Live betting: how in-game odds are priced
How in-game basketball odds are generated, why the square root of time governs how safe a lead is, what delays and suspensions do, and why live markets carry extra margin and extra temptation. Educational only.
This article is educational and describes how betting markets work. It is not advice, a prediction or an encouragement to bet. Sports betting is legal only in some places and only for adults, and over time the odds favor the house. If gambling is a problem for you or someone you know, call 1-800-GAMBLER in the United States, or look for the help line in your own country.
Pre-game betting is a slow market. A number is posted, it drifts through the day as news arrives, and by tip-off it has settled. Live betting, also called in-play betting, is the opposite. Once the ball goes up, the prices move constantly: after every basket, every turnover, every substitution. A team that was a seven-point favorite at the start can be a long underdog by the middle of the third quarter, and the sportsbook's screen shows the change as it happens.
Behind that moving screen is a machine. No trader is typing in new spreads after every possession. A pricing model reads a live feed of the game, updates an estimate of each team's chances, and publishes a fresh set of numbers many times a minute. Understanding that machine tells you why live prices move as they do, why they cost more than pre-game prices, and why some of the stories fans tell about "momentum" do not appear in the numbers. This article explains the mechanics in plain language, with a simple model you can use to sanity-check any live number. It does not recommend any bet.
What live markets cover
Live betting offers many of the same markets as pre-game betting, repriced as the game goes on.
- Live moneyline. Who wins, given the current score and time.
- Live spread. A fresh handicap based on the current margin and time left.
- Live total. An over/under on the combined final score, with the points already scored counted toward the result.
- Next-score and race markets. Which team reaches a certain number of points first, or who scores next.
- Quarter and half markets. Bets on the outcome of the current period.
- Live player props. Updated lines on individual players, adjusted for minutes played so far and foul trouble.
The principles are the same as for other markets: each price is a probability with a margin added, as described in our article on the vig and the hold. What changes is the speed at which the underlying probabilities change.
The core of the model: score, time and possession
At any moment, the most important inputs to a live price are the score and the time remaining. A team leading by 10 with 40 seconds to play is almost certain to win. The same team leading by 10 with most of the game still to play is a favorite but far from a certainty, not because 10 points is nothing but because a long stretch of uncertain basketball still stands between it and the finish.
A pricing model formalizes this with the idea that the remaining margin is random with a size that shrinks as time runs out. Think of the final margin as the current margin plus whatever happens over the rest of the game. Whatever happens in the remaining time has an expected value, based on the pre-game estimate of team strength scaled to the time left, and a spread of possible outcomes that depends on how much time remains.
The square root of time
The spread of outcomes grows with the square root of the time left, not in proportion to it. This is a standard feature of random processes built from many small independent events, and basketball scoring is a good fit for the idea. If the standard deviation of the final margin in a full game is about 12 points (a rough figure used here for illustration; the real number varies by league and era), then with a quarter of the game left, the standard deviation of the remaining margin is about 12 × √(1/4) = 6 points. With one-sixteenth of the game left, it is about 3 points.
Use this to see how safe a lead is. The table below gives a stylized estimate for a lead when the two teams are otherwise equal and the starting figure is a 12-point game-wide standard deviation. It is a simplification, not a betting model.
| Lead | Time left | Standard deviation of what remains | Approximate chance the leader wins |
|---|---|---|---|
| 10 points | 24 minutes | about 8.5 | about 88 percent |
| 10 points | 12 minutes | about 6.0 | about 95 percent |
| 10 points | 5 minutes | about 3.9 | about 99 percent |
| 5 points | 12 minutes | about 6.0 | about 80 percent |
| 5 points | 2 minutes | about 2.4 | about 98 percent |
The last row is worth a pause. A five-point lead with two minutes left in a stylized model is nearly safe. In real basketball, the final two minutes feature timeouts, deliberate fouling and a flurry of three-pointers, which make leads less safe than the smooth formula implies. Real models include those effects, which are a major reason that real late-game win probabilities are more cautious than this table. The qualitative point stands: leads become safe rapidly as time runs out, and the speed is faster than the intuition "a point per minute" would suggest.
Adding the pre-game estimate
The stylized table assumes equal teams. In practice, the model blends the current score with the pre-game spread, scaled to the time that remains. If a team was a six-point favorite before the game, its expected margin over the remaining minutes is six points times the fraction of the game left. As the game goes on, that prior fades, and the current score dominates. A good way to think about it: the pre-game number is the first draft, and the live score gradually replaces it. Our article on simulating a season and how win probabilities are built discusses similar methods at the scale of a full schedule.
Possession and other context
Beyond score and clock, models track who has the ball, the number of team fouls, which players are on the floor, and who is in foul trouble. A star on the bench with four fouls changes the expected output of the next few minutes. A team in the bonus gets free throws on every foul, which changes the expected value of a possession. These adjustments are small individually and meaningful in aggregate. They are also why a well-run model reprices after substitutions as well as scores. For how foul rules work, see the NBA's foul rules.
How live totals are priced
A live total is not a simple extrapolation of the current pace. If a game is on pace for 240 points after a fast first quarter, the live total is usually not 240. The model recognizes that early pace is a noisy estimate of the rest of the game and blends it with the pre-game expectation. Fast starts usually slow down, and slow starts usually speed up, as the model's mixture of the two sources suggests.
