Laz Diaz is an Above-Average Home Plate Umpire All of a Sudden

On what lessons umpiring holds for the importance of crime data

Don’t click away just yet, hear me out about why this is relevant to this newsletter that is ostensibly about crime data!

MLB umpires used to be really bad. I mean, laughably bad. Check out the below clip of Greg Maddux pitching to Sammy Sosa in 1997 (Excellent source video from Foolish Baseball).

And here’s Charlie Hough pitching to José Offerman in the first ever at bat of the Miami Marlins in 1993.

And marvel at the strike zone for Livan Hernandez in a 1997 playoff game against the Braves.

Laz Diaz is frequently seen as one of the worst umpires out there. Diaz has been an umpire since 1999, and the internet is filled with tales of his struggles. There is, of course, a lot more to umpiring than getting balls and strikes right, but I’m going to focus on just the balls and strikes part for the sake of this piece.

We can actually measure just how bad an umpire he has been thanks to MLB’s investments in the Statcast system (I wrote about it a bit two years ago and we actually built a fun little dashboard for evaluating umpires). Statcast was implemented in force in 2015 and measures everything that happens on the field to include the strike zone, so we can tell with pinpoint accuracy which balls were called strikes and which strikes were called balls.

There were 64 umpires that called 30,000 or more pitches from 2015 to 2022, and Laz Diaz ranked 5th worst in terms of accuracy at just over 90 percent (accuracy here to mean strikes called as strikes and balls called as balls). Laz Diaz was 0.8 percent below the MLB average between 2015 and 2022 which works out to nearly 300 more pitches over that span called incorrectly compared to an average MLB umpire (not even a good one!). So the data backs up the anecdote that Laz Diaz has been a bad ump.

But, a funny thing has been happening with every MLB umpire — including Laz Diaz — over the last few years. They’re improving.

Data from Statcast has become more widespread and so too has accountability. Umpires are graded on performance related to hard data, not anecdote, and umpires have shown that they can clearly get better in response. The improvement has been even more pronounced in 2025 and 2026 with the spread of the automatic balls and strikes (ABS) system to the minor leagues and (in 2026) MLB. Now batters and catchers can challenge an umpire’s call and, using video, determine whether a called strike should actually stand as one.

Of the 64 umpires in the initial sample, only one has improved more than Laz Diaz (that ump is Ramon De Jesus) relative to where they were on average from 2015 to 2022. Laz Diaz is now slightly above average in terms of ball/strike accuracy, coming in with the 23rd best accuracy so far in 2026.

He is, of course, not alone. Every single umpire is more accurate in 2026 than they were in the 2015-2022 timeframe with the worst umpires tending to show even more improvement.

The thing is that umpires aren’t getting better because they suddenly decided to get good, they’re getting better because there was suddenly an unprecedented level of accountability driven by improved data collection and reporting.

Which brings us (finally!) to crime data.

The United States has entered its golden age of crime data collection and reporting. More data is now available faster than ever before to allow us to see crime trends as they develop.

Right now, the crime trends are almost universally positive. Murder is down, robbery is down, car theft is down. Sure, there are outliers, but there are always outliers.

At some point, these trends will switch and yesterday’s huge decline may become tomorrow’s increase and then we are talking about historical context and base rates. Having such excellent and fast crime data will enable us to identify and properly contextualize these changing trends literally as they change.

The bigger piece, however, that this democratization of crime data provides is the ability to bring accountability on complex issues like never before.

Think about clearance rates. They’re frequently how we measure policing performance in terms of what share of crimes are ending in an arrest, but they’re also badly flawed. Last year’s crimes can count towards this year’s statistics and a wide range of exceptional clearance types can plausibly dramatically impact clearance rates.

Thanks to widely available NIBRS data and a huge uptick in NIBRS participation, we can realistically ask why, for example, more than 1 in 5 rapes in Delaware and 1 in 6 in Pennsylvania were cleared by exception in 2025 compared to 6.8 percent nationally.

Or we can see in near real-time as the share of firearms used in a murder is falling after reaching record highs in 2021/2022.

Or what percentage of vehicle burglaries in big Texas cities in 2025 featured a firearm being stolen.

Or track how carjackings surged post-COVID and, like murders, have absolutely plummeted in the last few years.

These are big problems, they’re complex problems, and — thanks to this golden age of crime data — they’re widely identifiable problems. And if a problem can be identified then responses can be implemented, the impact of solutions can be measured, and accountability can, in theory, be attained.

Crime data is basically in the 2015 era of Statcast. The FBI is publishing monthly data and now we have a fancy new data source that we are just learning how to use impactfully. The questions of today are answerable with consistent, standardized data that is not perfect but is far better than it has ever been before.

I wish that there was a straightforward solution like ABS out there that could be applied to crime, but obviously there is not. Crime data is imprecise and it is flawed, and it will always be both imprecise and flawed in ways that tracking the flight of a baseball is not.

If data can prompt accountability which turned Laz Diaz into an above average umpire, however, I like to think that massive improvements in the crime data infrastructure can lead to better strategies and more accountability in the crime and policing world.

---

New on the Jeff-alytics Podcast

The national conversation around gun violence tends to revolve around what laws should be passed next. But a lot of the work of reducing violence doesn’t happen in Congress. It happens in cities, hospitals, community organizations, police departments, schools, and increasingly through coordination between all of them.

My guest is Rob Wilcox, the president and CEO of the Fund for a Safer Future and the former deputy director of the White House Office of Gun Violence Prevention.

In this episode, we talk about what it was like helping build such a critical office from the ground up, how he tried to approach gun violence as more than just a legislative issue, what happens when the federal government starts thinking about violence reduction as an operational challenge instead of simply a political one, and what the future of gun violence prevention should look like.

Apple

Spotify

You can also catch it on the Jeff-alytics YouTube page where I’ll be posting episodes and video clips, so be sure to like and subscribe there if you’re so obliged!

And while you’re here, be sure to check out these other recent great episodes:

Niskanen Center Director of Criminal Justice Greg Newburn

Center for American Progress CEO Neera Tanden

Senator Chris Murphy

Manhattan Institute Senior Editor Charles Fein Lehman