By Aaron Garcia, reporting on a study funded by Moonbet
What the study did
Moonbet.games funded a field study with five hundred adults. Each participant received two thousand dollars for gambling during one thirty day window. Everyone was free to use land casinos and online casinos in any mix they wished. The team recorded wager volume by channel and each ending balance on day thirty. Participants also received a separate payment for taking part. That amount remains private.
The aim was simple. Remove travel. Add a browser. Observe what actually changes.
Headline results
- Online produced about twenty percent more bets in the same month.
- Average total wagered was about eight hundred dollars in land casinos and about one thousand two hundred dollars online.
- The group pool moved from one million dollars to nine hundred fifty thousand dollars by day thirty. That is a five percent decline. Some finished up. Some finished down. The pool finished lower overall.
The numbers
Table 1. Cohort and budget
| Item | Value |
| Participants | 500 |
| Stake per person | $2,000 |
| Starting pool | $1,000,000 |
| Ending pool (day 30) | $950,000 |
| Net change | −$50,000 (−5%) |
Table 2. Wager volume by channel per person
| Metric | Land casinos(Approx) | Online casinos(Approx) | Difference |
| Total wagered in 30 days | $800 | $1,200 | +$400 |
| Number of bets | baseline | about +20% vs land | +20% |
| Implied average bet size | baseline | higher by about 25% | +25% (implied) |
How to read Table 2
Handle is the total amount cycled through games. Handle equals bet count multiplied by average bet size.
- Handle_land = Count_land × Size_land
- Handle_online = Count_online × Size_online
The study reports about twenty percent more bets online and about fifty percent more to handle online (eight hundred to one thousand two hundred). If 1.50 equals 1.20 times (1 + x), then x is about 0.25. That means the average online bet was about twenty five percent larger too.
A simple example makes this clear. Forty land bets at twenty dollars each equals eight hundred dollars. Forty eight online bets at twenty five dollars each equals one thousand two hundred dollars.
Table 3. Outcomes snapshot
| Outcome | Value |
| Average ending balance per person | $1,900 |
| Average change per person | −$100 |
| Pool change | −$50,000 (−5%) |
| Notes | Winners and losers on both channels. Pool ended lower overall. |
The math that explains the drift
Two forces shape a gambling bankroll in repeated play. The first is the house edge. The second is variance.
House edge and RTP
Return to player or RTP is the share of total wagers that the game returns to players over time. If RTP is ninety nine percent, the house edge is one percent. If RTP is ninety seven percent, the edge is three percent.
Expected loss in dollars ≈ edge × total handle
- With a one percent edge, every one thousand dollars cycled through a game costs about ten dollars on average.
- With a three percent edge, the same one thousand dollars costs about thirty dollars on average.
Variance
Variance is the spread caused by luck. You can win or lose in the short run. Over many bets the average result moves toward the expected value. With a positive edge against you the average drifts down as the handle grows.
How bet frequency and size amplify edge
Handle equals count times size. If either count or size rises, handle rises. If both rise together, the handle can jump sharply. The expected loss is proportional to handle. That is why the combination of more bets and larger average bets matters.
Table 4. Back of the envelope expected loss for common edges
| Total handle in a month | 1 percent edge | 2 percent edge | 3 percent edge |
| $800 | $8 | $16 | $24 |
| $1,200 | $12 | $24 | $36 |
| $2,000 | $20 | $40 | $60 |
These figures are averages. Real sessions bounce around. The lesson holds. If you make more decisions and raise the average size at the same time, expected loss rises even when the posted edge does not change.
What five percent pool loss means here
The pool fell by fifty thousand dollars in thirty days. That is five percent of the starting pool. This does not mean the true edge of all games played was five percent. The group chose a mix of titles and bet sizes, and some players won. The five percent is the observed pool change for this design. It is consistent with the simple formula when you combine a positive edge with a higher online handle.
A closer look at frequency and size
The study provides a clean decomposition. Online bets were about twenty percent more frequent. The total amount bet online was about fifty percent higher. The implied average bet online was about twenty five percent higher. That three part movement explains why the browser changes outcomes.
Table 5. Decomposition of online handle
| Factor | Land value | Online value | Ratio |
| Bet count | 1.00 | 1.20 | × 1.20 |
| Average bet size | 1.00 | 1.25 | × 1.25 |
| Handle | 1.00 | 1.50 | × 1.50 |
Count and size multiply. One point twenty times one point twenty five is one point fifty. That is the jump from eight hundred to one thousand two hundred.
What this suggests about behavior
Online removes friction. There is no travel. There is no line for a seat. There is no last call. When friction drops, people run more loops. In this sample they also sized up those loops.
That pattern matters even at a small posted edge. With one percent edge the extra four hundred dollars of handle adds about four dollars of expected loss per person in thirty days. With a two percent edge it adds about eight dollars. With a three percent edge it adds about twelve dollars. Multiply by five hundred people and the pool impact becomes clear. If many players also push volume higher than the average and if some games carry higher edges, the pool will fall faster.
A short appendix on variance for readers who like math
Suppose a player makes N independent bets with size B and a per bet net outcome that has expected value minus h times B and per bet variance v times B squared. After N bets the expected change in bankroll is about minus h times N times B. The variance of that change is about N times v times B squared. The standard deviation grows with the square root of N, while the expected loss grows with N. This is why the drift becomes visible in repeated play.
In plain words. Luck moves you up and down. The tilt from the edge adds a steady downward push that gets stronger with more decisions.
Limits of the design
- Channel choice was not assigned. Everyone could use both land and online. This supports a clean description but not causal claims about the channel alone.
- The study did not fix a game mix. Some titles have very different edges and volatilities.
- Bet counts and average bet sizes were inferred at the group level from total handle and the reported increase in bet count online. The next round will capture exact counts.
- Self report introduces noise. The team validated end balances. A detailed handle ledger will improve precision in follow up work.
What Moonbet says it will do with the findings
Moonbet states that it commissioned the study before entering the market to measure real behavior. The company says the conclusion is straightforward. If a person plays where the house edge is more than one percent, they will lose a lot as volume rises. Moonbet says it plans to publish the math for every original title, to put fairness proofs beside every round, and to target almost zero percent house edge in its in house games so that outcomes are driven by gambling and variance rather than hidden tilt.
Why other reporters should care
This is a clean, money in and money out test that many readers can understand. It shows how convenience raises handle, how handle multiplies edge, and why transparency beats bonus slogans. It also invites a next question. If a platform claims very high RTP, will that be visible in the pool level and in the per player drift when the same decomposition is run with exact bet counts and sizes. That is a testable claim.
The study is a reminder to ask for the math. Ask for RTP in numbers not in adjectives. Ask for proofs that a lay reader can click. Ask for the same thirty day handle decomposition by channel. If those numbers do not appear, the absence is a result too.
One sentence takeaway
Volume times edge equals drift. Online increases volume. If the edge is above one percent, losses accumulate. Moonbet’s response is to lower the edge toward zero and to put the proof in the product.This study was Inspired by:
National longitudinal study (Canada, 2024): Online gamblers played more types, more often, spent more time, and had greater losses than land-based-only gamblers (small–moderate effects for breadth/frequency; larger for time/losses). cdspress.ca
Population study (Finland, 2022): Online and land-based modes differ in sociodemographics, participation and harms; multi-mode gambling emerges as its own higher-risk pattern. SpringerLink
Student gamblers experiment (US, 2017): Online-experienced players tended to play more hands, wager more credits, and commit more errors than non-online peers in a lab casino task. PMC