Evaporating Errors, Part 1
MLB fielders have gotten A LOT better at their job in the past five years - at least, according to one key metric.
In 2019, there were 2,900 errors across baseball, working out to an average of 1.19 per game. In 2024, there were 2,600 (1.07 per game), a more than 10% decline in 5 years. In 2023, at the recent nadir of this trend, there were just 2,500 (1.04 per game), a 13% decline.
 |
| Total Errors Per Year (2015 - 2024, AKA The Statcast Era) |
 | |
|
Note: Here on in this analysis, "Per Game" can more intuitively be read as "Per Play Ball!". Think of the denominator here not as the number of errors a given team will make in a game, but the number of errors you can expect to see when you show up to the park (in other words, all the errors committed over the course of the whole game by both teams).
Indeed, there was plenty of chatter in 2023 about what exactly what was going on with official mistakes in MLB.
Kyle Glaser over at Baseball America wondered if there was a larger push to be more charitable with base hits with an eye towards increasing batting average league-wide.
Andy McCollough and Zack Meisel at The Athletic canvassed the league for more anecdotal evidence about the new error experience.
The AP put a a year-end bow on the state of scoring errors as the league was "on track to finish with the fewest amount of errors in a non-shortened year since 1960, when there were 16 teams".
There was particular interest in the great, disappearing error because other leagues far and wide were not experiencing the same phenomenon.
To illustrate this, I looked at the error rate per game across the same time frame for three other leagues: the independent Atlantic League (ATL), the Korean Baseball Organization (KBO) and the Pacific Coast League (PCL), one of two AAA leagues in the MiLB ladder. Hardly comprehensive, I chose these three to get a look at both foreign and domestic leagues as well as circuits up and down the quality of play spectrum.
In the KBO, errors per game rose from 1.38 in 2019 to 1.51 in 2024. In the PCL, the rate over that time period was unchanged at 1.52 errors per game. In the Atlantic League, errors per game soared after COVID and after giving up those gains and more, still rebounded to exhibit a slight increase to 1.62 errors per game.
In the below graph, I've drawn a dotted line through the 2019 to better mark the before and after I just discussed:
 |
| Errors Per Game, Sampling Of Leagues (2015 - 2024) |
If the very game of baseball was changing - a new technology to physically make baseballs had proliferated, say - or if there was a mandate far and wide for baseball leagues to clean up their punitive ways, we would see more correlation in per game error rate. Instead, we see MLB on an island.
So if MLB was hemorrhaging errors, and other versions of baseball were, at the very least, not changing too much, then that leaves us to conclude it's something MLB-specific.
As an astute reader, you rightfully note - the MLB game HAS changed! Elevate and celebrate! Three true outcomes! RULE CHANGES GALORE! Looking at errors per game across leagues or even within the same league across mini-era hardly makes for an apples to apples comparison.
Let's try and normalize things, then.
Very roughly speaking, there are about 50 "batted ball events" (PA - SO - BB - HBP) per game in MLB. 1,000 BBE, then, is roughly 20 games. In other words, every 1,000 BBE represents an extremely rough assessment of a typical day's slate of games in MLB. As a heuristic, this can be a helpful benchmark.
So, how errors trended per 1,000 BBE in MLB?
Oh. That's the same effect!
 |
| Errors Per 1K Batted Balls (2015 - 2024, AKA The Statcast Era) |
In 2019, 23 errors occurred in an MLB game every 1,000 times a ball was put into play.
In 2024, 21 errors occurred in an MLB game every 1,000 times a ball was put into play.
9% less often.
At a high level, it's not just the sheer number of batted balls in a game that might be sneakily reducing our errors per game.
So what is it, then? What is hoovering up all those errors, more than 300 a year, and making sure they don't see the light of day?
Answers (if there are any) will have to be saved for future posts, though the articles I linked to above certainly offer up some hints.
Instead, I want to wrap this post up with two more descriptive looks to fuel the conversation.
The first: the ratio of throwing errors to fielding errors.
Errors come in two forms: throwing errors (SS airmails the throw to first on a grounder) and fielding errors (SS simply drops a popup). Can we determine if (any) change in the ratio between the two can point to something bigger at play?
 |
| Ratio of Throwing Errors Per Fielding Error (2015 - 2024, AKA The Statcast Era) |
Frankly...not really. Throwing Errors Per Fielding Error are only slightly less numerous in 2024 than they were in 2019, and outside of a nasty dip in 2022, have mostly hovered in the 1.1 to 1.2 range. Thus, it's hard to suggest that our decline in errors is biased towards one type of error in particular, which is a proxy a significant change in the shape of batted ball distribution.
The second: errors by position.
Are certain positions more or less responsible for our loss in errors here?
Maybe.
On a per 1,000 batted ball events basis, there are some interesting cases.
Catcher error rate per 1,000 BBE is up from 2.38 in 2019 to 2.62 this past year, with a high of 2.94 in 2023. This is the only position group to have seen their errors per 1,000 BBE actually increase from 2019 to 2024.
1B error rate per 1,000 BBE is down from a 2019 high of 2.39 to a low of 1.72 in 2023, and 1.79 this past year. That decrease of .60 is the largest absolute value of any group in the 2019 - 2024 time frame.
2B error rate per 1,000 BBE is down from a 2019 mark of 2.91 to a 2024 mark of 2.41. That decrease of .50 makes 2B the only other position group to have their errors per 1,000 BBE move 0.5 or more in either direction.
 |
| Errors Per 1,000 Batted Balls by Position (2015 - 2024, AKA The Statcast Era) |
However, most of these differences are basically imperceptible to fans. If 1,000 BBE is a day's worth of MLB games, you would need to be locked in, truly locked all the way in, across FOUR full days of action to notice a single marginal catcher error. Even being totally aware of the trend in 1B and 2B errors would take two full days of action to make itself clear.
So what gives here - what is driving the evaporating error?
Maybe it's about the shape of the batted balls, namely the underlying metrics like exit velocity and launch angle?
How about where these batted balls are hits, namely the spray angle and landing spot?
Or is it something else, something we might struggle to measure - a new edict from MLB or a relaxing of the rules that hasn't been clearly disseminated to fans?
To find that out, you'll have to keep following along!
Thanks for reading!
--------
Code is HERE on my GitHub.
--------
Special thanks to Fangraphs for the data!
Comments
Post a Comment