NBA Prop Betting Calculators: EV Math and Line Shopping

Updated August 2026

Licensed

Available in US

Fast payouts

18+ Only

Table of Contents

1. ssential Prop Calculators: Implied Probability and No-Vig
2. Implied Probability: Reading What the Line Says the Odds Are
3. No-Vig Fair Value: Stripping the Bookmaker’s Cut
4. Expected Value Maths Step by Step
5. Kelly Criterion for Prop Sizing and Why Half-Kelly Is Smarter
6. Line Shopping Across UKGC-Licensed Sportsbooks
7. Building a Minimal Prop Tracking Sheet
8. Where Calculators Fail: Garbage In, Garbage Out
9. Numbers Stay Honest, Even When You Don’t

Editorial Team «nba Props Betting» June 9, 2026 Reading time: 25 min

ssential Prop Calculators: Implied Probability and No-Vig

I have a spreadsheet open on my second monitor every time I look at an NBA prop. Three calculators on three tabs. Implied probability on tab one. No-vig fair value on tab two. Expected value with a Kelly fraction on tab three. The whole thing took me about ninety minutes to build the first time, and roughly nine years of refinement since to make it useful. The principle behind it has not changed.

The principle is that the price the operator is showing you on a player prop has been engineered. It is not a guess at fair odds. It is a fair-odds estimate plus a margin loaded into the line so that, on average, the operator wins. Your job as a bettor is to strip the margin out, work out what the implied probability of the line actually says, compare it to your own probability estimate, and only place the bet when the gap between the two is large enough to overcome the margin and produce a long-run positive return.

The arithmetic is not hard. The arithmetic actually works. What separates bettors who survive a long season from bettors who do not is whether they have built the calculator habit, run the numbers every time, and respect what the numbers say even when they want to take the bet anyway. At standard prop pricing of 1.91 decimal — the equivalent of −110 in American format — you need a true win probability of at least 52.4% just to break even. Anything less than that, and the maths is grinding you down whether you feel it on a single bet or not.

This piece is the practical walk-through of the three calculators. We will go through implied probability conversion across the three odds formats, then strip the vig out of two-sided lines to get a fair-value estimate, then work through expected value step by step, then apply Kelly sizing properly — and yes, half-Kelly properly, which is what almost every responsible recreational bettor should be running. We will close with line shopping across UK-licensed sportsbooks, building a minimal tracking sheet, and the failure modes that come from trusting the maths without sanity-checking the inputs.

Applying these formulas becomes much more effective once you master the art of calculating NBA usage rates for star players.

Implied Probability: Reading What the Line Says the Odds Are

Quick test. A line is priced at 1.91 in decimal odds. What probability of winning is the operator implying? Answer in your head, then read on.

Implied probability is the conversion from odds to the percentage chance of winning the bet must have for the bet to be a break-even play. The formula in decimal odds is straightforward: divide one by the decimal price. For 1.91 odds, that is 1 / 1.91 = 0.524, or 52.4%. The operator is implying — through the price they have set — that this side of the bet wins 52.4% of the time. If your estimate of the true win probability is higher than that, you have a positive expected-value bet. If it is lower, you do not.

For UK fractional odds, the conversion has one extra step. Take the fraction, divide the denominator by the sum of numerator and denominator. A 10/11 line — equivalent of 1.91 decimal — works out as 11 / (10 + 11) = 11/21 = 0.524, the same 52.4%. A line of 4/5 (decimal 1.80) is 5 / (4 + 5) = 5/9 = 0.556. The conversion is identical in meaning, just a different arithmetic path to the same answer.

For American odds, you have two cases depending on whether the price is positive or negative. Negative American odds — the kind you usually see on prop favourites — convert as: |line| divided by (|line| + 100). So −110 becomes 110 / (110 + 100) = 110 / 210 = 0.524, again 52.4%. Positive American odds convert as: 100 divided by (line + 100). So +110 becomes 100 / 210 = 0.476, or 47.6%. Two sides of the same vigged line, adding up to more than 100% precisely because the vig is built into both prices.

