What does coefficient mean?
The short version of coefficient meaning: a coefficient is a number that multiplies something else. In algebra, the 3 in 3x is a coefficient. In betting, that same idea shows up in two very different places, and mixing them up costs people money.
First, plenty of sportsbooks (especially outside North America) call the odds themselves “coefficients.” A 1.91 coefficient just means decimal odds of 1.91. Second, in any statistical model, a coefficient is the weight attached to an input. Pace, rest days, opponent defense: each gets a number that says how much it moves the projection.
Same word, two jobs. One is a payout multiplier. The other is a measure of influence. Below we’ll separate them, then show how coefficients actually change how you read a player prop.
Coefficient as odds: the payout multiplier
When a European or Latin American book shows “coefficient 2.40,” that’s decimal odds. Multiply your stake by it and you get the total return, stake included. A $10 bet at 2.40 returns $24, so $14 profit.
The useful move is converting that multiplier into an implied probability. Divide 1 by the decimal odds:
- Coefficient 1.50 → 1 / 1.50 = 66.7% implied
- Coefficient 1.91 → 1 / 1.91 = 52.4% implied
- Coefficient 2.40 → 1 / 2.40 = 41.7% implied
- Coefficient 4.00 → 1 / 4.00 = 25.0% implied
That 52.4% figure is the one to memorize. It’s the break-even rate on a standard -110 line, and it’s why the vig quietly eats casual bettors. We broke the math down in our implied probability guide, and it’s worth ten minutes of your time.
Here’s the trap: implied probabilities across a market add up to more than 100%. That surplus is the book’s margin. To get a fair price you have to strip it out, a process called de-vigging. StatsBench’s Positive EV Scanner does this with the power method, then compares the fair number against soft-book prices.
Coefficient as model weight: how much an input matters
In a regression model, each coefficient tells you how much the output changes when one input moves by one unit, holding everything else steady. If a WNBA rebound model gives “opponent pace” a coefficient of 0.08, then a one-possession bump in pace adds roughly 0.08 rebounds to the projection.
Three things to check before you trust any coefficient:
- Sign. Positive means the input pushes the projection up. Negative means down. A wrong sign usually means the model is picking up noise, not signal.
- Size. A coefficient of 0.08 rebounds is real but tiny. It won’t move a 9.5 line on its own.
- Sample. Coefficients fit on 12 games are guesses wearing a lab coat. Our piece on sample size in betting covers where the line sits.
You’ll also run into the correlation coefficient, usually written as r. It runs from -1 to +1 and measures how tightly two variables move together. That one matters a lot for parlays, since correlated legs don’t multiply the way books’ payout math assumes. The correlation explainer goes deeper.
Why coefficients get abused in betting content
A coefficient is only as good as the model behind it. Feed a regression enough variables and it will hand you confident-looking weights for things that don’t matter at all. Statisticians call that overfitting. Bettors call it a bad week.
Watch for these red flags:
- A model with more inputs than it has games of data.
- Coefficients quoted without any error range or confidence interval.
- Weights that flip sign every time the model gets refit.
- Inputs that overlap heavily (points and PRA in the same model, for instance).
Honest framing sounds like this: “pace explains part of the variance, matchup explains more, and a chunk stays unexplained.” Nobody’s coefficient set predicts a basketball game. It just tilts the odds a little, and a little is all you need over hundreds of bets.
Reading coefficients on a real prop slate
Here’s where the abstract stuff gets concrete. Take a snapshot of StatsBench cheatsheet props from a July 2026 WNBA slate, Washington hosting Connecticut. Four Shakira Austin markets showed the same 60% hit rate over her last 10 games, but the prices and edges were not identical.
| Prop | Line | Hit rate (L10) | Edge | Best price |
|---|---|---|---|---|
| Points + rebounds + assists | o/u 28.5 | 60% (6/10) | 1.2% | +128 |
| Points | o/u 15.5 | 60% (6/10) | 1.3% | +115 |
| Rebounds + assists | o/u 11.5 | 60% (6/10) | 1.4% | -110 |
| Rebounds | o/u 9.5 | 60% (6/10) | 1.1% | -110 |
Convert those prices to coefficients and the picture sharpens. +128 is a decimal coefficient of 2.28, implying about 43.9%. The -110 markets sit at 1.91, implying 52.4%. Same player, same hit rate, wildly different payout multipliers.
