Units in Betting: How to Read a Tipster’s Record
Nine settled picks, five of them hits, and the record still ends in the red. It looks like a spreadsheet error, but it is plain unit arithmetic: the profits from five hits did not make up the total losses from four missed picks. That is exactly why tipsters count results in units rather than in hit percentages. In Poland there is also the stake tax, which lowers the result of any record kept at the odds on the slip. Betting is entertainment, not a source of income, so the numbers below are there to help you read other people’s records and promise nothing.
The unit: a measure that does not depend on the wallet
A unit (in Polish “jednostka”) is a predetermined measure of stake. It can mean a fixed sum in zloty or a percentage of a stated bankroll. The author should say whether the base is the starting capital or the current balance, and when the unit may be recalculated. The basic terms are explained in our betting glossary.
With a unit equal to 2% of the starting capital, a player with PLN 500 stakes PLN 10, and a player with PLN 5,000 stakes PLN 100. Units make records easier to compare when the stake is defined in the same way, but on their own they do not prove the quality of the forecasts. In the examples below the value of the unit stays constant across the whole series. If it changes during the series, the cash flows in zloty have to be shown as well, because the sum of units no longer has a single conversion rate.
How big a unit should be for you is a separate question, and you will find the answer in the guide on bankroll management. Here we do not repeat percentages; we learn to read a record. The word “unit” means nothing until the author says how many zloty it is, or what percentage of the bankroll. A record without such a definition cannot be read. And a record in which units grow after a loss is no longer a confidence scale but one of the staking systems.
The 1–3 scale: why not 1–10
In the records we plan to publish, a three-step scale applies. One unit is the standard stake, three is the strongest pick in a given series, and two sits in between. This is our editorial choice, not an industry rule. The reasoning is purely arithmetical: the wider the range, the greater the influence of a single pick on the whole result.
Take ten picks at odds of 2.00. Nine stand at one unit each, and one was given the “strongest” weight. Each time exactly five of the ten hit (four or five of the other nine, so that the total comes to five); only the fate of that one pick changes:
| Weight of the strongest pick | Units staked | If it hits: profit / yield | If it misses: profit / yield |
|---|---|---|---|
| 1 u (flat record) | 10 | 0.0 u / 0.0% | 0.0 u / 0.0% |
| 3 u (1–3 scale) | 12 | +2.0 u / +16.7% | −2.0 u / −16.7% |
| 10 u (1–10 scale) | 19 | +9.0 u / +47.4% | −9.0 u / −47.4% |
The same hit rate, the same odds, and the result swings from −47.4% to +47.4%. On a 1–10 scale one pick steers the whole record. A ceiling of three units limits this swing in our example to ±16.7%. A scale without a ceiling also favours what we call unit inflation: the number of units grows during the series, so late results weigh more than early ones, and the record stops telling the truth about the whole.
How to read a record: an example with ten picks
The table below is an example made up for the arithmetic. It does not describe a real tipster or real results.
| No. | Units | Odds | Result | Profit (u) |
|---|---|---|---|---|
| 1 | 1 | 1.85 | hit | +0.85 |
| 2 | 2 | 2.10 | miss | −2.00 |
| 3 | 1 | 2.40 | miss | −1.00 |
| 4 | 3 | 1.70 | hit | +2.10 |
| 5 | 1 | 1.95 | hit | +0.95 |
| 6 | 2 | 1.80 | hit | +1.60 |
| 7 | 1 | 3.10 | miss | −1.00 |
| 8 | 2 | 2.00 | refund (void) | 0.00 |
| 9 | 3 | 1.90 | miss | −3.00 |
| 10 | 1 | 2.20 | hit | +1.20 |
| Total | 15 (excluding the refund) | – | 5 of 9 | −0.30 |
Units staked
We add up the units of all settled picks: 1 + 2 + 1 + 3 + 1 + 2 + 1 + 3 + 1 = 15. Pick no. 8 was refunded, so in our calculation it does not count towards turnover (with it the total would be 17). This is a convention, not a law of nature, which is why a good record states how it treats refunds.
Profit in units
A hit pays units × (odds − 1), a miss is minus the units staked, and a refund is zero. Pick no. 4: 3 × (1.70 − 1) = 3 × 0.70 = +2.10 u. Pick no. 9: −3.00 u. Five hits add up to +6.70 u, four misses to −7.00 u, a balance of −0.30 u.