The reason the implied probability matters is that it is the threshold against which your own projection has to clear. If your model says the player has a 56% chance of going over his points line, and the price is 1.91 with implied probability 52.4%, you have a 3.6 percentage point gap of perceived edge. Whether that edge is real depends entirely on whether your model is well calibrated, which we will get to. But the implied probability is the floor under everything else.

One subtle point that catches bettors out. Implied probability includes the vig. It is not an unbiased estimate of the operator’s “true” view of the probability — it is the price at which the operator is willing to take action while preserving their margin. The “true” probability the operator implicitly believes in is somewhere a little below the implied figure on the side they expect to win, and a little above on the other side. To get to the operator’s true view, you have to strip the vig out. That is the next calculator.

For UK bettors looking at a typical prop slip, decimal odds are the easiest format to run these calculations in. If you are reading a line in fractional or American format, convert to decimal first. The conversion is mechanical, and your calculator gets simpler. Decimal at 1.91 is intuitive — the inverse is 0.524, you are above or below it on your projection, you know what to do.

No-Vig Fair Value: Stripping the Bookmaker’s Cut

Picture a coin-flip game. Heads pays 1.91, tails pays 1.91. Both implied probabilities are 52.4%. They sum to 104.8%. That extra 4.8% is the vig — the operator’s margin built into a market that should sum to exactly 100% in a fair game. When you place a bet on a vigged line, you are paying that 4.8% as a cost of doing business.

The no-vig fair value calculation is how you find what the line would be priced at if there were no operator margin. It is not the same as your projection of the true probability — it is the operator’s implied two-sided probability with the vig stripped out. For a clean two-way market with both sides at 1.91, the no-vig probability for each side is 52.4% / 104.8% = 50.0%, exactly as you would expect on a fair coin flip.

The general formula for two-sided no-vig probability is: implied probability of the side you care about, divided by the sum of implied probabilities of both sides. If side A is priced at 1.80 (implied 55.6%) and side B is priced at 2.05 (implied 48.8%), the sum is 104.4%. The no-vig probability of side A is 55.6% / 104.4% = 53.3%. The no-vig probability of side B is 48.8% / 104.4% = 46.7%. Those two now sum to exactly 100%, as they should.

Why this matters in practice. The no-vig fair value gives you the operator’s implicit best estimate of the true probability — what the operator believes after stripping out their margin. If your own projection of the true probability is significantly higher than the no-vig fair value, you have an edge. If your projection equals the no-vig fair value, you have no edge regardless of what the implied probability says, because the operator’s true view matches yours.

The classic mistake is to compare your projection directly to the implied probability of the side you want to bet, which automatically inflates your perceived edge by the vig. A 56% projection compared to a 52.4% implied probability looks like a 3.6 percentage point edge, but if the no-vig fair value of that side is 53.3%, your real edge is only 2.7 percentage points. The vig makes every line look more biddable than it actually is. Strip it before deciding.

One complication that pops up on prop markets specifically. Not every UK book offers a clean two-sided over/under at quoted prices on every prop. Some offer only the over with the implied under priced behind it, or the lines are slightly asymmetric, or the under is priced on a different line. When the two sides are not symmetrical at the same line, the no-vig calculation gets messier. The workaround I use is to find a UK book that does offer a clean two-sided market at the same line, run the no-vig there, and use that as the benchmark for evaluating prices at other operators.

For a deeper walk-through of the formula, including the algebraic derivation and the edge cases at three-way and multi-way markets, our companion piece on no-vig fair value formulas goes through the maths in full. For day-to-day prop betting, the two-sided formula above covers about 95% of what you actually need.

Expected Value Maths Step by Step

Here is the simplest definition of expected value I know, paraphrased from a maths-side analyst whose work I respect: you calculate expected value by multiplying the probability of each outcome by its payout, then adding the results together — and if the total is positive, the bet is +EV. That is the whole concept. Most of what comes next is just arithmetic.

