'Yu Darvish's Pitching Evolution (2021-2025): A Statcast Data Analysis'

Source: Dev.to

Introduction

In November 2025, Yu Darvish underwent UCL (ulnar collateral ligament) surgery, ruling him out for the entire 2026 season.

Over his five years with the San Diego Padres (2021‑2025), his game appearances declined from 30 to 15. Looking at the Statcast data from this period, the numbers suggest significant shifts in his pitch mix and pitching style.

In this article I use pybaseball + DuckDB to pull five seasons of Statcast data and examine how his pitch selection, velocity, and whiff rates changed over time.

Note: This is a purely data‑driven exploration of observable trends – it is not intended to make definitive claims about the pitcher’s intentions or decision‑making.

The Google Colab notebook used for this analysis is linked at the bottom of the article.

Before diving into the data, here is a brief timeline of key events during this period.

Season‑by‑Season Summary

Elbow‑related injuries have been a recurring theme since 2023. Let’s see what changes appear in the data from this period.

Data Extraction

from pybaseball import statcast_pitcher
import duckdb
import pandas as pd

PITCHER_ID = 506433          # Yu Darvish
YEARS = [2021, 2022, 2023, 2024, 2025]

dfs = []
for year in YEARS:
    df_year = statcast_pitcher(f'{year}-03-01', f'{year}-12-31', PITCHER_ID)
    df_year['season'] = year
    dfs.append(df_year)

df_raw = pd.concat(dfs, ignore_index=True)

# Regular season only
con = duckdb.connect()
df = con.execute(
    "SELECT * FROM df_raw WHERE game_type = 'R'"
).df()                     # → 10,387 pitches (5 seasons combined)

# Important: Filtering by `game_type = 'R'` is necessary to exclude spring‑training
# and postseason data.

Pitch‑Usage Rates (2021‑2025)

• In 2021, slider (31 %) + sweeper (23 %) = 54 % of all pitches, making horizontal breaking balls the core of his arsenal.
• By 2025 that combined share had dropped to 24 %, while sinkers (20 %), curveballs (15 %), and cutters (12 %) rose sharply.
• The 2023 season stands out: slider usage fell from 31.5 % to 17.7 % while sinker usage jumped from 8.5 % to 18.6 %. That year featured a delayed start (WBC) and an August shutdown for an olecranon stress reaction, which may have prompted the shift.

Generally, sliders and sweepers are thought to place relatively more stress on the elbow, whereas sinkers involve a more natural arm motion and curveballs rely more on speed differential. The observed shift could reflect an adjustment for elbow health, though strategic considerations and other factors may also be at play.

Average Velocity by Pitch Type

A 0.5‑0.6 mph drop in four‑seam velocity over five years is modest. The pitch‑mix changes therefore appear to be driven by factors other than pure velocity loss—perhaps condition management, strategic adjustments, or a combination of both.

Velocity Drop‑Off by Inning

In 2021 Darvish maintained his velocity through seven innings; by 2025 the data show a 1.4 mph drop by the sixth inning. The 2024‑2025 figures include starts made shortly after returning from injury, which likely accentuates the late‑inning decline.

Whiff Rate (Swinging‑Strike %)

The four‑seam whiff rate fell from 30 % to 14.3 %, while the curveball’s whiff rate peaked at 44 % in 2023. The shift in which pitches generated the most swings and misses appears to correlate with the changes in pitch selection.

Two‑Strike Pitch Selection

In 2021‑2022 the four‑seam and sweeper were his primary “put‑away” pitches. By 2025 the curveball had become his most‑used two‑strike offering.

Key Takeaways

Across Yu Darvish’s five seasons with the Padres, Statcast data reveal the following trends:

These patterns suggest a combination of health‑driven adjustments (especially concerning elbow stress) and strategic evolution as Darvish aged and dealt with recurring injuries.

Notebook Link

The full Google Colab notebook used for this analysis can be found here: Yu Darvish Pitch‑Mix Exploration (Colab).

Weeper to Curveball

The changes seem to have accelerated around 2023, coinciding with his WBC participation and the olecranon stress reaction. The data suggests that changes in physical condition may have influenced his pitch selection over time.

Note: The 2024 (16 games) and 2025 (15 games) seasons have small sample sizes, so the numbers should be interpreted with caution. Whiff rates for less‑frequently thrown pitches are especially volatile with fewer observations. Additionally, the data includes starts made shortly after injury rehabilitation, which may not reflect typical performance levels.

You can run the full analysis using the notebook below. Change PITCHER_ID or YEARS to apply the same analysis to other pitchers.

• pybaseball GitHub Repository:
• Baseball Savant: