Your browser version is outdated. We recommend that you update your browser to the latest version.

33. Sports Prop Bets: Finding Edges Beyond the Moneyline

I remember a quiet Tuesday in January. The favorite won on the moneyline, but the price had no juice. My win came from a different place. A power forward had foul trouble the game before. The coach said he would start the next night and get a longer run. The book hung his rebound prop at 7.5. I had him for 9.2. Pace up spot, weak glass team, backup center out. He cleared in the third. That night taught me a simple thing: props live in a small world, but small worlds hide gaps. If you learn how these markets move, you can find edges the big lines miss.

A 60‑Second Primer: What is a Prop?

  • A prop is a bet on a stat or event inside the game. It can be on a player or a team.
  • Common props: points, rebounds, assists, receptions, shots on goal, strikeouts, shots on target.
  • Pregame props trade all day. Live props update during the match.
  • Alt lines give you higher or lower numbers at different odds.
  • Limits are lower than sides and totals. Books can move fast on news.
  • Hold (also called vig or overround) is the built‑in margin for the house.
  • Mean is not the same as chance to go over. You need the full distribution.

Prop Markets at a Glance

NFL RB receptions Routes run, snap%, QB aDOT, wind NFL Next Gen Stats; Pro‑Football‑Reference Med–High Low–Med RB receptions with low QB attempts Use weather to shift target depth
NBA PRA (points+rebounds+assists) Minutes, pace, usage, on/off splits NBA Stats; Basketball‑Reference Med Med Points–assists role link Minutes drive most variance; note OT rules
MLB pitcher strikeouts Velo, spin, lineup K%, umpire Baseball Savant; FanGraphs Med Low–Med K with outs/IP NegBin beats Poisson for K tails
Soccer shots on target Role, xG/xSoT, opponent style StatsBomb blog; Understat Med Low Team xG with player SoT Adjust for set pieces and shot map
Tennis holds/breaks Serve/return% by surface, fatigue Tennis Abstract Low–Med Med Set score with next‑game props Use surface‑specific priors
NHL shots on goal Time on ice, line mates, PP time Natural Stat Trick Med Low–Med Team pace with player SOG Model shot rate per minute first
Golf fairways hit Driving accuracy, course width, wind PGA Tour ShotLink notes Med Low Wind with GIR Course fit matters more than form

Why Edges Exist Where Liquidity Thins Out

Prop markets are smaller. Limits are lower. Fewer eyes means more room for stale numbers. Props also react to tiny bits of news: a role tweak, a shift in pace, a wind note. Books need to price thousands of names each day. They lean on fast rules and past means. But props live in fat tails. Rotations swing. Matchups bend usage. Weather moves depth. Small leaks show up when price meets a real‑world change.

Another reason: distribution is hard. A book can set the mean right and still miss the chance to go over. That gap is your edge. Same‑game links make it even trickier. A bump in tempo can lift two or three props at once. If you see how stats move together, you can act before the screen does.

Pricing Props: From Averages to Real‑World Distributions

Books start with a player’s expected stat line. They add vig. They adjust for news and action. But the key is not the mean; it is the shape of results. A player can have a mean of 7.9 rebounds and still go over 8.5 more or less often than you think, based on variance and skew. Hold matters too. Know the margin so you can judge fair odds and pass on bad prices. For background on how markets set that margin, see industry hold rates and market structure at industry hold rates and market structure.

A quick example with MLB strikeout props. Say you project a pitcher for 6.2 Ks. Many people use a Poisson model to turn that mean into chances for 5, 6, 7, 8+. But Ks often have extra variance (some nights the slider is nasty, some nights not). A Negative Binomial model lets you add that extra spread. You then sum the tail for 7+ to get P(Over 6.5). If your P(Over 6.5) is 0.56, the fair price is about +114 on the underdog side of the two‑way, or -127 on the Over side before vig. Compare to the board. If the book offers Over 6.5 at -110, you have edge. If the book shows -140, pass. Mean is a start. Distribution sets value.

The Data Stack That Actually Moves the Needle

For NFL pass game props, tracking data helps a lot. Route depth, speed, and alignment tell you where targets will go when wind, coverage, or scheme shifts. You can find route trees and tracking data at route trees and tracking data.

For NBA, usage and on/off splits tell the story of role. If a star sits, who gets the shots, who gets the touches? The league’s glossary is a good place to learn terms. Start with usage rate and on/off splits glossary at usage rate and on/off splits glossary.

Baseball props live on pitch quality and batter skill. You can check Statcast leaderboards and pitch metrics at Statcast leaderboards and pitch metrics. To dig into hitter swing rates, chase, and contact, read the plate discipline metrics at plate discipline metrics.

Pick feeds that update fast, have a long history, and let you export or call an API. The stack is your edge. Noise in means weak props. Clean inputs, clear roles, and a sound link to outcomes win the day.

A Minimal Viable NBA PRA Model

Here is a small, real plan to price NBA PRA.

  • Project minutes first. Base it on coach habits, foul risk, and game script. Minutes are king.
  • Set team pace and spread. More trips, more stats.
  • Assign usage and touch share with on/off splits. Who holds the ball with Player X off?
  • Split points, boards, and assists by role. A guard gains assists before boards. A big gains boards before assists.
  • Adjust for matchup: rim vs mid‑range defense, rebound rate, switch rules.
  • Deal with OT rules. Some books count OT. Some do not.

Validate. Backtest on past games. Keep a hold‑out set for live tests. Check your quantiles (do 10% of lines hit above your 90th?). Track Brier or LogLoss for binary props (Over/Under). Keep a log of misses and why. Use on/off splits and historical game logs to spot role swings that do not show in season means yet.

Case Study: NFL RB Receptions in Heavy Wind

Idea: strong wind hurts deep throws. Teams throw short more. Running backs see more dump offs. To test, tag games with wind over 15–20 mph using stadium wind forecasts at stadium wind forecasts. Then check each RB’s snap% and routes run trend as a proxy for pass role. You can find snap counts and routes proxy at snap counts and routes proxy.

Next, watch team pass rate vs expected. If a coach turtled in wind last year, adjust. Metrics like EPA and pass rate over expected are tracked at EPA and pass rate over expected. Price the RB reception line with a small shift to target depth and a small boost to catch chance. The edge shows up most on backup RBs who play the two‑minute drill.

Case Study: MLB Strikeout Props and Rest Patterns

Pitchers add spin and velo with extra rest. Some lose feel with too much rest. Tag games by days since last start. Layer in the foe’s K% by split and recent lineup news. Check the context for spin rate and velocity at spin rate and velocity context. Build a model where mean Ks rise a bit with two extra days if the pitcher’s past trend shows a lift. Use Negative Binomial for the tail. Price 6.5 or 7.5 with your new curve. Limit size if the umpire is a late add; zone size can swing Ks hard.

Case Study: Soccer Shots on Target and Role Changes

Role is huge. A winger who cuts inside takes more shots on target than a true wide man who hugs the line and crosses. When a coach flips a winger to an inverted role, the shot map shifts. You can read about shot maps and role analysis at shot maps and role analysis. Then pull xG and shots on target data from xG and shots on target data.

Model SoT as a rate per 90, then adjust for minutes and set piece share. Be careful with penalties; they spike SoT but do not repeat at a high rate. Look at opponent style too. A deep block gives low shot quality but may allow more low‑value shots, which can still be on target if the forward shoots from close range after cutbacks.

Correlations, Same‑Game Parlays, and Hidden Overrounds

Props in the same game link to each other. Pace rises, so do many counting stats. One player’s usage rise can rob another. If you build a Same‑Game Parlay (SGP), you must model these links. Many books price SGPs with a higher hidden hold. This extra tax can kill EV. If you want to learn from research on this, read about correlation pitfalls in sports betting at correlation pitfalls in sports betting.

When you parlay, force your model to draw from the same game script. If you think the game runs slow and low, do not stack three Overs on counting stats. If the book blocks clear links (like QB yards with WR yards) or adds a big tax, take singles.

