PREMIER BETTING TIPS • FOOTBALL GUIDES
Updated 4 October 2026
Quick answer: Judge a betting record by net profit, return on stakes, the prices actually taken and the size of the sample. Strike rate tells you how often bets win; it does not tell you whether they make money. Track refunds, partial settlements and losing runs too.
All figures below are invented teaching examples, not Premier Betting Tips performance claims. Decimal odds and ordinary cash stakes are used; fees are excluded unless stated.
What should a results record measure?
| Measure | Meaning | What to check |
|---|---|---|
| Strike rate | Winning bets ÷ resolved win/loss bets × 100 | The odds and how partial results are classified. |
| Total stakes | Money committed to the bets counted in the record. | The treatment of refunds and voids. |
| Net profit | Total returns, including returned stakes, minus total stakes and fees. | Every losing bet and applicable charge is included. |
| ROI on stakes / yield | Net profit ÷ total stakes × 100. | The denominator is stakes, not starting bankroll. |
| Average odds | The mean decimal price taken. | Whether it is weighted by stake; individual prices still determine profit. |
| Maximum drawdown | The largest fall from a previous bankroll peak. | The period, stakes and treatment of deposits or withdrawals. |
Terminology varies. In this guide, ROI means return on stakes, also called yield. A £100 profit on £1,000 of stakes is 10% ROI on stakes. If the starting bankroll was £500, that same profit is a 20% return on starting bankroll. Label the denominator so readers can compare like with like.
Why a higher strike rate can lose money
| Illustrative record | 55% strike rate | 70% strike rate |
|---|---|---|
| Bets and stakes | 100 bets at £10 each | 100 bets at £10 each |
| Odds on every bet | 2.00 | 1.30 |
| Wins / losses | 55 / 45 | 70 / 30 |
| Total returns | 55 × £20 = £1,100 | 70 × £13 = £910 |
| Net profit | £100 | −£90 |
| ROI on £1,000 stakes | 10% | −9% |
The second record wins more often but pays too little to cover its losses. For binary bets at one fixed price, the break-even win rate before fees is 1 ÷ decimal odds: 50% at 2.00 and approximately 76.92% at 1.30. This shortcut does not describe a mixed portfolio with partial settlements.
When odds or stakes vary, calculate each bet’s actual return and add the results. Multiplying an overall strike rate by a simple average price can misstate profit because the prices on winning bets matter.
Asian handicap: record half wins and refunds correctly
At a £10 stake and decimal odds of 1.90, the standard settlement amounts are:
| Settlement | Return including stake | Net profit |
|---|---|---|
| Full win | £19 | +£9 |
| Half win: one half wins, one half pushes | £14.50 | +£4.50 |
| Full push / refund | £10 | £0 |
| Half loss: one half loses, one half pushes | £5 | −£5 |
| Full loss | £0 | −£10 |
One bet of each type produces £48.50 in returns on £50 staked: a £1.50 loss and −3% ROI. Calling both full wins and half wins simply “winners” hides their different financial effects.
Reporting convention: For ROI examples here, count the original stake of every settled bet, including pushes and quarter-line bets. Exclude fully voided or cancelled bets from both stakes and returns. Keep them listed separately. Other records may exclude pushes from turnover; disclose the rule and use it consistently. For Asian handicap strike rates, show the five settlement counts rather than compressing them into an unexplained percentage.
Use the operator’s actual settlement and rules if a market is abandoned, voided or adjusted. Read our Asian handicap guide for line-by-line examples.
Fees, odds and staking change the result
Deduct applicable commission and other betting charges. Exchange commission rules can depend on net market winnings and the account’s terms, so use the charge actually applied rather than assuming a universal percentage. Smarkets’ commission explanation illustrates how a charge reduces winning returns.
Record the odds available when the selection was issued and the odds actually taken. A profitable record at 2.00 does not establish the same result at 1.75. Keep flat-stake performance separate from actual-stake performance: larger stakes on a few winners or losers can change the outcome substantially. Treat free bets and promotional stakes separately because their return rules differ from ordinary cash bets.
Losing runs and drawdown
A positive long-run expectation does not prevent consecutive losses. Under the simplified assumption that each bet independently has a 60% chance of winning, the chance that the next five all lose is 0.40⁵ = 1.024%. That is the chance of one specified five-bet sequence, not the chance of encountering a five-loss run somewhere in a season.
Real selections can share teams, matches and conditions, so independence may be unrealistic. A confidence label is not enough to establish a 60% win probability. Do not increase stakes to recover losses or treat a losing sequence as proof that a win is due.
Drawdown measures the fall from a previous peak. If a bankroll rises from £1,000 to £1,120 and then falls to £980, the drawdown from that peak is £140, or 12.5%. Deposits and withdrawals must be accounted for separately so they do not disguise performance.
How many bets are enough?
There is no universal number of bets that proves a strategy works. Ten wins can be luck; a larger record still needs scrutiny if it mixes changing rules, prices or markets. Statistical sample requirements depend on the effect being measured, the uncertainty tolerated and assumptions about the observations. The NIST guide to sample sizes for proportions explains those dependencies.
Report the number of bets and date range alongside ROI. Review complete, consecutive periods and test a fixed strategy on later selections that were not used to design it. Avoid choosing only the strongest league or month after seeing the results. Many bets from the same match do not provide the same independent evidence as unrelated observations.
Track FIS, confidence and Delta alongside results
FIS is a football assessment score; it is not automatically a win probability. Confidence must be interpreted using the report’s definition: it may describe evidence quality rather than the chance a bet wins. Delta (ΔFIS) is the gap between the home and away FIS ratings in reports that use that definition. It is not FIS minus confidence.
Save these values before kick-off, together with the report version and market. Then compare complete groups over time: for example, preselected FIS bands within one strategy, at comparable prices. Include unsuccessful selections too. A higher FIS band can win more often yet deliver worse ROI if the odds are shorter. Delta does not establish value without the market price and a relevant probability estimate.
Do not pool scores from different model versions as though their meanings are identical. Where confidence is explicitly a predicted probability, assess calibration separately: events assigned similar probabilities should occur at roughly that frequency over a sufficiently large comparable sample. Evidence-quality confidence should not be tested as if it were a probability.
A practical results log
- Record timestamp, fixture, strategy, market and exact line.
- Save issued odds, odds taken, stake and any applicable fees.
- Freeze FIS, confidence definition, ΔFIS and model/report version before the result.
- Record settlement type, cash return and net profit.
- Keep research-only and “no score / no bet” entries separate from placed bets.
- Review full periods by strategy, with counts, stakes, net profit, ROI and drawdown.
For accumulators, calculate returns and stakes by ticket. Do not count each leg as a separate winning bet while counting the ticket stake only once. Keep singles and accumulators in separate summaries, and flag overlapping selections.
Before accepting a performance claim: Can you see the full dated record, prices, stakes, settlement rules and fees? Is the sample complete? Were the rules fixed before results were known? A headline strike rate or screenshot of winners cannot answer those questions.
Frequently asked questions
Is a 70% strike rate good?
It depends on the prices, fees and settlements. At fixed odds of 1.30, 70% is below the break-even rate for ordinary win/loss bets before fees.
Does positive ROI prove a future advantage?
No. It describes the recorded sample under its stated accounting rules. Chance, selection bias, changed prices and changed conditions can affect future results.
Should refunds count as wins?
No. A push returns the stake and produces zero betting profit before any charges. Show it separately and explain how it is treated in turnover.
Related guides
18+. Betting involves risk. Set affordable limits and never chase losses. These examples explain measurement; they do not promise profit.
