FIP (Fielding Independent Pitching)

FIP Explained: Fielding Independent Pitching

Fielding Independent Pitching (FIP) is one of the most important advanced metrics in modern baseball analytics. Developed by Tom Tango, FIP measures what a pitcher's ERA should look like based solely on the outcomes they can control: strikeouts, walks, hit by pitches, and home runs.

What FIP Measures

FIP isolates pitcher performance from defensive influence by focusing on three true outcomes plus home runs:

• Strikeouts (K): Pitcher retired batter without ball in play
• Walks (BB): Pitcher allowed batter to reach without hitting
• Hit by Pitch (HBP): Pitcher hit the batter
• Home Runs (HR): Only hit type FIP considers (can't be fielded)

The FIP Formula

FIP = ((13×HR) + (3×(BB+HBP)) - (2×K)) / IP + FIP_constant

Component Weights:

• Home Runs (13): Highest weight—worst possible outcome
• Walks/HBP (3): Free baserunners hurt run prevention
• Strikeouts (-2): Subtracted—best outcome for pitcher

The FIP constant (typically ~3.10) scales FIP to match league-average ERA.

FIP vs ERA: When to Use Each

FIP Benchmarks

Advanced Variants: xFIP and SIERA

xFIP (Expected FIP)

Replaces actual HR with expected HR based on fly ball rate × league HR/FB%. Normalizes for home run luck.

SIERA (Skill-Interactive ERA)

More complex metric incorporating strikeout rate, walk rate, and ground ball rate with interaction terms. Generally most predictive of future ERA.

Python Implementation

from pybaseball import pitching_stats
import pandas as pd
import matplotlib.pyplot as plt

# Get pitching statistics
stats_2024 = pitching_stats(2024, qual=100)

# Calculate FIP-ERA differential
stats_2024['FIP_ERA_diff'] = stats_2024['FIP'] - stats_2024['ERA']

# Top FIP performers
print("Top 10 Pitchers by FIP (2024)")
print("=" * 60)
top_fip = stats_2024.nsmallest(10, 'FIP')[['Name', 'Team', 'IP', 'ERA', 'FIP', 'xFIP', 'K/9', 'BB/9']]
print(top_fip.to_string(index=False))

# Identify regression candidates
print("\nPitchers Likely to Regress (ERA much better than FIP):")
lucky = stats_2024.nsmallest(10, 'FIP_ERA_diff')[['Name', 'ERA', 'FIP', 'FIP_ERA_diff']]
print(lucky.to_string(index=False))

print("\nPitchers Likely to Improve (ERA much worse than FIP):")
unlucky = stats_2024.nlargest(10, 'FIP_ERA_diff')[['Name', 'ERA', 'FIP', 'FIP_ERA_diff']]
print(unlucky.to_string(index=False))

# Visualization: FIP vs ERA
plt.figure(figsize=(10, 8))
plt.scatter(stats_2024['FIP'], stats_2024['ERA'], alpha=0.6)
plt.plot([2, 6], [2, 6], 'r--', label='Perfect Agreement')
plt.xlabel('FIP', fontsize=12)
plt.ylabel('ERA', fontsize=12)
plt.title('FIP vs ERA (2024 Season)', fontsize=14)
plt.legend()
plt.grid(True, alpha=0.3)
plt.tight_layout()
plt.savefig('fip_vs_era.png', dpi=300)
plt.show()

# Component analysis
stats_2024['HR_component'] = (13 * stats_2024['HR']) / stats_2024['IP']
stats_2024['BB_component'] = (3 * (stats_2024['BB'] + stats_2024['HBP'])) / stats_2024['IP']
stats_2024['K_component'] = (2 * stats_2024['SO']) / stats_2024['IP']

print("\nFIP Component Analysis (Top 5 by FIP):")
components = stats_2024.nsmallest(5, 'FIP')[['Name', 'FIP', 'K/9', 'BB/9', 'HR/9']]
print(components.to_string(index=False))

# Year-to-year stability comparison
stats_2023 = pitching_stats(2023, qual=100)
merged = pd.merge(
    stats_2024[['IDfg', 'Name', 'FIP', 'ERA']],
    stats_2023[['IDfg', 'FIP', 'ERA']],
    on='IDfg', suffixes=('_2024', '_2023')
)

fip_corr = merged['FIP_2023'].corr(merged['FIP_2024'])
era_corr = merged['ERA_2023'].corr(merged['ERA_2024'])

print(f"\nYear-to-Year Correlation:")
print(f"FIP: {fip_corr:.3f}")
print(f"ERA: {era_corr:.3f}")
print(f"FIP is {((fip_corr - era_corr) / era_corr * 100):.1f}% more stable")

R Implementation

library(baseballr)
library(dplyr)
library(ggplot2)

# Get pitching data
fg_pitching <- fg_pitch_leaders(
  startseason = 2024,
  endseason = 2024,
  qual = 100
)

# Calculate FIP-ERA differential
fg_pitching <- fg_pitching %>%
  mutate(
    FIP_ERA_diff = FIP - ERA,
    luck_status = case_when(
      FIP_ERA_diff > 0.5 ~ "Unlucky",
      FIP_ERA_diff < -0.5 ~ "Lucky",
      TRUE ~ "Neutral"
    )
  )

# Top FIP performers
cat("Top 10 Pitchers by FIP (2024)\n")
top_fip <- fg_pitching %>%
  select(Name, Team, IP, ERA, FIP, xFIP, `K/9`, `BB/9`) %>%
  arrange(FIP) %>%
  head(10)
print(top_fip)

# Regression candidates
cat("\nLucky Pitchers (ERA << FIP):\n")
lucky <- fg_pitching %>%
  select(Name, ERA, FIP, FIP_ERA_diff) %>%
  arrange(FIP_ERA_diff) %>%
  head(10)
print(lucky)

# Visualization: FIP vs ERA with luck indicators
ggplot(fg_pitching, aes(x = FIP, y = ERA, color = luck_status)) +
  geom_point(size = 3, alpha = 0.7) +
  geom_abline(intercept = 0, slope = 1, linetype = "dashed", color = "red") +
  scale_color_manual(values = c(
    "Unlucky" = "#E74C3C",
    "Neutral" = "#95A5A6",
    "Lucky" = "#3498DB"
  )) +
  labs(
    title = "FIP vs ERA: Identifying Lucky and Unlucky Pitchers",
    x = "FIP (Fielding Independent Pitching)",
    y = "ERA (Earned Run Average)",
    color = "Performance"
  ) +
  theme_minimal()

ggsave("fip_era_comparison.png", width = 12, height = 8, dpi = 300)

# FIP category summary
fip_summary <- fg_pitching %>%
  mutate(
    FIP_category = case_when(
      FIP < 3.00 ~ "Elite",
      FIP < 3.75 ~ "Above Avg",
      FIP < 4.25 ~ "Average",
      TRUE ~ "Below Avg"
    )
  ) %>%
  group_by(FIP_category) %>%
  summarise(
    count = n(),
    avg_ERA = mean(ERA),
    avg_FIP = mean(FIP),
    avg_K9 = mean(`K/9`),
    avg_BB9 = mean(`BB/9`)
  )
print(fip_summary)

Key Takeaways

• FIP isolates pitcher skill: Focuses on outcomes pitchers control
• More predictive than ERA: Better for forecasting future performance
• FIP-ERA gaps reveal luck: Large differences indicate regression potential
• xFIP normalizes HR luck: Uses expected HR based on fly ball rate
• SIERA is most sophisticated: Accounts for skill interactions
• Use both metrics together: FIP for projection, ERA for actual results