Catcher × Pitch-Type Cooperation
Most public framing models score a catcher with a single number. CPC v0 says that single number leaves the strongest signal on the floor: the variance of (catcher × pitch-type) interaction is 4.47× larger than the catcher main effect alone. The same catcher can be the league's best framer on a sinker and middle-of-the-pack on a slider.
Built on 1.13M taken pitches across 2022-2024 — 117 qualified catchers, 541 quality (catcher × pitch_type) cells. Transfer r=+0.843 across non-overlapping windows (2022-23 → 2024) means catcher framing skill is real and persistent, not a luck artifact. Pitcher-control survival 3.40× — the signal strengthens after removing pitcher main effect, not weakens.
Catcher Leaderboard
click any name to drill into pitch-type breakdownSorted by cat_cs_rate — controlled-for-pitch-mix called-strike rate on taken pitches. Spread = (max − min) CS% across that catcher's qualifying pitch types; high spread means the catcher's framing is very pitch-specific. Austin Barnes leads in spread.
Pitch Drilldown
select a catcherPer-pitch-type CS% for the selected catcher. The grey baseline mark on each bar is the league pitch-type main effect — delta measures how much this catcher's CS% on this pitch type beats or trails the league baseline.
Pitch Type Main Effects
league baselinesLeague-wide CS% per pitch type. SI (sinker, in-zone late tail) is easiest; CH (changeup, out-of-zone bait) is hardest. These are the baselines the catcher drill-down compares against.
Methodology
cpc v0 · 2022-2024 · taken pitchesSubstrate:
- Pitch-level: 1.13M taken pitches 2022-2024 (TAKEN-only removes the batter-swing strategic confound)
- Within-cell residualization: for each (catcher × pitch_type) cell, we subtract the catcher main effect AND the pitch-type main effect — what remains is the interaction
- 117 qualified catchers (≥ 1500 taken pitches across the window)
- 541 quality cells (≥ 200 taken pitches per cell to qualify)
- 7 pitch-type bins: FF (fastball), SI (sinker), SL (slider), CU (curveball), CH (changeup), FC (cutter), OTHER
Headline finding: σ²(catcher × pitch_type) / σ²(catcher_main) = 4.47. The interaction variance is four-and-a-half times larger than the variance of catcher main effects. Translation: when you're shopping for a framer, you should be asking "framer for what pitch shape?" — not just "good framer?".
Robustness checks:
- Cross-window transferability: r=+0.843 (n=389 cells, 2022-23 vs 2024). Highest transfer in the post-compact Resolvemetrics arc — catcher framing is the most persistent signal we've measured.
- Pitcher-control survival: σ²(C×PT) / σ²(pitcher_main) = 3.40×. Removing pitcher main effect STRENGTHENS the catcher×pitch-type signal — it's intrinsic to the catcher, not an artifact of which pitchers they happen to catch.
- Face validity: Patrick Bailey #1 (consensus #1 in public framing models), Alejandro Kirk, Yadier Molina, Austin Hedges all in top 15. Bottom: Jason Castro, Tyler Soderstrom, Luis Torrens, Korey Lee.
What this is NOT: not a complete catcher value metric. CPC measures framing on taken pitches; complementary skills (batted-ball receiving, pitch-calling, blocking, throwing) are separate axes — xCRA measures batted-ball receiving and gives a different leaderboard (no overlap with CPC top-15 — they're orthogonal catcher skill axes). Report jointly, not averaged.