Hit Rate, ROI and Sample Size: Reading Prediction Results

Hit rate tells you how often settled selections won. Return on investment (ROI) tells you what those selections earned or lost relative to the amount staked. They answer different questions, so a high hit rate does not automatically mean a profitable record.

Sample size supplies the missing context. Six wins from ten decisions may look impressive, but ten results are still a small and volatile record. Read all three figures together, using the same market, date range, staking method and settlement rules.

Real Sport Insider shows these figures on results views such as the Football xG page. This guide explains what each figure means, which selections belong in each denominator and what the numbers cannot prove.

What is hit rate?

For a record containing only wins and losses, the calculation is wins ÷ (wins + losses) × 100. Six wins and four losses therefore produce a 60% hit rate.

A push, void or refunded selection is not a win or a loss. A pending selection has not been decided. Neither belongs in the win-rate denominator. If a page contains 6 wins, 4 losses, 1 push and 2 pending selections, it has 13 published picks but only 10 win-or-loss decisions. Its hit rate is still 6 ÷ 10 = 60%.

Why can a 60% hit rate still lose money?

Hit rate ignores the price attached to each selection. Imagine ten $10 bets at decimal odds of 1.50. Six win and four lose. The six winners return $90 in total, including the returned stakes, while all ten bets cost $100. The result is a $10 loss.

The hit rate is 60%, but the ROI is −10%. At odds of 1.50, the break-even hit rate is about 66.7% before any other costs or differences in staking. Winning more often than losing is not enough when the average winning return is small.

How is ROI calculated?

ROI is net profit or loss ÷ total amount staked × 100. In the example above, −$10 ÷ $100 × 100 equals −10%.

With a flat one-unit stake, a winning selection at decimal odds of 1.80 earns 0.80 units of profit; a losing selection loses 1 unit. The returned stake is not profit. If ten one-unit selections produce a net profit of 1.5 units, the ROI is 1.5 ÷ 10 = 15%.

A valid saved price is necessary for a trustworthy return calculation. If the odds are missing, you can grade the selection as a win or loss, but you should not invent the profit that would have been earned. Odds movement matters too: a result calculated from a recorded price may differ from the return available to someone who bet later.

Can a low hit rate be profitable?

Yes. Consider ten $10 bets at decimal odds of 3.00. If four win and six lose, the hit rate is only 40%. The four winners return $120, including their stakes, against $100 staked. Net profit is $20 and ROI is 20%.

This does not make long-odds betting automatically better. It only shows why hit rate must be interpreted alongside average odds and return. Different market prices require different break-even hit rates.

What does sample size tell you?

Sample size is the number of observations behind the reported figure. A 60% hit rate from 10 decisions means 6 wins. The same 60% from 500 decisions means 300 wins. The percentages match, but the second estimate contains much more information about the longer record.

Small samples can swing sharply. After a 6–4 start, one more loss changes the hit rate from 60% to 54.5%; one more win changes it to 63.6%. Larger samples usually reduce that sensitivity, but they do not remove bias, guarantee future results or turn poor data into good evidence.

The NIST explanation of binomial proportions defines the observed proportion as successes divided by trials and notes that common approximations can be inaccurate with small samples. That is why a short streak should be presented as a record, not as proof of a stable winning rate.

How should results be compared?

Compare like with like. Use the same settlement status, market, date range, odds source and staking rule. A record built from flat one-unit stakes cannot be compared directly with a record where stake sizes change unless the method is clearly disclosed.

Do not remove losses, count pushes as wins or mix pending selections into the hit-rate denominator. Do not select only the strongest league after seeing its results and treat that filtered run as a general forecast. A Top Leagues panel is a historical view of the displayed graded picks; it is not a promise that the same rate will continue.

How do Real Sport Insider result summaries work?

On the daily Football xG results summary, Win and Loss describe graded selections. The hit rate uses those decided outcomes. Pending selections can make the displayed Total larger than Win plus Loss because they have been published but not settled.

ROI is based on settled selections and their recorded decimal odds using a flat one-unit calculation. Read it as the return for that displayed historical sample. It does not predict the ROI of the next day, league or market.

Keep any betting spend within a personal limit you can afford. The Gambling Commission’s guidance on spending limits explains how financial limits can help control gambling spend.

Open the Football xG predictions and results to read Win, ROI and Total with their date and sample context.