Hold percentage vs house edge: why they are different numbers
House edge is the expected loss per shilling staked, fixed by the game's rules. Hold is the share of money brought to a table, or staked in total, that the casino actually keeps over a period. Hold depends on how much players re-stake and on luck, so it can sit far from the house edge.
Two numbers, two questions
House edge answers: on average, how much does this game take from each shilling staked? Hold answers: of the money that came to the table over a period, how much did the casino end up keeping? The first is a property of the rules. The second is a measurement of what happened.
The two are often used as if they were one number, which leads to wrong conclusions about cost. The guide to RTP and house edge covers the first number; this page shows where the second one comes from and why it can look so different.
House edge: fixed by the rules
House edge is the expected loss per unit staked over the long run. It comes from the probabilities and payouts, so it does not change with the player, the table or the night. Single-zero roulette is the cleanest case: a straight-up bet on one number pays 35 to 1 and wins in 1 of 37 pockets.
The expected result of a KES 100 straight-up bet is (1/37 × 3,500) + (36/37 × −100) = 94.59 − 97.30 = −2.70 shillings, or 2.70% of the stake (exactly 1/37 = 2.7027%). The derivation is set out in the roulette house edge guide.
Hold: a ratio of money that actually moved
Hold is the casino's win divided by a base, and the base is a choice. At a table, the usual bases are the drop (the money players bought chips with) or the turnover (the total of all bets placed). Win is simply what the casino took in minus what it paid out in the period.
The Gambling Control Act, 2025 defines gross gambling revenue as "total revenue received from a gambling activity before any deductions are made". Regulation 92(1) of the Conduct of Gambling Operations Regulations, 2026 requires a licensee to keep records of total sales, total payout and monthly gross gaming revenue. Those are the raw amounts a hold figure is built from; the regulations as read here do not set a formula or a target for hold.
hold on drop = (buy-ins − cash-outs) ÷ buy-ins
hold on turnover = (buy-ins − cash-outs) ÷ total bets placed
Why hold on drop is not the house edge
Chips are bet again and again. A player who buys in for KES 1,000 and places KES 100 bets can stake the same chips many times before leaving, so total bets can be several times the buy-in. The link between the two numbers is the turnover ratio.
expected hold on drop = house edge × (turnover ÷ drop)
A single-zero roulette table (edge 1/37 = 2.7027%) has three nights. On night A, players buy in for KES 100,000 in total, place KES 600,000 of bets in all, and cash out KES 82,000. The casino's win is 100,000 − 82,000 = KES 18,000, so hold on drop is 18% and hold on turnover is 18,000 ÷ 600,000 = 3.0%.
The expected win from edge alone is 600,000 ÷ 37 = KES 16,216, so night A ran slightly ahead of the average. Nights B and C use the same game and the same edge, with different play and different luck.
| Night | Drop (KES) | Turnover (KES) | Turnover ÷ drop | Expected win at 2.7027% (KES) | Cash-outs (KES) | Actual win (KES) | Hold on drop | Hold on turnover |
|---|---|---|---|---|---|---|---|---|
| A | 100,000 | 600,000 | 6 | 16,216 | 82,000 | 18,000 | 18.0% | 3.00% |
| B | 100,000 | 1,500,000 | 15 | 40,541 | 60,000 | 40,000 | 40.0% | 2.67% |
| C | 100,000 | 600,000 | 6 | 16,216 | 115,000 | −15,000 | −15.0% | −2.50% |
The edge is the same on all three nights, yet hold on drop runs from −15% to 40%. Night B is high because chips were staked 15 times each, not because the game was changed. Night C is negative because a few large results went the players' way, which is ordinary variation.
What makes hold wander
Two forces move hold away from the edge, and they work on different parts of the formula.
- Turnover ratio. Faster play, smaller bets relative to the buy-in, and longer stays all raise turnover per shilling of drop. Hold on drop rises with it, even though the edge per bet is unchanged.
- Luck. Over a finite number of bets, results scatter around the expected win. The scatter is larger when the game has high volatility or when the sample is small, such as one table for one night.
As turnover grows, the win measured against turnover tends to settle near the edge. Night A's 3.00% and night B's 2.67% are both within sight of 2.70%, while night C's −2.50% is a reminder that a single night is a small sample.
Where the cost to a player sits
For a player, the figure that tracks cost is the edge applied to total bets, not to the buy-in. This is the same arithmetic as expected value: the average loss is edge × turnover.
A player buys in for KES 2,000 and, over an evening, places bets that add up to KES 30,000 on single-zero roulette. The expected loss is 30,000 ÷ 37 = KES 811, about 40.5% of the buy-in, even though the edge is 2.70%.
Hold figures reported for a whole table would hide this, because they blend many players, some ahead and some behind.
Online games: the same ideas with different records
An online game has no chip tray, so drop is not measured the same way. Records such as total sales, total payout and gross gaming revenue take its place, and a ratio of revenue to total stakes is a hold-style measure. Whether a figure is stated against deposits or against stakes changes its size greatly, for the same turnover reason as above.
The figure the rules point players to is different. Regulation 20(2) says a licensee shall display clearly "rules of games, odds, house edge and average return to the player", and regulation 20(3) extends this to online gambling. Regulation 45 adds that the theoretical return to player percentage for each game is to be disclosed, and that the game's random number generator is certified to test it. That is the edge-type number, because it is calculated from the rules rather than observed from past play.
How the two relate to a short session
Because hold depends on luck and re-staking, a high or low hold for one period says little about the game's edge. The same reasoning explains why actual results in short sessions stray from the theoretical RTP. Edge is a prediction of the average, hold is a record, and the record only approaches the prediction when the amounts wagered are large.
Gambling is addictive. Play responsibly, set a limit on total bets rather than on one balance, and treat any loss as a cost, not as a way to earn.
Questions and answers
Can hold be higher than the house edge?
Yes. If hold is measured against the money players bought in with (the drop), and players re-stake those chips many times, the casino keeps a larger share of the drop than the edge alone. In the table above, a 2.70% edge produces a 40% hold on drop when the chips are staked 15 times.
Can hold be negative?
Yes. If players cash out more than they bought in for, the casino's win for that period is negative, and so is the hold. The house edge stays positive because it describes the long-run average, not one period.
Which number should a player look at?
House edge, or the equivalent RTP, is the one that describes a game before you play. Hold is a record of what happened across many players in a past period, so it says nothing about the next session.
Do Kenyan rules mention house edge?
Yes. Regulation 20(2) of the Conduct of Gambling Operations Regulations, 2026 says a licensee shall display clearly the rules of games, odds, house edge and average return to the player, and regulation 20(3) applies this to online gambling with the necessary modification.
Is hold the same as gross gambling revenue?
No, but they are linked. The Gambling Control Act, 2025 defines gross gambling revenue as total revenue received from a gambling activity before any deductions. Hold is a ratio, and a revenue or win amount is the top line of that ratio.