Why measurement is the job
Lesson 1 described an electronic gaming estate as machines that each earn rent for their space. Running one well starts with knowing what each machine earns, what it should earn, and how that compares with the space it takes. Every decision in the rest of this course, from which games to buy to which to retire, is only as good as those numbers.
The industry uses a small vocabulary for them: coin-in, win, hold, win per unit per day, occupancy and index. Each is simple. Most mistakes come from mixing them up, or from comparing a figure from one estate with a figure from another that is measured differently.
The meters and what win means
Every machine keeps electronic meters, and the main ones are set by regulators' technical standards. Illinois's Technical Standards for Video Gaming Terminals (version 1.5, revised June 2024) are typical. They require, among others:
- Coin in: the total value of all wagers, whether from inserted cash, the credit meter or any other means.
- Coin out: the total value paid by the machine for winning wagers.
- Bill in, ticket in and ticket out: the cash and vouchers going into and out of the machine.
- Games played: the count of games since power reset, door close and initialisation.
Two families of meters matter. The play meters (coin-in and coin-out) record what was wagered and won. The cash meters (bills and tickets in, tickets and hand pays out) record money crossing the machine's edge; the cash and tickets collected from a machine are its drop. Win can be calculated from either:
Win = coin-in minus coin-out (and jackpots paid by hand) = cash in minus cash out
The Illinois Gaming Board's public data shows the two routes agreeing. For September 2026 the Board reports $2.997 billion played, $2.719 billion won, $1.162 billion of funds in and $884 million of funds out. Played minus won gives $278.42 million; funds in minus funds out gives $278.45 million, the reported net terminal income. The two differ by about $27,000. Lesson 4 covers how such differences are reconciled.
The same figures show why coin-in is much larger than drop. Players recycle winnings: in September 2026, every dollar put into Illinois terminals was wagered about 2.6 times ($2.997 billion played on $1.162 billion of funds in). Win as a share of coin-in was 9.3%; win as a share of funds in was 24%. Both are "hold". Only the first is comparable with a game's theoretical return, so always ask which base a hold figure uses.
Theoretical and actual hold
Hold is win divided by coin-in over a real period. House edge is fixed by the maths of the game; hold is what actually happened. The Casino Game Mathematics course sets out why they differ and why, over enough play, they converge.
On the floor the theoretical side is turned into money. Each game has a theoretical hold (often called par), and the system multiplies it by coin-in:
Theoretical win = coin-in x theoretical hold
Illustration: a machine with a par of 8.5% takes $5,000 of coin-in in a day. Its theoretical win is $425. If the meters show coin-out of $4,600, actual win is $400 and actual hold is 8.0%.
The gap between $425 and $400 is not a problem. Over one day on one machine actual hold can be anything, including negative. The volatility lesson shows that the spread around expectation shrinks only with the square root of the amount played, so a high-volatility title needs far more coin-in than a low-volatility one before its actual hold means much.
This is why individual games are best judged on theoretical win and the floor, in aggregate, on actual win. Theoretical win removes the luck from a game-level comparison. Actual win, summed across hundreds of machines and a full month, is close enough to theory to be the number that pays the bills. A persistent gap between the two at floor level is worth investigating: wrong par settings in the system, a meter fault, or something worse.
Win per unit per day
The core performance measure is win per unit per day (WPUD): win divided by the number of machines and the number of days. It answers the rent question directly: what does one machine position earn in a day?
Nevada's monthly report gives the inputs. For the twelve months from August 2025 to July 2026, nonrestricted licensees statewide reported an average of 125,721 slot units and $10.90 billion of slot win, at a win percentage of 7.10%. Dividing:
- WPUD = $10.90 billion / 125,721 units / 365 days = about $238
- Coin-in = win / win percentage = $10.90 billion / 0.0710 = about $154 billion
Both are derived here from the report's figures. The coin-in step works because Nevada defines slot win percentage as the reported win divided by the total amount played.
Use theoretical win per unit per day for comparing games, and actual WPUD for comparing periods, sections of the floor and properties. Write down which one a figure is, and whether the units are average units in service or units on the floor at month end, because the denominator changes the answer.
