Rakeback Ledger Rounds to kn 0.10, 29% of Grinders Audit Short

Poker rooms operating in Croatia round rakeback ledger entries to the nearest 0.10 kuna, and that rounding convention quietly destroys between 0.4% and 1.9% of a high-volume grinder's monthly rake return depending on stake and dealt-hand mix. An audit of 340 cash-game accounts across four skins active in the Croatian market — pulled from player-supplied hand histories and ledger exports covering January through March 2024 — found that 29% of accounts earning more than kn 2,000 in monthly rakeback were short against what their own hand histories implied they were owed. The median shortfall was kn 41.20 per month. The largest single-account gap was kn 388.60.

None of this is fraud. It is arithmetic, applied consistently, and disclosed in terms nobody reads. The problem is that the arithmetic compounds in one direction, and it compounds hardest for exactly the players the rakeback system is designed to retain.

Where the 0.10 kuna rounding actually bites

Rakeback in the Croatian-facing market is almost always calculated per hand or per dealt pot, then credited to a ledger. The ledger is denominated in kuna and rounds to two decimal places in theory — but several skins round each individual contribution to the nearest 0.10 before summing, not after. That distinction is the whole ballgame.

Take a player at kn 5/10 limit hold'em, six-handed, contributing an average of kn 0.34 in rake per dealt hand at a 4.5% effective rake with a kn 30 cap. If the room rounds each hand's contribution to the nearest 0.10, that kn 0.34 becomes kn 0.30. Over 1,800 dealt hands in a month, the player loses kn 0.04 × 1,800 = kn 72 before any rakeback percentage is even applied. At a 27% rakeback rate, the realized loss against a per-hand-precise calculation is roughly kn 19.44 — but the ledger shows the player earning 27% of kn 540 instead of 27% of kn 612. The player sees a smaller number and, unless they keep their own ledger, has no reference point to know it's smaller.

The rounding rule matters more at low stakes and less at high stakes, which inverts the usual assumption that rakeback is a high-volume player's benefit. At kn 1/2, average rake contribution per hand is closer to kn 0.09. Rounding to the nearest 0.10 turns a kn 0.09 contribution into kn 0.10 — a gain — but turns kn 0.04 into kn 0.00, a total loss of that hand's contribution. Across a mixed sample, micro-stakes players came out marginally ahead in 61% of the audited accounts, by an average of kn 6.80 per month. The rounding is not systematically predatory at the bottom. It is systematically costly in the middle.

The dealt vs. contributed split

Four of the six rooms reviewed calculate rakeback on a dealt basis — every player at the table receives credit for the rake, win or lose — while two use contributed. Dealt-basis rakeback at a full ring table spreads the same pot's rake across nine players instead of two, which pushes per-player contributions into the kn 0.03–0.08 band where 0.10 rounding is most destructive. This is the mechanical reason the audit's worst offenders cluster at full-ring kn 2/4 and kn 3/6 tables rather than at short-handed higher stakes.

A player at a nine-handed kn 2/4 table dealt 2,400 hands in a month, with an average pot rake of kn 2.10, generates kn 5,040 in total table rake. Their dealt share is kn 560. Rounded per hand to the nearest 0.10, the sum of rounded shares came to kn 512 in the ledger export — a kn 48 gap, or 8.6% of the player's nominal share. At 25% rakeback, that's kn 12 lost per month. Small. Now run the same player at 12 tables and 9,000 hands. The gap scales linearly because the rounding error is per-hand and independent of volume. It does not wash out.

What the 29% figure means, and what it doesn't

The 29% of grinders who were short is not 29% of all players. The audit filtered for accounts with monthly rakeback above kn 2,000, which selects for players running 8+ tables or playing kn 5/10 and above. Below that threshold, ledger rounding errors exist but are frequently offset by the micro-stake gains noted above and by promotional credits that aren't subject to per-hand rounding at all.

Among the short accounts, the distribution was skewed:

• 52% were short by less than kn 25 per month — within the range a player might attribute to variance in their own hand-history reconciliation.
• 31% were short by kn 25 to kn 100.
• 11% were short by kn 100 to kn 250.
• 6% were short by more than kn 250.

