What Makes a Pitch "Filthy"? The Data Behind Nasty Stuff

Every dugout uses the word and nobody defines it. A pitch is filthy, or it is nasty, or it has late life — all of which mean roughly "that was really hard to hit" and none of which tell a pitcher what to do differently on Tuesday. The interesting question is whether the thing everyone recognizes instantly can actually be measured. It mostly can.

Start by separating two things that get conflated. A pitch that looks spectacular in a slow-motion clip and a pitch that hitters cannot do anything with are not the same category, and the gap between them is where a lot of pitch-design effort gets wasted.

A curveball with enormous break, thrown from a visibly different slot, that starts so far outside the zone the hitter never considers swinging, is beautiful and harmless. Hitters take it for a ball. Meanwhile a slider that moves half as much but comes out of the hand looking exactly like a fastball produces awkward swings all night. The second pitch is the filthy one, and the highlight reel prefers the first.

Break the word down and it resolves into four things, all of which are individually measurable with tracking data you already have.

• Velocity — how much time the hitter has to decide. Every mph removes reaction time, and it compounds with everything below.
• Movement — how far the pitch deviates from where a spin-less ball would end up. Induced vertical break and horizontal break capture this.
• Deception — how much the pitch resembles another pitch out of the hand. This is tunneling, and it is the ingredient people underrate most.
• Location — whether it finishes somewhere a hitter can do damage. The nastiest shape hung middle-middle is a souvenir.

Notice that only two of those are properties of the pitch in isolation. Deception only exists relative to the pitcher's other pitches, and location only matters relative to the zone and the count. That is why a single-pitch grade will always be incomplete — filth is contextual.

If filth were just movement, the filthiest arsenal in baseball would belong to whoever throws the biggest curveball, and it does not. The reason is timing: a hitter has to commit to his swing while the ball is still roughly 20 to 25 feet from the plate.

Movement that happens before that commit point is information the hitter gets to use. He sees the ball diverging from the fastball path early, identifies it, and either takes it or adjusts. Movement that happens after the commit point is movement he cannot respond to, because the swing is already in motion.

That is the whole argument for late separation over large separation, and it is why two pitchers with identical movement profiles can have completely different results. The one whose pitches stay together longer out of the hand is the one who is filthy.

Since filth is about the hitter's experience, the cleanest measurement is the hitter's response. That is what CSW% captures — the share of pitches that produced a called strike or a swinging strike, meaning the pitch beat either the hitter's judgment or his bat.

It is not a perfect stand-in. A pitch can generate weak contact rather than whiffs and still be excellent, and CSW% will not credit that. But as a single number describing "how often did this pitch simply win," it is the most useful one available, it stabilizes quickly, and it separates cleanly by pitch type — so you can grade an individual slider rather than a whole pitcher.

The practical move is to look at CSW% alongside the ingredients. A pitch with a high CSW% tells you it is working. Velocity, movement, and release consistency tell you why, which is the part you can actually train.

If you wanted to collapse all of that into one number — and this is exactly what a Filth Score does — the honest construction combines the inputs with the outcome rather than picking one:

• Velocity, relative to the level being played. 91 means something different in a high school game than in the Cape.
• Movement, weighted toward the type of movement that pitch is supposed to have. Low vertical break is good on a changeup and bad on a four-seam.
• Release consistency against the rest of the arsenal, as a proxy for deception.
• Whiff and called-strike rates as the outcome check on all of the above.

The weighting is where judgment enters, and anybody who tells you their grade is objective is overselling it. What a grade genuinely buys you is speed — a coach scanning a staff can see which pitches deserve attention without reading four charts per pitcher. The charts still matter when you get there.

The useful takeaway is that filth is mostly trainable, and not in the way people assume. Velocity gains are slow and have a ceiling. Movement can be changed with grip and intent but also has limits. Release consistency and sequencing — the deception ingredients — are the most improvable and the most neglected.

If a pitcher's fastball and slider leave his hand from measurably different release points, fixing that will do more for how his stuff plays than another inch of break. It shows up in the data, it responds to work, and it costs nothing but attention.

Which is the honest answer to the question in the title: a filthy pitch is usually not the one with the most movement. It is the one the hitter could not identify in time.