Bullpen Volatility: MLB Late-Inning Edges and Totals

TL;DR: Full-game totals live or die in the 7th–9th inning. You can find an edge by tracking who is fresh, who is tired, how managers use relievers, park and weather, and small tells before first pitch. Use data from trusted places like FanGraphs, Baseball Savant, Baseball-Reference, Umpire Scorecards, and Ballpark Pal. Then build a simple “rest penalty” model and a plan for live spots. Shop lines and bet with care.

  • Why late innings decide tickets
  • Why bullpens drive totals more than you think
  • What drives bullpen volatility
  • Metrics that predict bullpen performance
  • A simple late-inning model you can build
  • Market timing and information edges
  • Case studies
  • Practical pre-bet checklist
  • Where to compare lines and read reviews
  • FAQ
  • Sources and method notes
  • About the author and trust
  • Responsible gambling

Why late innings decide tickets

Totals do not end in the 5th. They often start in the 7th. One bad slider from a tired setup man can turn 4–3 into 7–6 fast. Starters face the lineup two or three times. Books price them well. But pens are a mix of roles, rest, matchups, and manager choices. This is where the market misses most. If you track who is fresh, who is taxed, and who will face the top of the order, you can price the last three innings better than a simple season ERA.

This article shows you clear steps. We will keep the words simple. We will still use key terms like gmLI (game leverage index) and xFIP, and we will link to trusted guides so you can go deeper when you want.

Why bullpens drive totals more than you think

Many models love the starting pitcher. That is fine for First 5 (F5) totals. But full-game totals depend on 9–12 pitchers, not two. The gap between a fresh pen and a tired pen can be over one run. The market does not always price that in.

  • Reliever roles change fast. Managers ride hot arms or rest a closer after back-to-back days. See the three-batter minimum rule here: MLB three-batter minimum.
  • Leverage changes outcomes. High leverage means tougher plate spots. Learn about leverage index (LI) and gmLI at FanGraphs Library.
  • Usage is not just days worked; pitch counts and up-downs in the bullpen matter too, even if not in the box score.

Think of a “Bullpen Heat Index.” It blends rest, pitch load, role certainty, and matchups. A high number means risk and more runs. A low number means strength and fewer runs. You can build this in a simple sheet and adjust your total by 0.5–1.0 runs in some games.

What drives bullpen volatility

1) Usage and rest

  • Days since last pitched: Track 0 days rest (pitched yesterday), back-to-back (B2B), and 3-in-4 days. You can check days rest and usage at Baseball-Reference team pages and reliever logs.
  • Pitch counts per outing: A reliever who threw 28 pitches yesterday is not the same as one who threw 8. The first will likely lose 0.2–0.4 runs per 9 in his next outing.
  • Hidden load: Warm-up throws count too, but they are not logged. Follow beat writers and pregame notes to catch “closer unavailable” hints.

2) Matchup and context

  • Handedness and pitch mix: Check pitch info and platoon splits at Baseball Savant. A slider-heavy righty vs a lefty stack is risk. A sinker guy vs a groundball lineup is safer.
  • Who comes up in the 7th–9th: If the top of the order is due, the pen will face better hitters. If it is 7-8-9, run risk drops.
  • Inherited runners: Some pens are great with traffic. Some are not. See “inherited runners” on Baseball-Reference reliever team stats.

3) Environment

  • Park factors: Some parks boost runs and homers. See park factors at FanGraphs Park Factors.
  • Weather: Wind out to right at Wrigley can add a full run. Use Ballpark Pal to see run and HR boosts by game.
  • Umpires: A tight zone means more walks and longer innings. Check recent reports at Umpire Scorecards.
  • Travel and schedule: A pen coming off a Sunday night game and a flight may fade on Monday.

4) Rules and trends

  • Three-batter minimum: If a lefty reliever starts an inning and loses feel, he must face three batters or finish the inning. That can spike runs fast.
  • Pitch clock: Fatigue can show as lost command late. Velocity dips late can be a red flag. You can track velo at Baseball Savant.

Metrics that predict bullpen performance

Do not lean on ERA for relievers. It is too noisy in small samples. Use skills-based stats and simple signals that hold up.

For each reliever

  • K-BB%: Strikeouts minus walks. Higher is better. Stabilizes faster than ERA. Read why sample size matters at FanGraphs: Sample Size.
  • xFIP / SIERA: Skill-based run estimators. They smooth out luck on homers and defense. See xFIP and SIERA.
  • Stuff+ and movement/velo changes: If Stuff+ or velo drops 1–2 mph after B2B, expect worse results. You can spot pitch trends on Baseball Savant.
  • GB% under wind: Groundball arms lower HR risk. With wind out, fly ball arms are risky. Learn HR/FB regression here: FanGraphs: HR/FB.
  • Release/timing stability: Sudden release changes can mean command issues. See pitch release charts on Savant.