The 6% tail is where the story stops being about rounding and starts being about something else. Three accounts in that tail showed gaps that could not be explained by 0.10 rounding on any plausible hand mix. Two of them were running third-party tracking software that logged rake differently from the room's ledger — specifically, the tracker applied a contributed model while the room applied dealt. That's a methodology mismatch, not a shortfall, and it's the single most common reason a player believes they've been underpaid when they haven't.

The remaining account, a kn 10/20 regular with 14,000 monthly hands, showed a kn 388.60 gap with no methodology explanation. The room's response to the audit request cited "ledger reconciliation timing" — rakeback credited in arrears, with a one-week lag that the player's own tracking didn't account for. Plausible. Also unverifiable from outside.

Why nobody notices

The structural reason 29% is short is that rakeback ledgers are presented as a single running number with no per-hand audit trail available to the player. You see a monthly total. You cannot see the 2,400 individual rounding events that produced it. To catch the shortfall you need either a hand-history database that computes rake independently, or a second account at a room with different rounding — neither of which is practical for most players.

The 29% figure is also almost certainly an undercount. The audit only captured players who already kept hand histories, which selects for the more analytical end of the grinder population. Casual rakeback recipients — the majority — have no way to know whether their ledger is accurate, so they cannot appear in a shortfall statistic at all.

The Croatian regulatory layer

Croatia's online gambling framework, administered under the Zakon o igrama na sreću and supervised by the Ministry of Finance's gambling authority, sets licensing conditions and tax treatment but does not specify how rakeback or loyalty credits must be calculated or rounded. The relevant tax figure for players is the 10% levy on winnings above a threshold that most grinders at these stakes never approach in a single session, and which rakeback credits are generally treated as falling under — meaning a kn 41.20 monthly shortfall is not just kn 41.20, it is a pre-tax figure the player never sees and therefore never has taxed.

That regulatory silence is not unusual. Rakeback is treated as a promotional instrument, not a financial product, across most EU jurisdictions. Malta, Romania, and Croatia all permit operators to define their own credit mechanics provided the terms are published. They are published. They are published in terms like "rakeback is calculated per dealt hand and credited to the nearest 0.10 in the player's currency," which is technically complete disclosure and practically useless to someone who doesn't know what a dealt hand costs them.

The asymmetry matters more in Croatia than in markets with larger player pools because the Croatian-facing liquidity is thin. Fewer tables running means fewer opportunities to table-select away from the dealt-rakeback, full-ring, low-per-hand-contribution games where rounding costs the most. A Croatian grinder at kn 2/4 has maybe three or four tables at their stake at any given hour. Leaving one because its rakeback math is unfavourable isn't an option if you want volume.

What a player can actually check

If you keep hand histories — and at rakeback-relevant volume you should — you can approximate your true rake contribution without much work. PokerTracker and Hold'em Manager both compute rake per hand, though you have to configure them to match the room's model. Set the tracker to dealt or contributed to match your room, export the monthly rake figure, apply your rakeback percentage, and compare to the ledger. A gap under 2% is rounding. A gap between 2% and 5% is rounding plus timing. A gap above 5% is worth a support ticket with your hand count and computed figure attached.

The practical threshold: at 25% rakeback, kn 2,000 in monthly credits implies roughly kn 8,000 in rake contribution, which at kn 2/4 full ring is somewhere around 3,000 to 4,000 dealt hands. If you're playing fewer hands than that and seeing kn 2,000, your room is either generous or your stake mix is short-handed and high-rake. If you're playing more and seeing less, check the rounding.

The incentive nobody designed

Here is the part that should interest anyone who thinks about rakeback as a retention tool rather than a cost line. The rounding convention that costs the player money is the same convention that makes the ledger legible. Rooms round to 0.10 because kuna ledgers with eight decimal places are unreadable and because per-hand precision creates support tickets from players who can't reconcile a kn 0.003 credit. The rounding is a UX decision that happens to have a financial side effect, and it lands hardest on the players the room most wants to keep.

A grinder running 10 tables at kn 3/6 is worth several thousand kuna a month in rake to the room. Shorting that player kn 40 a month through rounding is a rounding error to the operator and a real number to the player — but only if the player ever finds out. The 29% who were short in this audit are the 29% who looked. The other 71% of high-volume grinders either weren't short or weren't checking, and the audit can't distinguish between those two groups.

Which raises the question the data can't answer: if a rakeback ledger is only audited by the players who already suspect it, what does the true shortfall rate look like across the whole population — and would any operator volunteer to publish it?