Yield and ROI: the sources disagree
Yield is profit divided by turnover: −0.30 ÷ 15 = −0.02, that is −2.0%. This is how the Polish portal meczyki.pl defines it. The trouble starts with the word ROI. In our guide on tipping, ROI means profit relative to the sum of stakes. The site thebettordiary.com uses a different denominator: the starting capital. Before comparing two records, then, you have to check the formula given.
So there is no single, binding definition, and we will not pretend there is. In this text yield is profit ÷ units staked. When you see “ROI” in somebody else’s record, check the denominator. The same example calculated on the starting capital, say 100 units, gives −0.30 ÷ 100 = −0.3%, so the result looks milder only because the denominator is larger. Setting one tipster’s yield against another’s ROI on capital proves nothing.
Hit rate versus unit-weighted hit rate
Hit rate is the hit picks divided by the settled ones: 5 ÷ 9 = 55.6%. Unit-weighted hit rate divides the units staked on hit picks by all units staked: (1 + 3 + 1 + 2 + 1) ÷ 15 = 8 ÷ 15 = 53.3%. The difference arises because the two biggest misses (2 u and 3 u) landed on picks of a higher weight. Hit rate alone is easy to dress up, because it knows neither the odds nor the weights. Without odds you cannot work out profit either, so a record that gives only “hits out of settled” does not say whether the result was in the black.
The 12% tax: how it changes a record in Poland
The gambling tax on fixed-odds betting is 12% of the stake, not of the win, and the bookmaker settles it (more in the guide on tax on betting winnings). So 88% of the deposit goes into play. A slip pays 0.88 × stake × odds, and the order of multiplication does not matter, as the article the 12% tax and the effective odds shows. A record calculated on the slip odds alone is “paper”: it does not see this tax.
The conversion below applies to single bets with a constant unit value in zloty, a full stake subject to the 12% tax, and no promotion compensating for that tax. Refunds are excluded from turnover. At this stage we leave out the additional PIT, discussed separately below. If the yield before tax is Y, then after the stake tax it equals 0.88 × (1 + Y) − 1. Our example record: the payouts from hits come to 14.70 u, after tax 0.88 × 14.70 = 12.94 u, and with a turnover of 15 u that gives −2.06 u, that is −13.8% instead of −2.0%.
| Yield before tax | Yield after the 12% tax |
|---|---|
| −2.0% (our example) | −13.8% |
| 0.0% | −12.0% |
| +5.0% | −7.6% |
| +10.0% | −3.2% |
| +13.6% | 0.0% |
| +20.0% | +5.6% |
We read it like this: a record kept at the slip odds needs a yield of about +13.6% to break even after tax. A result of +10% looks impressive in a table and after tax is still in the red.
The same shows up in the hit-rate threshold, that is the percentage of hits above which picks stop producing a loss at constant odds. Without tax the threshold is 1 ÷ odds, with tax 1 ÷ (0.88 × odds).
| Odds | Threshold without tax | Threshold with the 12% tax | Difference (percentage points) |
|---|---|---|---|
| 1.80 | 55.6% | 63.1% | 7.6 |
| 2.00 | 50.0% | 56.8% | 6.8 |
| 2.50 | 40.0% | 45.5% | 5.5 |
The differences are calculated from unrounded values. A tipster who hits 55% of picks at odds of 2.00 is above the threshold without tax (50%) and below the threshold with tax (56.8%): the yield is 0.55 × 2.00 − 1 = +10%, and after tax −3.2%. At odds of 1.40 the thresholds are 71.4% and 81.2%, and the gap grows to 9.7 percentage points. The lower the odds, the wider the tax pushes the two thresholds apart.
Income tax is a separate matter: a 10% flat-rate PIT is paid on a single win above PLN 2,280, and it is calculated on the whole amount, not on the excess. The bookmaker collects it as the withholding agent. To include it in a unit record you need to know the payout in zloty. With a constant unit of PLN 80, the tax calculated in zloty can be divided by 80 and subtracted from the result in units. A record of odds and numbers of units without their monetary value does not allow you to check the PIT threshold.
