Plain-English Definitions

Every betting + sabermetric term you'll see on Bookie Bullies, explained without jargon. If we use a word that isn't here, tell us and we'll add it.

Betting Markets

Moneyline ML

Straight pick on which team wins. Odds shown as +150 (bet $100 to win $150) or -130 (bet $130 to win $100). No spread, no total — just who wins.

Run Line RL

MLB's fixed spread, almost always ±1.5. The favorite wins by 2+ runs to cover (-1.5), or the underdog can lose by 1 run and still cover (+1.5). Pays close to even money on most matchups.

Spread NBA

NBA's variable spread, e.g. -7.5 means the favorite has to win by 8+. Set by the book to attract roughly equal action on both sides.

Total O/U

Over/Under on the combined final score. Over 8.5 means both teams' runs together total 9+. Doesn't matter who wins.

First 5 Innings F5

An MLB-only side market on just the first 5 innings of a game. Strips bullpen variance out — only the starting pitchers matter. Good when you trust an SP but not their bullpen.

F5 Total F5_OU

Over/Under on combined runs scored in the first 5 innings only. Typically lines around 4.0-5.5 runs.

Confidence + Edge

Win % prob

Our model's estimated probability that the pick wins, expressed as a percentage. A 60% pick is expected to win 60 times out of 100 over a long sample. Includes the "± X%" uncertainty band, which widens when the model and market disagree.

EV Expected Value

Average profit per unit wagered if our probability is correct. +5% EV means a $100 bet returns $5 profit on average over many repetitions. Anything positive means the model thinks we have an edge.

Edge

The gap between our win % and the de-vigged market implied %. If the market says a team wins 50% but we say 55%, our edge is +0.05. Sharp bettors hunt for any positive edge over a fair-priced market.

Stake

Recommended bet size in units (1u = 1% of bankroll by default). Computed via 1/8-Kelly — a fraction of the optimal Kelly criterion to reduce variance, since model probabilities aren't perfect.

Stars (★)

Confidence tier from 1-5 derived from final probability + edge. 5★ requires both ≥58% prob AND ≥4% edge over market. 4★ needs ≥55% prob.

Lock

A 5★ pick where the model strongly disagrees with market in our direction. Highest conviction call of the day. Note: locks lose too — frequently. There is no such thing as a guaranteed bet.

CLV Closing Line Value

The single most important metric for evaluating long-term skill. Measures whether our published odds beat the closing market line. Positive CLV over a large sample = you're +EV regardless of short-term wins/losses. We track and publish ours on the Track Record page.

MLB Stats

FIP Fielding Independent Pitching

Pitcher's ERA equivalent based only on outcomes they fully control: strikeouts, walks, home runs, hit-by-pitch. Strips out defensive luck. More predictive of next-start performance than raw ERA.

xFIP Expected FIP

FIP normalized to league-average HR/FB rate. Removes home-run-allowed luck on top of FIP's defensive-luck strip. The most predictive single pitcher metric.

fipMinus FIP-

FIP indexed to league average and adjusted for ballpark. 100 = league average, 90 = 10% better than league, 110 = 10% worse. Removes both park and league effects, so it's directly comparable across pitchers.

xERA

Statcast's quality-of-contact-derived ERA. Built from exit velocity + launch angle of every batted ball, predicts what ERA should be if outcomes converted at average rates. Less fluky than raw ERA.

wOBA Weighted On-Base Average

Single number summarizing offensive output per plate appearance, weighted by run value (a HR is worth more than a single, a walk less than a single). League average ≈ .320.

OAA Outs Above Average

Statcast's defensive metric — how many outs a fielder/team converted that an average defender wouldn't. +10 OAA = saved 10 hits over the season. Big team gaps swing 0.4 R/game.

Catcher Framing

Runs saved by a catcher's ability to "steal" extra strikes on borderline pitches. League ranges from +8 R / 150G (elite, like Posey) to -6 R / 150G. Translates directly to lower SP ERA.

Bullpen Quality

Team-level bullpen xERA, lower = better. A 3.20-xERA pen vs a 4.20 pen has roughly a 0.4 R late-game edge. We amplify this signal in close games (where the late innings matter most).

BABIP Batting Avg on Balls In Play

Hits ÷ (balls in play). League norm ≈ .300. A pitcher running .220 BABIP is getting lucky; expect regression toward .300 next start.

LOB% Left-On-Base %

Fraction of baserunners a pitcher strands. League norm ≈ 72%. A pitcher at 85% has been getting lucky with timing.

Park Factor

Multiplier vs neutral. Coors = 1.23 means Coors plays 23% above average for run scoring. We adjust totals using park factors specific to each park, plus separate by-handedness HR factors (Yankee Stadium boosts LHB HRs more than RHB).

NBA Stats

Net Rating

Points scored minus points allowed per 100 possessions. +5.0 net = elite; -5.0 = bottom-tier. Most predictive single team metric.

Pace

Possessions per 48 minutes. Faster pace → more possessions → higher totals. League avg ≈ 99-101.

ORtg / DRtg

Offensive / Defensive Rating — points scored or allowed per 100 possessions. Pace-adjusted versions of points per game.

VORP

Value Over Replacement Player. Used to estimate the run-scoring impact of an injury (e.g. Kawhi out = -3.5 VORP = ~5 pts off the team's expected score).

Four Factors

The four stats that explain ~90% of basketball outcomes: eFG% (effective FG%), TOV% (turnover rate), ORB% (offensive rebound rate), FTR (free throw rate).

Advanced + Quant

Skellam Distribution

The probability distribution of the difference between two Poisson variables. Used to compute MLB margin probabilities (home runs - away runs). The correct family for integer-count differences — most public models use Gaussian, which underestimates tail risk.

Negative Binomial

Used for MLB run total probabilities. Real run totals are overdispersed Poisson (variance > mean) because of streakiness. Negative Binomial models this; Gaussian doesn't.

Bayesian Uncertainty

The "± X%" next to each win probability. Computed from the spread between our model, market, ensemble, and closing-line components. Tight band = sharp consensus = high conviction. Wide band = strong disagreement = low conviction.

Kelly Criterion

The mathematically optimal stake size given win probability and odds: f* = (p·b - q) / b. We use 1/8-Kelly (1/8th of the optimal) because real probabilities have noise, and full Kelly amplifies estimation error catastrophically.

De-vigged Market Probability

Market odds with the bookmaker's vig (~4-5%) stripped out, leaving the "fair" probability the book actually believes. Computed as side_juice / (side_juice + opp_juice + vig_constant).

Steam Move

A sharp, sudden line movement caused by professional money hitting the market. We track this in line_movement.json and surface "steam with us" / "steam against us" as a model driver.

Closing Line

The final odds posted just before a game starts. The market's most accurate read since all available information has been incorporated. Beating the close consistently = positive long-term edge.

Brier Score

Standard calibration metric: mean squared error between predicted probability and actual outcome (0 or 1). Lower = better. A perfectly calibrated 60% pick that hits 60% of the time has Brier ≈ 0.24.

Last updated: 2026-05-08. See something missing? Email hello@bookiebullies.com.