For the bullpen as a unit

  • Pen WAR and depth: Deep pens survive off days. Use team reliever WAR on FanGraphs leaderboards.
  • gmLI and manager patterns: Who gets big spots? How strict is the closer role? Check gmLI at FanGraphs and read manager notes from beat writers.
  • Inherited runner scoring%: This matters for unders. See team pages at Baseball-Reference reliever stats.
  • Framing and passed balls: A good framer can save runs. See framing data at StatCorner Catcher Framing and Baseball Prospectus (framing).

Operational signals you can track in real time

  • “Unavailable” notes: If a manager says “closer is down today,” bump the total up 0.2–0.4 runs.
  • Pregame bullpen sessions: Reporters often post who is throwing and how they look.
  • Lineup timing: When lineups drop and umpires are announced, totals can move. Confirm umps at Umpire Scorecards.

A simple late-inning model you can build

You do not need code to do this. A spreadsheet is fine. Here is a clear plan you can copy.

Step 1: Project starter innings

  • Use last 5 starts, pitch count trend, and opponent patience (walk rate). Set a range: 5.2–6.2 IP, for example.
  • Make a low, mid, and high case. This will feed inning-by-inning pen use.

Step 2: Map bullpen availability

  • List top 6 relievers with role tags: CL (closer), SU (setup), MR (middle), LR (long).
  • Mark rest buckets: Fresh (1+ day), 0 days rest, B2B, 3-in-4. Note last pitch count.
  • Assign a rest penalty (runs per 9) for today: Fresh (1+ day, under 20 pitches last time): +0.00 0 days rest: +0.10 Back-to-back: +0.25 3-in-4 days: +0.45 4-in-6 days: +0.55
  • Fresh (1+ day, under 20 pitches last time): +0.00
  • 0 days rest: +0.10
  • Back-to-back: +0.25
  • 3-in-4 days: +0.45
  • 4-in-6 days: +0.55
  • These are starting points. Calibrate to your data. I got close to these after two seasons of tracking.
  • Fresh (1+ day, under 20 pitches last time): +0.00
  • 0 days rest: +0.10
  • Back-to-back: +0.25
  • 3-in-4 days: +0.45
  • 4-in-6 days: +0.55

Step 3: Assign likely innings by score state

  • If ahead by 1–2: SU in 7th–8th, CL in 9th (unless CL is down).
  • If tied: Best two arms get 8th–9th. If both are down, risk jumps.
  • If behind: Middle relief first; high leverage arms may sit.
  • Note the lineup pocket due (1–3, 4–6, 7–9). Good managers target pockets, not only innings.

Step 4: Convert to inning run rates

  • Start with a neutral league run rate per inning. Adjust for park and weather with Ballpark Pal.
  • Adjust each reliever’s baseline by rest penalty and skills (K-BB%, xFIP).
  • Sum the 6th–9th expected runs for both teams.

Step 5: Simulate or use ranges

  • Simple way: Make low/median/high totals by swapping a fresh arm for a tired one and by lineup pocket swings. Compare to the market total.
  • Advanced way: Run 10,000 sims with inning-by-inning draw. Use binomial for runs per inning with park- and pitcher-adjusted mean.

How to use it

  • Full game vs F5: If your edge comes from a tired pen, target full game, not F5.
  • Team totals: If only one pen is stretched, look at the other team’s over.
  • Live betting: Best windows are end-6th to mid-8th. Watch who is warming. If the closer is down and the setup man threw 24 pitches last night, a tied game leans to the over.

Market timing and information edges

  • When to bet: Early lines move on starting pitcher news. Late lines move on lineups, umps, and wind. Bullpen info is often last to price in. Many totals settle 30–60 minutes before first pitch.
  • Where the market misses: Overweighting reliever ERA in small samples. Use xFIP/SIERA instead. Ignoring the three-batter rule. You cannot “paper over” a wild lefty with a quick hook. Underpricing wind and heat for fly-ball pens. Use park and weather boosts from Ballpark Pal. Not tracking inherited runner skill. You can see this on Baseball-Reference.
  • Overweighting reliever ERA in small samples. Use xFIP/SIERA instead.
  • Ignoring the three-batter rule. You cannot “paper over” a wild lefty with a quick hook.
  • Underpricing wind and heat for fly-ball pens. Use park and weather boosts from Ballpark Pal.
  • Not tracking inherited runner skill. You can see this on Baseball-Reference.
  • Overweighting reliever ERA in small samples. Use xFIP/SIERA instead.
  • Ignoring the three-batter rule. You cannot “paper over” a wild lefty with a quick hook.
  • Underpricing wind and heat for fly-ball pens. Use park and weather boosts from Ballpark Pal.
  • Not tracking inherited runner skill. You can see this on Baseball-Reference.

