Being right about the direction is not enough to make the price worth taking.
A bettor projects 48 points and sees Over 47.5 on the board. The choice looks obvious—until the Over is priced at -110. That price requires a win probability of about 52.4% just to break even, not merely a forecast half a point above the line.
The projection also carries uncertainty. If plausible outcomes are widely dispersed, 48 may translate into only a slim chance above 50% that the total clears 47.5. Sportsbook margin raises the hurdle further by inflating the implied probabilities on both sides. A wager becomes positive expected value only when the bettor’s estimated probability exceeds the break-even probability at the available odds. A likely winner can still be a poor bet.
Read the complete market quote
A usable totals quote has three parts: the line, the price on each side, and the settlement rules. The line defines the scoring threshold; the prices reveal the sportsbook’s weighting and juice. For a broader grounding, see how Over/Under betting works.
- Over: wins when the final total finishes above the line.
- Under: wins when it finishes below the line.
- Juice: the sportsbook’s margin embedded in the odds.
- Stake: the amount risked; profit is the net amount won, excluding the returned stake.
- Push: a tie with the line, usually resulting in the stake being refunded.
| Offer | Over | Under | Practical difference |
|---|---|---|---|
| Total 47.5 | -120 | +100 | Over costs $120 to profit $100 |
| Total 48.5 | -105 | -115 | Better Over price, but a harder line to beat |
| Total 48 | -110 | -110 | Exactly 48 may push |
Settlement terms can change the effective bet. Check whether overtime counts, how abandoned or postponed events are handled, and whether listed participants must start. Without those rules—and both prices—offers cannot be compared reliably.
Convert odds into break-even probability
American odds can be converted into an implied break-even probability with two formulas:
- Negative odds:
|odds| / (|odds| + 100) - Positive odds:
100 / (odds + 100)
At -110, the calculation is 110 / (110 + 100) = 0.5238, or 52.38%. A bettor staking $110 to win $100 therefore needs to win more than 52.38% of comparable bets to have positive expected value, assuming no pushes.
Account for the overround
If both sides of a total are -110, each has a 52.38% implied probability. Together they equal 104.76%, not 100%. The extra 4.76 percentage points are the overround, or the sportsbook’s built-in margin.
A simple proportional no-vig estimate divides each side’s implied probability by their combined total. For an asymmetric market:
- Over -120:
120 / 220 = 54.55% - Under +100:
100 / 200 = 50.00% - Combined implied probability: 104.55%
After proportional normalization, the Over is 54.55 / 104.55 = 52.18%, while the Under is 47.82%.
These no-vig figures describe the market’s relative pricing. They are not the bettor’s forecast. Value appears only when an independent estimate of the Over’s true probability exceeds the relevant break-even rate after accounting for settlement rules and uncertainty.
Build a distribution around the expected total
A projected average of 47.5 is only a starting point. Two forecasts can share that mean but have very different spreads: a volatile matchup may clear 48 far more often than a predictable one, even if their expected totals match.
Establish a baseline from expected possessions and scoring efficiency, or from a weighted blend of recent performance and season-long data. This provides a practical foundation to model game totals for better value. Then adjust only for information likely to affect scoring:
- Injuries: quarterback, offensive line, and defensive absences
- Weather: wind matters more than light rain
- Matchup: pace, red-zone efficiency, and explosive-play tendencies
- Venue and context: surface, altitude, travel, and game incentives
Finally, estimate the distribution using historical forecast errors or simulation. The betting decision depends on the probability above or below the market line—not merely whether the average differs from it.
Turn projections into probabilities
Once an expected total and realistic variance are set, simulation can estimate how often each side wins. For example, run 10,000 plausible game scores from the chosen distribution, then count outcomes above and below the sportsbook line. If 5,500 finish above 47.5, the estimated Over probability is 55%.
Whole-number lines require a separate count for pushes. An Over 47 simulation might produce 53% wins, 44% losses, and 3% pushes; expected value should preserve all three outcomes rather than treating pushes as wins or losses.
Historical comparisons offer a simpler alternative. Games with similar teams, venue, injuries, weather, and market expectations can form an empirical distribution, though small or outdated samples can mislead.