That gap is the whole point. A 60% historical rate on a market priced near 43.9% implied is a different proposition than 60% against 52.4% implied. Ten games is a small sample, so treat both as a starting question, not an answer.
The edge number is a coefficient too
StatsBench’s edge percentage is essentially a scaled difference between the fair price and the offered price. On that same July 2026 Washington slate, teammate Kiki Iriafen’s rebounds o/u 9.5 carried a 60% hit rate but a negative edge of -0.3% at -110, while her double-double market at +125 sat at roughly break-even, 0.1%.
Two takeaways from that pair:
- Hit rate and edge are separate numbers. A 60% clip means nothing if the price already assumes 60%.
- Small edges are normal. Anything advertised as a 25% edge on a mainstream WNBA market usually means stale data, not a gift.
Price first, trend second. That ordering is the habit that separates grinders from people chasing hot streaks. We wrote about why raw recent-form splits mislead in this piece on last-10-games traps.
How to actually use this
You don’t need to fit your own regression. You need three habits.
1. Convert every price to a probability
Before anything else, turn the coefficient or American odds into an implied percentage. If your own estimate of the player’s true rate isn’t clearly higher, pass. Most props fail here.
2. Shop the multiplier
The same prop can be +115 at one book and +105 at another. That difference is pure return with no extra risk. StatsBench tracks roughly 60 sportsbooks, including Pinnacle, FanDuel, DraftKings and prediction-market exchanges like Kalshi and Polymarket, and surfaces the best available price.
3. Respect the sample behind the weight
Six-for-ten is six-for-ten. It’s suggestive, not conclusive. Look for props where the hit rate, the matchup data and the consistency grade point the same direction, then check that the price hasn’t already caught up.
For a broader statistical vocabulary, the Wikipedia entry on coefficients covers the math cleanly, and WNBA.com’s official stats hub is a solid free source for raw box-score inputs.
Where the tools fit
Doing this by hand across a full slate is tedious. That’s the job the Positive EV Scanner handles: it recomputes fair prices from sharp books roughly every minute, de-vigs them, and flags where a soft book is lagging. Kelly staking is built in, with a 5% bankroll cap.
If you’d rather work off hit rates and matchups directly, the Prop Finder lets you filter by SB Score, EV%, consistency grade and defensive matchup. WNBA bettors can start with the WNBA props preview, and there’s a full WNBA player props guide if you want the sport-specific version.
One honest note. Positive expected value is a long-run statistical edge, not a promise. Losing weeks happen even when every number is on your side. Bet only what you can afford to lose, keep it fun, 21+ only, and call 1-800-GAMBLER if betting stops feeling like a hobby.
Start with the price, not the story
Coefficients are just weights. Some weight your payout, some weight a projection, and both deserve a skeptical look before money moves. Once you’re converting odds to probabilities automatically, most bad bets disqualify themselves.
Ready to see de-vigged fair prices next to live sportsbook lines? The Positive EV Scanner updates every minute across ~60 books. The free tier shows edges up to 0.5%, which is plenty to see how the math works before you commit.
Frequently Asked Questions
What is the simplest definition of a coefficient?
A coefficient is a number that multiplies another quantity. In 3x, the 3 is the coefficient. In betting it either multiplies your stake (decimal odds) or weights an input inside a statistical model.
Are betting coefficients the same as odds?
Often yes. Many sportsbooks outside North America use “coefficient” as another word for decimal odds. A coefficient of 2.40 means a $10 bet returns $24 total, and it implies roughly a 41.7% chance before removing the book’s margin.
How do I convert a coefficient into a probability?
Divide 1 by the decimal coefficient. A 1.91 coefficient gives 1 / 1.91 = 52.4% implied probability. Remember that implied probabilities across a market sum to over 100%, and that surplus is the vig.
What is a correlation coefficient in betting?
It’s a number between -1 and +1 that measures how closely two variables move together. It matters most for same game parlays, where correlated legs are not truly independent and the standard payout math overstates your return.
Does a big model coefficient mean a good bet?
No. A large coefficient only means an input has a strong effect inside that model. The bet is only good if the market price hasn’t already accounted for that effect, which is what expected value measures.