NBA totals, also called the over/under, are a bet on the combined score of both teams. The line starts from two things: how many possessions the game should have, and how many points each team should score per possession. In 2025-26 teams averaged 115.6 points a game, so about 231 combined (Basketball-Reference). Near the centre of a market, one point of line is worth roughly 2 percentage points of chance, which is why half a point and a few cents of price matter more than they look.
This guide shows how to build a total, turn it into a fair price, and account for overtime and the margin. If you're new to the market, start with NBA odds explained or the general over/under betting guide. The worked examples use invented numbers, not selections.
What an average NBA game looks like
Here's the baseline, from Basketball-Reference. The combined column is simply points per team doubled.
| Season | Points per team | Combined points | Pace | Offensive rating |
|---|---|---|---|---|
| 2014-15 | 100.0 | 200.0 | 93.9 | 105.6 |
| 2019-20 | 111.8 | 223.6 | 100.3 | 110.6 |
| 2024-25 | 113.8 | 227.6 | 98.8 | 114.5 |
| 2025-26 | 115.6 | 231.2 | 99.4 | 115.7 |
| 2026-27, as of early November | 113.4 | 226.8 | 99.6 | 113.2 |
Two terms do most of the work. Pace estimates a team's possessions per 48 minutes, and offensive rating is the points a team scores per 100 possessions (Basketball-Reference glossary). Defensive rating is the same measure for points allowed.
The current season's row is a small sample, so don't read too much into it yet.
How to build a total
The formula is short:
Total = possessions per team × (Team A points per 100 + Team B points per 100) ÷ 100
Test it on last season's league averages: a pace of 99.4 and a rating of 115.8 for each team (Basketball-Reference) give 99.4 × 231.6 ÷ 100 = 230.2. The real average was 231.2. The formula gets close, and part of the gap is overtime, because pace is counted per 48 minutes while real games sometimes run longer.
How wide the range is
Teams differ a lot. In 2025-26 (Basketball-Reference):
| Measure | Top of the league | Bottom of the league |
|---|---|---|
| Pace | Heat, 103.4 | Celtics, 94.8 |
| Offensive rating | Nuggets, 122.6 | Nets, 108.7 |
| Defensive rating (lower is better) | Thunder, 107.7 | Wizards, 122.7 |
Pace alone moves the number a long way. With both teams at league-average efficiency, a game played at the Heat's pace comes to about 239.5 points, and one at the Celtics' pace about 219.6. That's a gap of about 20 points before any difference in shooting or defence.
A worked example
Here are two invented teams. A simple shortcut is to average each team's offensive rating with the other team's defensive rating, and do the same for pace. It's a starting point we're using for illustration, not how a bookmaker sets a line.
| Input | Team A | Team B |
|---|---|---|
| Pace | 101 | 97 |
| Offensive rating | 118 | 113 |
| Defensive rating | 114 | 116 |
- Possessions. Average the paces: (101 + 97) ÷ 2 = 99.
- Team A per 100. Its offence against Team B's defence: (118 + 116) ÷ 2 = 117.
- Team B per 100. Its offence against Team A's defence: (113 + 114) ÷ 2 = 113.5.
- Total. 99 × (117 + 113.5) ÷ 100 = 228.2.
What moves the inputs
Season averages are only a start. They can't see who is playing tonight, whether a team is on the second night of a back-to-back, or whether a rotation has changed. If a key player sits, both the team's rating and the share of points left for teammates shift, which is where usage rate comes in.
Pace also tends to drift through a season. NBA.com's table shows league pace falling from the first 10 days to the full season in each of three straight seasons: 100.6 to 99.8, 101.1 to 99.2 and 100.5 to 99.6 (NBA.com). So early-season numbers can overstate possessions.
From a total to a fair price
A total on its own isn't a price. For that you need to know how far final scores tend to land from the expected figure. A common way to model this is the normal distribution, described by a centre and a spread called the standard deviation.
We couldn't find a solid published measurement of how far NBA finals land from the closing total, so the 18 points we use below is our own assumption. Check it against your own records. Here's what fair prices look like for a market centred on 225 points:
| Line | Chance of over | Fair over | Fair under |
|---|---|---|---|
| 220.5 | 59.9% | $1.67 | $2.49 |
| 222.5 | 55.5% | $1.80 | $2.25 |
| 223.5 | 53.3% | $1.88 | $2.14 |
| 224.5 | 51.1% | $1.96 | $2.05 |
| 225.5 | 48.9% | $2.05 | $1.96 |
| 226.5 | 46.7% | $2.14 | $1.88 |
| 227.5 | 44.5% | $2.25 | $1.80 |
| 229.5 | 40.1% | $2.49 | $1.67 |
Near the centre, each point of line is worth about 2.2 percentage points. With a narrower spread of 14 points it's about 2.9, and with a wider spread of 22 it's about 1.8. The same ladder is what sits behind NBA alternate lines.