Pace belongs in the model only where it reflects scoring opportunities. Basketball possessions, football play volume, or baseball plate appearances work differently, so methods to adjust totals for sport-specific pace should match the competition rather than apply one blanket multiplier.
Calculate expected value
Expected value combines each possible outcome with its estimated probability:
EV = (P(win) × win profit) − (P(loss) × stake) + (P(push) × $0)
A push returns the original stake, so it creates neither profit nor loss. Its EV contribution is therefore zero, although its probability still matters when the line can land on a whole number. The same framework can calculate EV for prop bets when their probabilities and payouts are known.
Example: $100 on Over 47.5 at -110
A $100 stake at -110 produces $90.91 in profit on a win. Because 47.5 cannot push, a 55% win estimate leaves a 45% loss probability:
- Win contribution: 0.55 × $90.91 = $50.00
- Loss contribution: 0.45 × $100 = $45.00
- Expected value: $50.00 − $45.00 = +$5.00
The bet has an expected return of $5 per $100 staked, or 5%. That does not mean the next wager should earn $5: it will either win $90.91 or lose $100. EV describes the average result expected across many comparable bets, assuming the 55% estimate is accurate.
Stress-test the estimated edge
At -110 odds, a $100 stake returns $90.91 profit when successful. Expected value changes quickly around the 52.4% break-even point:
| Estimated win probability | EV per $100 staked |
|---|---|
| 51% | -$2.64 |
| 53% | +$1.18 |
| 55% | +$5.00 |
A 53% estimate appears profitable, but an error of just one percentage point makes it negative. Even a 55% projection falls to roughly +$1.18 EV if the true probability is 53%.
Thin edges deserve extra skepticism when the model uses small samples, uncertain injury information, unstable lineups, or simplified assumptions about pace and scoring variance. A practical filter is to require a minimum projected probability or EV before betting. The threshold should increase as the inputs become less reliable; for example, a cautious bettor might reject marginal +EV plays and consider only estimates that remain profitable after reducing the forecast by one or two percentage points.
A positive EV calculation reflects the supplied probability—not certainty. Test weaker versions of the forecast before accepting the edge.
Compare the line and price
Totals cannot be compared by odds alone. Over 47.5 at -110 wins when the game lands on 48, while Over 48 at -105 only pushes. That half-point may be worth more than the five-cent discount.
Suppose a projection assigns 52% probability to scoring above 48, 3% to exactly 48, and 45% to below 48. For a $100 stake:
- Over 47.5 at -110: 55% wins, producing $5.00 EV.
- Over 48 at -105: 52% wins and 3% pushes, producing about $4.52 EV.
The apparently expensive -110 quote is slightly better under that forecast. But when the total is identical, price has a direct effect. If Over 48.5 has a 53% win probability, -105 returns about $3.48 EV per $100, while -115 returns about -$0.91—turning the same opinion from positive to negative.
Recalculate EV for every available quote rather than treating all Overs at a given number as equivalent. Before betting, compare the projection with consensus prices and review sportsbooks known for sharper totals lines. A large disagreement can signal value, but it can also expose stale injury news, lineup assumptions, or an overly confident model.
Turn each totals opinion into a recorded decision
- Freeze the forecast
Record the projected total, probability of Over and Under, model version, and key assumptions before checking whether the price looks attractive.
- Capture the complete offer
Log the sportsbook, timestamp, total, odds, push rules, and stake limits. Quotes can move quickly, so an undocumented number is difficult to evaluate later.
- Calculate the decision margin
Convert the offered odds to break-even probability, calculate EV, and subtract that threshold from the forecast probability. Bet only when the gap exceeds a preset allowance for model and market uncertainty.
- Record the decision
Mark the wager as bet or pass, with the intended stake and a brief reason. Keeping passes prevents reviews from becoming biased toward memorable action.
- Add the close and result
Log the closing total and price, then the outcome, profit, or push. After a meaningful sample, review closing-line value and compare predicted probability bands with actual win rates; results alone can be noisy.
An Over or Under opinion becomes +EV only when its estimated probability clears the price-implied threshold by a credible margin. Consistent records reveal whether that margin is real or merely optimistic.
Once a wager is placed, any adjustment should preserve expected value rather than simply reduce discomfort; use value-conscious hedging methods when exposure genuinely needs to change.