greater amount of more detailed and precise information available now than ever before, it has become increasingly important to differentiate between what is real and what is the byproduct of a facade created by luck.