Take the margin out first
Say a bookmaker lists over 224.5 at $1.87 and under 224.5 at $1.95. The implied chances are 53.48% and 51.28%, which add up to 104.76%, so the margin is 4.76%. Our guide to vig explains the idea, and the no-vig calculator on our calculators page does the sum.
Divide each chance by 104.76% and the fair chance of the over is about 51.05%. That's a useful detail: with uneven prices, the market's real centre isn't the printed line. Here it's about 225.0, half a point above 224.5.
The price also sets the bar you have to clear:
| Price | Chance you need to break even |
|---|---|
| $1.87 | 53.5% |
| $1.91 | 52.4% |
| $1.95 | 51.3% |
Your number against the market's number
Now put the example together. Your estimate says 228.2 and the market's centre is about 225.0, a gap of about 3 points. If your number were exactly right, the over at $1.87 would look generous. But that "if" carries the whole result, as this table shows:
| How much you trust your number | Centre used | Chance of over 224.5 | Fair over | Expected value at $1.87 |
|---|---|---|---|---|
| Fully | 228.2 | 58.1% | $1.72 | +8.7% |
| Half and half with the market | 226.6 | 54.6% | $1.83 | +2.1% |
| One quarter, market three quarters | 225.8 | 52.8% | $1.89 | -1.2% |
The same gap goes from a big edge to a small loss, depending only on how much you trust a spreadsheet built from season averages. A market total has usually absorbed team news and a lot of other people's estimates, so a gap like this more often means your number is missing something than that the line is wrong.
The research points the same way, although the samples are old. A study of more than 10,000 games from 2000-01 to 2007-08 found betting overs was a fair bet, consistent with an efficient market (The Sport Journal). Another found that 58.2% of Week 1 games finished under the total across the 2009 to 2012 seasons, while the market was generally efficient after that (Journal of Prediction Markets). Findings like that don't tend to stay put once they're published, so don't treat it as a system.
The better test is whether the prices you take tend to beat the close, which is what closing line value measures. A single total winning or losing tells you very little.
Overtime and landing on the number
Overtime is part of the total
An overtime period runs 5 minutes, against 48 minutes for the four regular periods (NBA rule 5). At last season's scoring rate, 5 extra minutes is worth about 24 points (231.2 × 5 ÷ 48). So a game heading under can finish over without anyone playing any differently.
How often does it happen? One analysis found overtime in 6.2% to 6.4% of games across 2013-14 to 2015-16 (RotoGrinders). That data is old and we couldn't find a newer public figure, so treat it as a rough rate. Using 6.3%, overtime adds about 1.5 points to the average game. In our model, on a line set at the regulation-time centre, it lifts the chance of the over from 50% to about 52.6%, which is more than the gap between $1.87 and $1.95.
Rules vary here. Many totals include overtime, but some markets are regulation time only (next.io). Read the market label before you compare two prices, as they might not be the same bet.
What happens if the total lands on the number
On a whole-number line, a final total matching the line is a push, and your stake is refunded (Action Network). Some operators grade pushes differently depending on the sport and market type (next.io), so check your bookmaker's rules. Half-point lines such as 224.5 can't land on the number.
In our model, a final total lands exactly on a line at the centre of the market about 2.2% of the time. That's the value of a whole number compared with a half point just beside it, and it's worth remembering when one bookmaker shows 225 and another 224.5.
A checklist before you take a total
- Check the settlement rules: overtime, and what happens on a whole number.
- Work out the margin, and the market's real centre, from both prices.
- Build your estimate from pace and efficiency, then adjust it for team news, rest and rotations.
- Give your estimate a range, not a single number, and work out the fair price at each end.
- If the bet only looks good at your most generous estimate, it isn't proven value.
- Compare the same line across bookmakers. Line shopping often matters as much as the estimate itself.
- Record the price you took and the closing price in a bet tracking spreadsheet.
If you also bet the line, our guide to NBA winning margin betting applies the same thinking to how far a team wins by.
See NBA totals side by side
BetBank's Odds Screen puts NBA lines, totals and player props from Australian bookmakers in one view, so you can see who has moved and who hasn't. If you haven't had a paid plan before, the NBA season special makes a weekly or monthly Silver or Gold plan half price for your first 2 months (the first 2 monthly or first 8 weekly payments), after the 7-day free trial, if you start by 30 November 2026.
BetBank is for adults 18+. Every bet can lose, a fair price doesn't make a bet a winner, and no method or tool removes the risk. Set a limit, take breaks, and gamble responsibly. If it stops being fun, our responsible gambling page lists free, confidential support, and BetStop lets you block yourself from all licensed Australian online and phone betting.