The formula in its standard form is: EV = (probability of win × profit if win) − (probability of loss × stake). For a 1.91 decimal bet at a 1-unit stake with a true win probability of 56%, you compute (0.56 × 0.91) − (0.44 × 1.00) = 0.5096 − 0.44 = +0.0696. That is positive expected value of about 7 pence per 1-unit stake. In ROI terms, you are looking at +6.96% in long-run return, assuming your 56% probability estimate is accurate.

At the standard 1.91 price, the break-even threshold of 52.4% true win probability is the line below which expected value flips negative. At 50% true probability — a coin flip on a vigged line — your EV is (0.5 × 0.91) − (0.5 × 1.00) = −0.045, or −4.5%. That is the cost of paying the vig on a market where you have no edge. Every bet placed at this implied breakeven on a coin-flip-true-probability market drains roughly 4.5p per pound staked over time.

The Wizard of Odds analytical framework on this point is worth internalising: at +5% EV on a stake of $1 with odds of +110 (1.91 decimal), the ROI is 5% — meaning a profit of 5 cents on every dollar staked in the long run. The simplicity of the maths is the point. The complexity is in being honest about your true probability estimate, because that is the input that drives everything.

The honest framing for an EV calculation is: this is the expected return assuming my probability estimate is correct. If my estimate is off, the EV calculation is off by the same amount. The discipline is to estimate conservatively. If your model spits out a 57% probability, run the EV at 55% as well to see how thin your margin is. If the bet is still positive at 55%, the call has built-in tolerance for projection error. If it flips negative at 55%, you are betting on the precision of your own estimate, and your estimate is almost certainly less precise than you think.

This is where I keep coming back to a discipline reminder I stole from a piece of editorial writing on the maths side: uncertainty is part of the analysis, and pretending you know the true probability with a precision you do not have is the fastest way to lose. Every projection has an error band. Treat the EV calculation as an expected value across the error band, not as a single-point estimate.

The longer-run consequence of running EV maths consistently is that you will reject most of the bets that look good. Most projections produce small edges, and small edges get eaten by vig and variance. The bets that survive an honest EV check are a small fraction of the bets you might consider, and that pruning is the edge. Volume is not strategy. EV-positive volume is strategy. The rest is gambling.

Kelly Criterion for Prop Sizing and Why Half-Kelly Is Smarter

You have done the maths. The bet has positive expected value. How much of your bankroll should you put on it? The Kelly criterion gives you the theoretically optimal answer for compounding bankroll growth, and the answer is almost always too aggressive for an actual recreational bettor.

The Kelly formula in its decimal-odds-friendly form is: fraction of bankroll to stake = (probability of win × decimal odds − 1) / (decimal odds − 1). For a bet at 1.91 decimal with a true win probability of 56%, the Kelly fraction is ((0.56 × 1.91) − 1) / (1.91 − 1) = (1.0696 − 1) / 0.91 = 0.0765, or about 7.65% of bankroll on this single bet. That is a big number for a single prop bet. It assumes you are right about the 56% probability, you have no other simultaneous bets correlated with this one, and you are comfortable with the variance.

The reasons full Kelly is rarely the right answer in practice are stacked. Probability estimation error is the biggest one — the formula assumes your probability is exact, and any overestimate of edge gets multiplied directly into stake size. A second issue is variance: full Kelly produces high return variance, and even with positive EV the drawdowns are uncomfortable enough to make most bettors deviate from the framework after a bad run. A third issue is that prop betting is typically not a single-bet-at-a-time activity — most slips have multiple correlated positions, and the correlation makes the Kelly maths underestimate true risk.

The standard fix is half-Kelly: stake half of what the full formula tells you. For our example, that becomes 7.65% / 2 = 3.83% of bankroll. Half-Kelly captures the bulk of the long-run growth advantage of full Kelly while substantially reducing variance and providing a buffer against probability estimation error. The technical maths-side argument is that running half-Kelly when your true edge equals your estimated edge produces about 75% of full-Kelly’s expected log-bankroll growth at roughly half the variance — a much better risk-adjusted profile.