How many picks make a sample
The ten picks in our example are an anecdote. Meczyki.pl advises calculating yield from at least several dozen to several hundred bets. For illustration, assume independent picks with an equal probability of hitting close to 50%. Under these simplifications the normal approximation of the 95% margin of error of the hit rate is ±1.96 × √(0.25 ÷ n). A constant stake alone does not make events independent.
| Number of picks | Margin of error of the hit rate (percentage points) |
|---|---|
| 30 | ±17.9 |
| 100 | ±9.8 |
| 300 | ±5.7 |
| 1,000 | ±3.1 |
This is a normal-distribution approximation for the variant with the greatest variance. The gap of 6.8 percentage points between the threshold with and without tax, at odds of 2.00, exceeds this margin after about 207 independent picks. It is a comparison of two quantities, not a test of a tipster’s skill or proof of a positive expected value. Correlated picks, for example on the same match, need a different estimate of uncertainty.
Our assessment: how to judge a tipster’s record
The list below is our editorial criteria. A quicker checklist to run before a slip for a specific day is in the guide betting tips for today (page in Polish); here the subject is counting.
- The time a pick is closed. A pick must have a date and time from before the first whistle, recorded so that it cannot be changed. An entry whose creation time cannot be checked proves nothing.
- No back-filling. Misses do not vanish from the record, and picks do not appear after the match. An account showing only wins most likely leaves out the rest.
- The odds from the moment of the pick. A record kept at the odds from the moment of the pick can be compared with the closing odds, and that is a signal of the quality of the analysis, independent of the outcome of a single match.
- Sample size. Dozens of picks are the minimum, hundreds are best. See the table above.
- Unit inflation. A constant, defined scale, preferably with a ceiling. A record in which units grow over time or reach ten deserves some distance.
- Profit only where there are odds. Without odds you can work out the hit rate, but not the yield. A pick “without odds” in the profit column is a hole in the record.
- Odds only from legal operators. A record kept at the odds of an operator outside the Ministry of Finance register describes a bet that cannot legally be placed in Poland. Such a result weighs nothing.
And one more thing: the bookmaker has an edge in the form of the margin, so even a clean, well-kept record does not promise that anyone will come out ahead. When counting results starts to serve recovering losses, it is no longer entertainment. Tools and contact details for help are described on the page about responsible gambling.
We apply the same counting rules in our own AI tipster league. Six artificial-intelligence models predict Ekstraklasa matches there for a virtual PLN 4,000 a month, on a constant scale of 1 to 3 units of PLN 80, with a 12% tax on every stake. Every pick has a submission time from before the match and cannot be changed.
Frequently asked questions
How many zloty is one unit?
It depends on the adopted definition. One unit can mean a fixed amount, for example PLN 80, or a set percentage of the starting or the current capital. It changes only when the open rules of the record provide for it. Without stating the base and the moment of recalculation, results are hard to compare.
How does yield differ from ROI?
Yield is profit divided by turnover, that is the sum of the units staked. ROI is defined in two ways: some sources calculate it the same way as yield, others on the starting capital. So always check what the author divides by.
What yield is enough after the Polish tax?
In the single-bet variant described above, staked with cash, with no promotion compensating for the tax and before the additional PIT, a record needs about +13.6% yield before tax to break even once 12% is deducted from the stake. This is arithmetic, not a promise: over a short series the result is decided mainly by volatility.
Does a refund (void) count towards the hit rate?
Not in our calculation. A refunded pick has a profit of 0 and counts neither towards turnover nor towards the hit rate. Others may count differently, so a reliable record states the rule it uses.
Why a 1–3 scale and not 1–10?
Because a wide scale lets a single pick dominate the whole record. In the example above, ten units on one pick move the yield by ±47.4%, and three units by ±16.7%.
Does a unit record include the 12% tax?
That depends on the author. A record kept on the slip odds alone leaves out the stake tax. In the single-bet variant described above, with no promotion compensating for the tax and before the additional PIT, we use the formula 0.88 × (1 + yield) − 1. Under other rules the actual cash flows have to be taken into account.
Sources: our guides betting glossary, bankroll management, tax on betting winnings (12% rate, PLN 2,280 threshold and legal basis: Act on Gambling Games, Journal of Laws 2025, item 595 (PDF in Polish); Personal Income Tax Act, Journal of Laws 2026, item 592 (PDF in Polish)), the 12% tax and the effective odds, how to tip and betting tips for today (page in Polish); meczyki.pl, “Yield – co to jest i jak go liczyć?” (in Polish); thebettordiary.com, “Profit, ROI, and Yield in Betting”; the register of legal entities of the Ministry of Finance at gov.pl/web/finanse/legalny-hazard. The sample records and tables are the editors’ own calculations, made up for the arithmetic.