Probability and odds

Probability ranges from zero to one, or zero to 100%. Odds against compare failure probability with success probability. If success probability is 40%, failure is 60%, so odds against are 60:40 or 1.5 to 1. Fair decimal odds are one divided by probability, giving 2.5.

These are alternative representations of the same assumed chance. The arithmetic can be exact while the chance estimate is poor. For an uncertain real event, probability quality matters more than converting it to four decimals.

Decimal odds and return

Decimal odds multiply the stake to give total return including the stake when the selection wins. At 2.60 with R100, total return is R260 and net win profit is R160. If it loses, the net result is -R100. The calculator uses these two outcomes.

Different markets may quote fractional or American odds. Convert them carefully and check whether a platform displays total return or profit. Promotions, dead-heat rules, deductions and commission can change settlement from the simple decimal formula.

Implied probability

The reciprocal of decimal odds is the raw implied probability. Odds of 2.60 imply about 38.46%. This does not mean the event objectively has that probability. A bookmaker’s prices include margin, and a betting exchange can include commission.

For a market with several mutually exclusive outcomes, convert every price to implied probability and add them. A total above 100% is an overround before adjustments. Normalising the values is one way to remove margin, but it still does not reveal the true probabilities.

Expected net value

Expected value multiplies each net outcome by its probability and adds them. With a 40% chance, R160 win profit and 60% chance of losing R100, EV is 0.4 × 160 − 0.6 × 100 = R4. That is an average per identical opportunity under the assumptions.

It is not the amount expected from one bet. The actual single result is R160 or -R100 in this simplified case. Even a genuinely positive EV can produce long losing sequences. A small estimated edge can be overwhelmed by probability error, commission or limits.

Edge in percentage points

The page subtracts offered implied probability from the user’s estimate. Forty per cent minus 38.46% is 1.54 percentage points. Percentage points are not the same as percent change. The result does not adjust for other outcomes’ margin.

An edge entered by intuition is not evidence. Keep forecasts before outcomes occur and score calibration over many independent events. Rewriting an old probability after seeing the result creates hindsight bias and makes performance look better than it was.

Independence and repeated trials

Multiplying EV by many bets assumes the probabilities, prices and settlement are correctly represented. Outcomes can be correlated, especially when bets depend on the same match, team, weather or economic event. Correlation increases concentration and invalidates simple diversification assumptions.

Bookmakers may also limit accounts, change odds or void markets, so identical repetition may be impossible. A theoretical long run is not a guarantee that a person has enough capital or time to experience the average.

Variance and bankroll risk

Two wagers can have the same EV but different outcome distributions. A long-shot price wins rarely and can create deeper losing runs. Expected value alone says nothing about the probability of ruin. Staking a large share of available money can cause catastrophic loss even with a positive model.

No staking system turns a negative expectation into a positive one. Doubling after losses increases exposure and eventually meets table limits or available funds. Money needed for housing, food, debt, health or family obligations should never be staked.

Data and model validation

A probability model should use information available before the event, avoid leakage from future data and be tested on an out-of-sample period. Calibration asks whether events assigned 40% occur about 40% of the time across a suitable set. Discrimination and profitability are separate questions.

Markets and participants adapt. Historical advantage can disappear after rule, team, pricing or data changes. Include transaction costs and rejected bets in evaluation. Do not publish an accuracy claim from a small selected set of winners.

Non-gambling probability uses

The conversion also supports risk communication, forecasting and decision analysis. A project with 40% success has odds against 1.5 to 1. Expected monetary value can compare choices only when consequences are reasonably valued and risk tolerance, liquidity and non-financial outcomes are considered.

High-stakes medical, safety and legal decisions should not be reduced to a monetary EV without professional frameworks and ethical constraints. Probability expresses uncertainty; it does not decide what consequence is acceptable.

Gambling harm

Gambling can lead to debt, secrecy, conflict, impaired work and mental-health harm. Warning signs include chasing losses, borrowing to gamble, hiding activity, increasing stakes for excitement or being unable to stop. A positive number on this calculator is not permission to continue.

Set hard limits before participation, never use credit and take breaks. If gambling is causing harm, use self-exclusion and seek confidential support from an appropriate South African service. In a crisis or risk of self-harm, contact emergency or crisis support immediately.

Promotions, scams and account terms

A bonus can impose turnover, minimum-odds, expiry and withdrawal conditions that change its effective value. Do not enter a promotional headline as ordinary decimal odds without modelling every condition. Free-bet stakes may not be returned on a win, making the payout formula different from the calculator’s standard stake.

Verify that an operator is lawfully authorised for the relevant jurisdiction and access official contact details independently. Guaranteed systems, paid tip groups, requests to share an account and demands for extra money to release winnings are warning signs. Never provide remote access, one-time passwords or borrowed identity documents to a betting contact.

Separate the estimate from the conversion

The arithmetic converts a probability into odds and expected value, but it cannot establish whether the input probability is credible. Record the data source, sample size, time period and assumptions behind that percentage. Outcomes may be dependent or conditions may change, so historical frequency is not automatically a valid forecast for the next event or wager.