For UK recreational bettors I would actually go further. A quarter-Kelly is plenty conservative, particularly if your probability estimates have not been calibrated against years of logged results. A 1.9% stake on a 7.65% full-Kelly recommendation is unaggressive enough to absorb the inevitable model errors without producing a bankroll death spiral when a few of them hit at once.

One specific Kelly trap I have watched bettors fall into. The formula assumes the probability you input is the true probability, not your estimated probability. If your model has been overestimating edge by 10% on average — a perfectly normal calibration error — then Kelly will overestimate stake size by something like 25-35% on average across the bet history. Half-Kelly turns that into a 10-15% overstake; quarter-Kelly into 5-8%. The buffer matters.

The longer-run discipline is to log every bet’s predicted probability, the price, the realised result and the implied stake size, and then check periodically whether your predicted probabilities cluster correctly against realised win rates. If your “55%” predictions win 55% of the time across a hundred-bet sample, your model is calibrated and you can trust the Kelly maths. If your “55%” predictions win 49% of the time, your model is biased and your Kelly stakes have been too aggressive. The log is the calibration tool.

Line Shopping Across UKGC-Licensed Sportsbooks

Two UK-licensed operators in 2026 are pricing the same player’s points over at different decimal odds — say, 1.91 at one and 1.95 at the other. Same player. Same line. Same fixture. Different prices. If you do not place the bet at the better price, you have left edge on the table. Line shopping is the simplest, lowest-skill, highest-impact discipline in prop betting, and most bettors do not bother because it requires opening a second tab.

The reason different UK operators price the same prop differently is that each operator’s risk book runs on slightly different inputs. They have different trading desks, different historical data inputs, different models and different exposure profiles to particular players or markets. Lines do not move in perfect sync. Operators correct toward each other over the course of a betting window, but during the window itself the gaps can be material.

The market context matters. Ladbrokes commands roughly 36.8% of UK Sports Betting PPC click share, and Sky Bet sits at 23.85%. The PPC dominance translates loosely to brand visibility and customer acquisition volume, not directly to line-pricing competitiveness. Some smaller operators have sharper opening lines than the market leaders precisely because they are competing on price rather than brand. Identifying which operators run sharper books for NBA props specifically is a process of observation across a season, not something you can read off a league table.

The mechanics of line shopping are simple. For any prop you are seriously considering, check the price across at least three UK-licensed sportsbooks before placing. Use the best price among them. The arithmetic on this is non-trivial — moving from 1.91 to 1.95 on a single bet improves your EV by roughly 2 percentage points, which compounds across a season into substantial bankroll difference. Pence here, pence there, it adds up.

Closing-line value is the related concept worth tracking. The closing line — the price the bet was at when the market closed — is the market’s final consensus on fair value. If you consistently bet at prices better than closing, you have a real edge. If your bets consistently move toward you between placement and close, you are reading the market accurately and the line is following you. Logging the closing line on every bet is the single most informative metric for whether you are sharp or whether you got lucky.

A practical note for UK bettors. Line shopping requires having accounts at multiple UK-licensed sportsbooks and being comfortable moving between them. Identity verification, deposit limits and KYC checks all apply per operator under UKGC rules, so building out a multi-operator workflow is not zero-cost in time terms. But the EV improvement is real, and once the accounts are set up and verified, the per-bet workflow is just a tab-switch.

Building a Minimal Prop Tracking Sheet

The simplest prop tracking sheet I run has eight columns and has been the single most valuable tool I have ever built. Date. Player and stat. Line. Operator. Price. My estimated probability. Stake. Result. That is it. Eight columns, one row per bet, one new row every time I place anything. The sheet does not need anything fancy. The sheet needs to be filled in every single time.