The model also accounts for what comes in the final minutes. Intentional fouling and free throws late in close games raise scoring, while a blowout may slow it. And overtime remains a possibility that adds to the combined score. Our article on totals explains the pre-game logic, and the live version builds on it with the points already scored as a fixed part of the final result.
Why live prices carry more margin
Live markets are costlier than pre-game markets for several reasons.
- The book is exposed to faster information. In a live game, someone watching the action may know of an injury or a lineup change a second or two before the price updates. The book defends itself by widening its margin.
- Lower limits. Maximum stakes are usually smaller because prices are less certain.
- Delays and suspensions. Books pause a market at moments of high uncertainty, such as a foul, a challenge or an injury. Wagers in flight may be rejected or delayed.
- Heavier margin per bet. The overround on a live market is commonly higher than on the standard pre-game spread. A bettor needs a bigger edge to break even.
- Many repeat bets. A fast-moving market invites a large number of wagers during a single game. Even at a modest margin per bet, the number of times that margin is paid is high.
The effect on a bettor is straightforward: the cost per bet is higher, and the number of bets tends to be higher. Both push toward bigger total costs. Our article on expected value explains why the cost per bet is the key quantity.
Delay, latency and the information race
The pricing machine depends on a data feed, and the feed has delays. Even official data takes time to reach a sportsbook, and the video a person watches on a screen may be seconds behind the arena. The gap between what is happening on court and what the model knows creates a risk for the book, and the book manages it in several ways.
- Bet acceptance delay. After a wager is submitted, the book may wait a few seconds before confirming it, then compare the price at that moment with the price when you clicked. If the price changed against the customer, the bet may be rejected or repriced.
- Market suspensions. When something important happens, such as a scoring play or a stoppage, the book can pause betting for a moment while the model catches up.
- Source monitoring. Books monitor where information is coming from and who is betting at what moment. They have rules and technology to deal with customers who appear to have access to faster information than the book's own feed.
For the ordinary fan, the practical consequence is that live betting is a game of information speed that happens at a scale of seconds, and the person watching a stream is, in the game's terms, usually slower than the people who set the prices. A fan who thinks a team "has the momentum" and then taps a button is acting on information that the pricing machine incorporated several seconds ago.
Momentum and the stories fans tell
Basketball is full of runs: a 12-0 stretch, a team that cannot miss, a crowd that comes alive. It is natural to believe these runs predict what comes next. Pricing models, on the whole, do not treat runs as strongly predictive. They update for the change in score and time, and they apply modest adjustments for personnel, but they generally assume that the next possession is not much different from what the teams' underlying quality implies.
That stance rests on a long-running debate. Studies of streak shooting and momentum in basketball have produced mixed and contested findings. The hot-hand idea has been studied for decades, and researchers have disagreed about how large any real effect is. What can be said is that if there is an effect, it is small enough that the market does not need to treat it as a large factor. A live price that moves a lot after a run is mostly reflecting the change in the score, not a belief that the run will continue.
That is useful to know for two reasons. First, a swing in the live line after a run is not necessarily a sign that the market has changed its mind about the teams. It is the arithmetic of a smaller or larger lead with the same time left. Second, anyone who bets on the idea that a run will continue is betting against a model that has already considered the possibility.
Reading live numbers without betting
Live odds can be used as a lens on a game even by someone who never places a wager.
- Look at the live moneyline as a win probability. After removing the margin, it gives you a fast read on how much the market thinks the current score matters. Our guide to moneylines explains the conversion.
- Compare to the pre-game line. If a team that was a seven-point favorite is now an even-money proposition after trailing by eight at halftime, you can see how the market weighs the first half against the full game.
- Watch for the market pausing. A suspension at an odd moment often points to something the broadcast has not shown yet, like a player injury.
- Use the live total to understand pace. If the live total is close to the pre-game total despite a fast start, the market is treating the pace as noise.
You can follow games as they happen on the scores page, and the pre-game numbers for comparison are on the odds page.
Responsible play and live betting
Live betting is the format most associated with problems of speed and repetition. Prices change every few seconds, a new bet is always one tap away, and a losing bet can be followed immediately by another. This is the pattern that can turn a few modest wagers into a much larger total, and it is the setting in which chasing a loss is easiest. Our article on bankroll management discusses setting limits before play begins, and responsible gambling: signs of a problem and where to get help lists the warning signs. Many operators offer tools such as deposit limits, time limits and cooling-off periods, and they are worth using.
What to take away
Live odds are produced by models that blend a pre-game estimate with the current score, the time left, possession and personnel. Because the spread of outcomes shrinks with the square root of time, leads become safe quickly as the clock runs down. Live markets carry more margin than pre-game ones, have lower limits, are subject to delays and suspensions, and give the book a speed advantage over the fan watching a screen. Momentum, as fans experience it, is largely absent from the numbers. The most useful thing a fan can get from live odds is a running read of how the market sees the game, and the most important thing to remember if you bet is that the speed itself is part of the product.
If betting stops being fun or starts to cost more than you can afford, stop and talk to someone. In the United States the National Problem Gambling Helpline is 1-800-GAMBLER, and most other countries have a similar service.