Case studies

Case 1: Over due to tired setup men and wind out

Set-up A and Set-up B both threw 22+ pitches last night. Closer is “down” per beat note. Wind out to right is +15% HR at Wrigley (source: Ballpark Pal). Starter projects 5.2 IP. That leaves 10–11 outs for tired arms. Rest penalties push their run rates up ~0.5 runs/9 each. Model says full-game fair total is 9.3. Book posts 8.5 -110. This is a small but real edge on the over.

Case 2: Under with groundball pen, elite framer, and cold night

Opponent pen is average. Our pen has two groundball righties (55–60% GB), both fresh. Catcher is a top framer (see StatCorner). Umpire has a wide zone per last 30 days (see Umpire Scorecards). Temp is 50°F, wind in 8 mph. Model trims late-inning runs by ~0.4. Fair total 7.6, book 8.5 -105. Under has value.

Counterexample: Model miss on role change

The sheet assumed the usual closer for the 9th. But the manager used him in the 8th vs 1-2-3. The 9th went to a weak arm vs 4-5-6. A two-run homer flipped the under. Lesson: build “pocket-first” trees. Do not lock roles to innings. Tie roles to who is due up.

Practical pre-bet checklist

  • Confirm rest for the top three relievers on each team (0 days, B2B, 3-in-4). Note last pitch counts.
  • Check gmLI patterns and manager quotes (search beat writers and team notes).
  • Lock park and weather. Use Ballpark Pal for run and HR factors.
  • Confirm the plate ump and his zone at Umpire Scorecards.
  • Look at the 7th–9th lineup pockets likely to hit.
  • Have a live plan: who warms if tied, up 1, or down 1?
  • Compare numbers across books before you bet.

Where to compare lines and read reviews

Totals can move late when bullpen info hits. It helps to shop lines. If you want a simple, honest place to see book limits, baseball pricing, and bonus terms, this guide is a clean starting point. Always check rules and promos in your state. Read the small print. Use more than one book so you can grab the best number.

FAQ

Do bullpens matter more for totals or sides?

For full-game totals, bullpens often move the number more than starters do. For sides, bullpen gaps matter most in close games. In blowouts, it matters less.

Is F5 better than full game if a bullpen is gassed?

No. If the edge is bullpen fatigue, target the full game. F5 strips out that edge. Use F5 when the starter edge is big but pens look even.

How do I quantify bullpen fatigue fast?

Use rest buckets (0 days, B2B, 3-in-4). Add pitch counts. Apply small run penalties (about +0.10 to +0.55 runs/9). Then adjust for park, weather, and who is due up.

How does the three-batter minimum change late-inning plans?

Managers cannot yank a wild arm after one hitter unless he finishes the inning. This increases the cost of a bad matchup and can add runs. Read more at MLB.com.

What is the best time to place a totals bet for bullpen edges?

Often 30–60 minutes before first pitch, after lineups, umps, and weather confirm. Live, the best time is end-6th to mid-8th when you see who is warming and who is down.

Sources and method notes

  • Core data: FanGraphs (gmLI, WAR, park factors, xFIP/SIERA guides).
  • Pitch and batted ball: Baseball Savant (velo, movement, release charts).
  • Usage and inherited runners: Baseball-Reference.
  • Umpires: Umpire Scorecards.
  • Weather and park boosts: Ballpark Pal.

Method: I track rest buckets and pitch counts for key relievers. I set a base run rate per inning from park and weather. I add penalties for 0 days, B2B, 3-in-4, and 4-in-6. I assign innings by score state and lineup pocket, not only by inning number. I test and adjust each month. Please build your own sheet and check the math with your data.

About the author and trust

I model MLB totals and track bullpen rest and usage across the season. I focus on simple, testable rules, not hype. I cite public data and show limits. I do not claim sure wins. Baseball is high variance. Use this as a clear process you can verify.

Responsible gambling

Bet only if you are of legal age in your area. Set limits. Never chase. If you need help, visit the National Council on Problem Gambling (US) or BeGambleAware (UK). If you use any sportsbook links, assume there may be affiliate relationships. Read all terms. No bet is guaranteed.

Quick reference: Bullpen Heat Index (simple)

  • Start with 0.00
  • Add +0.10 if top setup or closer is on 0 days rest
  • Add +0.25 if any must-use arm is on B2B
  • Add +0.45 if any must-use arm is on 3-in-4
  • Add +0.10 if wind is +10% runs or more
  • Subtract -0.10 if elite framer catches
  • Subtract -0.10 if cold temp and wind in

High index: lean over or avoid unders. Low index: lean under or pass overs.

Extra learning links

  • FanGraphs Library: Reliever concepts
  • Plate discipline and pitch counts
  • Statcast glossary
  • MLB glossary

Totals are hard, but late-inning edges are real. Keep it simple, track rest, know roles, watch the weather, and shop the price. That is how you turn bullpen volatility into a plan.