How to Actually Read a Statcast Leaderboard

Open a Statcast leaderboard for the first time and it looks like a wall of red and blue: little colored bubbles next to every number, a column for exit velocity, another for barrels, another for chase rate, a few you’ve never heard of. It is, genuinely, the best free window into baseball that has ever existed. It is also a machine for fooling yourself, and most of the bad takes I see lifted off of it come from three or four mistakes people make over and over. So this is the guide I wish someone had handed me: how to read one of these boards without walking away believing something that isn’t true.

I’m going to talk in terms of the Baseball Savant percentile rankings — the player pages with the colored sliders — because that’s the version most people actually look at. But the same instincts carry over to any leaderboard with a sortable column and a date range.

The colors are percentiles, not grades

The single most important thing to internalize: those red and blue bubbles are percentiles, not raw values and not letter grades. A bright red 95th-percentile exit velocity means the player is harder-hitting than 95 percent of his peers in that population — it does not tell you the actual number, and it does not mean “95 out of 100” in any absolute sense. The percentile is relative to a pool, and the pool matters. Hitters are ranked against qualified hitters; pitchers against pitchers. Change the population and the same raw number lands on a different color.

This sounds pedantic until you watch someone compare a starter’s percentile to a reliever’s, or a part-timer’s to a full-season regular’s, and conclude something the data never said. Two players can post the identical 91.5 mph average exit velocity and sit at different percentiles in different years simply because the league around them moved. Percentiles are a fantastic way to say “how does this guy stack up right now,” and a terrible way to compare across seasons or across populations. When I want to track a player over time, I pull the raw value, not the color.

Raw value or percentile? Pick the right one for the question

Here’s the rule I use. If the question is where does this player rank among his peers this season, the percentile is exactly the tool — it’s built for that. If the question is did this player actually change, or how do two different seasons compare, you want the raw value, because the percentile silently re-baselines every year. A hitter whose average exit velocity held perfectly steady at 90 mph can drift from the 70th percentile to the 60th without swinging any differently, just because the league got harder around him. The slider moved; the player didn’t.

Neither is “better.” They answer different questions. The mistake is using the percentile — which is the prettier, more prominent number on the page — for the comparison the raw value is supposed to handle.

Signal versus noise: the sample-size problem

Now the big one. Every number on a leaderboard is computed over some sample, and a small sample is a liar. The board will happily show you a hitter sitting at a .650 wOBA over his first nine plate appearances of the year, bubble glowing the brightest red the page can render, and that number means essentially nothing. It is four good swings and a lot of luck. The leaderboard does not warn you. It just shows the number.

So before I react to anything, I look at the denominator. How many batted balls is this exit-velocity ranking built on — twelve, or two hundred? How many pitches feed this chase rate? A metric computed over thirty events and the same metric over a full season are not the same kind of object, even when the column header is identical. The leaderboard flattens that distinction into one tidy cell, and reading it well means mentally un-flattening it every single time.

Which metrics stabilize fast (and which don’t)

The good news is that not all metrics are equally noisy, and the differences are knowable. A statistic “stabilizes” when it has accumulated enough events that the number is mostly skill rather than mostly luck — when, roughly, half of what you’re seeing is the real player. The order is consistent, and it tracks how directly the stat measures the swing or the pitch itself versus a downstream outcome:

Treat those as the order of trust, not as exact thresholds — the precise sample where each one “stabilizes” varies by study and by how you define it, and I’m not going to quote you a specific event count as gospel. The pattern is what matters. The further down that list a stat sits, the longer you wait before believing the leaderboard. Why the order? Because the top of the list measures things the player controls directly and does constantly — a hitter makes a swing decision on nearly every pitch, so chase rate piles up evidence fast. The bottom measures outcomes that pass through fielders, ballparks, and the timing of when hits clump together, none of which the hitter controls and all of which add noise. This is the same logic that powers regression to the mean: the noisier the stat, the harder it regresses toward the player’s true level.

The practical payoff is enormous. In April, if a hitter’s chase rate has genuinely dropped, I’ll believe it well before I believe his shiny new batting average, because chase rate has already seen hundreds of decisions while the average is still mostly noise. The leaderboard shows both in the same red. Only one of them has earned it yet.

Common misreadings, collected

Most leaderboard disasters are one of these:

• Treating a percentile as a raw value. A 70th-percentile barrel rate is not “70 percent of balls were barreled.” It’s a rank. The actual barrel rate is a small single-digit number for almost everyone.
• Reacting to small samples. The most extreme bubbles in April are almost always the smallest samples. Big numbers and tiny denominators travel together; the loudest cell is frequently the least real.
• Comparing across mismatched populations. A starter and a reliever, or a regular and a bench bat, ranked against different pools, then compared as if the colors mean the same thing.
• Trusting outcome stats over skill stats too early. Believing a hot batting average while ignoring that the chase rate and whiff rate — the stats that stabilize first — haven’t budged. The outcome will usually drift back toward the skills.
• Forgetting expected stats exist. When wOBA and xwOBA diverge hard, the leaderboard is quietly telling you luck is involved. Read both columns, not just the one that flatters your take.

A reading checklist

When a leaderboard cell jumps out at me, I run the same four questions before I say anything out loud. What’s the sample — how many events is this built on? Is this a percentile or a raw value, and which one does my question actually need? Where does this metric sit on the stabilization list — is it the kind of stat I can trust early, or the kind that needs a full season? And does an expected-stat sibling agree, or is something lucky propping this up? Four questions, fifteen seconds, and most of the embarrassing takes never make it out of my head.

The bottom line

A Statcast leaderboard is not a verdict; it’s a pile of evidence of wildly varying quality, presented in uniform, confident colors. The skill is learning to discount the right cells — to see a glowing 99th-percentile bubble over nine plate appearances and feel nothing, while a quiet, unsexy chase-rate improvement over four hundred pitches makes you sit up. Read the denominator, know which stats have earned your trust, and keep the percentile and the raw value in their separate lanes. Do that and the same board that generates bad takes for everyone else starts handing you real ones.