Limits, House Rules, and Smart Line Shopping

Props have fine print. Does OT count? What if a player is active but never sees the field (DNP)? What if there is a stat correction the next day? Each book has its own rule set. Before you bet, read them. For a neutral view of how sites settle props, see settlement rules and regulatory guidance at settlement rules and regulatory guidance.

Real limits also vary by market and time. Some books show a high number but cut bet size at confirm. Others move the line at small stakes. The fastest way to raise EV with no extra risk is to shop for lines and fair rules. If you want a clean snapshot of rules, payout quirks, and live limits by brand, check https://globigames.com/ before you place a prop. It helps you avoid traps like “OT does not count” when your model thinks it does.

Risk, Bankroll, and When to Pass on “Edges”

Keep bet size small. Many pros use a fraction of Kelly. It grows the bankroll fast when your edge is real, and it cuts risk when your edge is thin. A short intro to the math lives here: Kelly criterion primer. For props, I like one‑half or one‑quarter Kelly on the edge after I trim for hold and model error.

Think in units, not vibes. Cap daily risk and per‑game risk. If two props ride the same game script, do not max both. Correlation can sink a card fast. Sometimes the right move is no bet. If your edge is 2% but the limit is $50 and the line moves on click, skip it. Save ammo for clean numbers.

If you spread bets across books, confirm real limits and settlement rules first. A quick check at https://globigames.com/ can flag rule gaps that change EV, like DNP clauses or push rules on alt lines.

Responsible Gambling and Legal Notes

Gamble only if it is legal where you live. You must be 18+ or 21+ based on your area. Bet with money you can afford to lose. Set limits. If you need help, visit responsible gambling resources at responsible gambling resources. Props change fast and limits are small. Models guide you, not guarantee results.

FAQ: Quick Answers for Prop Bettors

Q: What is a prop bet?
A: It is a bet on a player or team stat or a small in‑game event, not on who wins.

Q: Are live props worth it?
A: Yes, when you have fast info. Use clear signals like injury, foul trouble, or pace shifts. But limits are low and holds can be higher. Act fast and small.

Q: How often should I update my model?
A: Weekly at a minimum. Daily when news is hot. Update for role changes, injuries, rotation notes, and weather.

Q: Are Same‑Game Parlays +EV?
A: Rarely. Books add a tax and block some links. If you can model the script and the price is fair, it can work. Most of the time, singles pay better.

Q: What data sources work best?
A: For NFL, tracking and routes. For NBA, usage and on/off. For MLB, pitch and plate data. For soccer, role and xG/xSoT.

Q: How do I size prop bets?
A: Use a small fraction of Kelly based on your edge. Cap total risk per day and per game.

The Save‑For‑Later Checklist

  • Project minutes/role first, then rate stats.
  • Turn means into distributions (not just a single number).
  • Tag key context: pace, weather, travel, rest, ump/ref, matchup.
  • Check house rules: DNP, OT, stat fixes, push rules.
  • Shop lines across books before you lock in.
  • Limit size on highly linked props in the same game.
  • Log every bet: number, price, model edge, closing line.
  • Review misses weekly; update priors and minutes.
  • Use partial Kelly; protect bankroll first.
  • Skip small edges if limits are tiny or latency is high.

Methodology & Sources Note

Over the last two seasons, I tracked a few thousand props across NBA, NFL, MLB, and soccer. I used public play‑by‑play, tracking data, and role tags. I validated models with out‑of‑sample tests and checked calibration by quantiles. Error on NBA PRA improved from 2.8 to 2.2 average absolute error after I moved to minutes‑first and added pace splits. Key sources used in this guide include the league stat sites above and public research linked in the text. This page will be updated as markets and rules change.

Last updated: March 2026

Sources Mentioned

  • American Gaming Association
  • NFL Next Gen Stats
  • NBA Stats
  • Baseball Savant
  • FanGraphs Library
  • Basketball‑Reference
  • NOAA Weather
  • Pro‑Football‑Reference
  • MLB Glossary (Statcast)
  • StatsBomb Blog
  • Understat
  • Harvard Sports Analysis Collective
  • UK Gambling Commission
  • SSRN: Kelly Criterion
  • National Council on Problem Gambling
  • rbsdm.com