The columns I add for diagnostic purposes — and that pay off after about three months of consistent logging — are: closing line at the operator I bet through, no-vig fair value at placement, my predicted EV, and a brief note on the rationale. With those four extra columns, the sheet becomes a calibration tool. I can pull up every bet I made at predicted probability between 55% and 60% and see what fraction of them actually won. If the realised rate matches the predicted rate within reason, my model is calibrated. If it doesn’t, I know which direction to adjust.

The single most useful diagnostic the sheet produces is the closing-line-value comparison. Every bet I place gets compared to the closing line at the same operator on the same fixture. If my placement price beat the close by an average of 2 percentage points across a hundred-bet sample, I am beating the market consensus by a margin that explains my long-run profit. If my placement price lost to the close on average, I am betting at worse-than-consensus prices and my long-run results will reflect that even if individual bets cash.

The sheet also catches behavioural drift. If I notice that my recent stakes have been climbing without a corresponding climb in my predicted edges, I know I have started overstaking — usually because of a recent win streak. The sheet is the antidote to that. Numbers do not flatter you. They tell you what you actually did.

One final feature I added a year in is a “would I bet this again?” column, filled in retrospectively about a week after the result lands. The retrospective view is different from the in-the-moment view. Bets that felt obvious at placement sometimes look reckless in hindsight; bets that felt marginal sometimes look perfectly reasoned. The retrospective discipline is the closest thing to objective self-assessment a betting record can produce.

Where Calculators Fail: Garbage In, Garbage Out

I have spent ten years building progressively better prop calculators, and the most important thing I have learned about them is that they are downstream of the inputs. A perfectly engineered EV calculation with a perfectly correct Kelly fraction is meaningless if the probability estimate I fed into it is wrong. The calculator can only ever be as accurate as the projection that drives it.

The most common projection failure I see in bettors who have just discovered the maths is overconfidence in small samples. They watch a player put up three big games in a row and update their probability estimate from 52% to 60% based on three observations. The maths then dutifully produces a strong positive EV and a meaningful Kelly stake. The bet loses, because three games is not signal, and the calculator has just translated overconfidence directly into stake size.

The second common failure is forgetting that the input probability has to include all the relevant scenarios. A model that estimates probability of clearing the line based on average minutes ignores the chance of a blowout that compresses fourth-quarter minutes, the chance of foul trouble, the chance of a late scratch the bettor did not see. Each of those is a small adjustment downward on the over probability, but together they often shave 2-3 percentage points off the raw model output.

The third failure is correlation between bets on the same slip. A bet builder with five legs all on the same player creates correlated outcomes — if the player is hot, all five legs win; if he is cold, all five lose. Standard EV maths assumes independence, and the calculator will produce wildly inflated edges on highly correlated parlays. Bet builders are priced by operators who know exactly how to exploit this misperception.

The discipline that protects against these failures is the same one I have been pushing throughout: log everything, calibrate against realised results, and discount your raw model output toward conservatism. The calculator gives you the arithmetic. The discipline gives you the inputs that the arithmetic deserves.

Numbers Stay Honest, Even When You Don’t

The thing I keep returning to about the calculator stack is that it is not the strategy — the strategy is reading the matchup, projecting the player and finding edges in the inputs. The calculator stack is the audit. It is what tells you, in plain arithmetic, whether the bet you want to place is the bet you should place. Most of the time the answer is no. That is what a good calculator does. It rejects more than it approves.

Use these calculators daily alongside the insights provided by the top NBA props site to maximize your expected value.

The bettors I have watched lose money over the long run almost always have one thing in common: they ran the maths inconsistently or not at all. The bettors who survive run the maths every time, accept the answer when it says no, and refine their inputs every few months against logged results. The arithmetic is honest. The honesty in your inputs is what determines whether the arithmetic produces signal or noise. Build the habit, build the sheet, and the rest of the work — the projection, the matchup, the edge — sits on a foundation that does not betray you when the variance arrives.

Published by the nba Props Betting team.