Occupancy, utilisation and indexing
WPUD tells you what a machine earned; it does not tell you why. Breaking it into parts does:
WPUD = occupied hours per day x games per hour x average bet x hold
Occupancy is the share of available machine time in which someone is playing. Systems estimate it from games played or from carded sessions. Illustration: a 400-unit floor offers 9,600 machine-hours a day. If 3,360 of them are occupied, occupancy is 35%, or 8.4 hours a machine. With 500 games an hour at a $0.75 average bet and 8% hold:
- 8.4 x 500 x $0.75 x 0.08 = $252 WPUD
The decomposition separates problems that look identical in a WPUD report. Low occupancy with healthy spend per occupied hour points to placement, signage or too many machines. High occupancy with low spend points to the wrong denomination or game speed for the audience. Occupancy also has a time profile: a floor at 35% across a day may be at 85% on a Saturday evening, and the peak, not the average, decides whether players are turned away.
Utilisation is used loosely: some mean occupancy, others coin-in per unit against the floor average. Define it before comparing.
Indexing compares a game with the floor. Illustration: the floor averages $250 theoretical WPUD; a bank of ten units of one title averages $350. Its index is 350 / 250 = 1.40, so each of its positions earns 40% more than the average position. A game indexing below 1.0 for months, on theoretical win and after allowing for its location on the floor, is a candidate for conversion or removal, which lesson 3 takes up.
Reading a public slot report
Public reports turn these ideas into checkable numbers. The table below uses Nevada's twelve months to July 2026; WPUD and coin-in per unit per day are derived as above.
| Area | Avg units | Win | Win % | WPUD | Coin-in per unit per day |
|---|---|---|---|---|---|
| Statewide | 125,721 | $10.90bn | 7.10 | $238 | $3,346 |
| Las Vegas Strip | 35,125 | $5.03bn | 7.72 | $393 | $5,084 |
| Downtown Las Vegas | 10,025 | $635m | 8.23 | $174 | $2,109 |
| Reno | 8,846 | $670m | 5.55 | $208 | $3,740 |
Source: Nevada Gaming Control Board, Monthly Revenue Report, July 2026, pages 1, 8, 12 and 33.
Downtown holds the highest percentage and earns the least per machine. Reno holds the lowest and earns more than Downtown, because each machine there takes far more play. Hold percentage is a price; WPUD is revenue. A property that raises par to lift hold can lose more in play than it gains in price.
Illinois offers the route equivalent. With 50,141 terminals and $278.4 million of net terminal income in September 2026, the statewide figure is about $185 per terminal per day before tax, from about $1,990 of play. That is about four-fifths of Nevada's statewide WPUD, from terminals limited to a $4 bet, and it is shared with the state, the municipality and the venue before the operator sees it.
Other regulators publish their own versions. New South Wales ranks clubs and hotels by net profit per gaming machine each quarter. Oregon's lottery sets a floor on play rather than win: a retailer must generate at least $10,000 of "average dollars played video sales per week per terminal", averaged over 90 days or more, or risk losing terminals.
Using the numbers
What to check before relying on any slot figure:
- The base of the percentage. Hold on coin-in and hold on drop differ widely (9.3% against 24% in Illinois in September 2026). Only hold on coin-in compares with par.
- The denominator. Average units in service, units at month end, or units licensed. State it.
- Theoretical or actual. Judge games on theoretical win; judge floors and months on actual.
- The model. A route terminal's win is split between tax, venue and operator; a casino keeps its win after tax. Lesson 1's habit applies.
Common mistakes:
- Removing a game after a month of low actual hold, when its theoretical win was fine.
- Reading a higher win percentage as better performance, when WPUD fell.
- Comparing a whole-floor average with a high-traffic bank without adjusting for location.
- Ignoring occupancy, so that adding machines to a floor already at 35% looks like growth when it only spreads the same play thinner.
Lesson 3 uses these measures to manage the floor: the mix of games, where they stand, and the deals under which they are bought, leased or shared.