Corrections
Last updated 3 August 2026
When a published figure turns out to be wrong, we change it and say so here, with the old value, the new one and the reason. Corrections are not made silently.
Policy
A correction is issued whenever a figure, chart or stated finding on a published study is found to be wrong. The study is amended, a note is added, and an entry appears below recording what changed and why. Corrections that alter a study’s conclusion are marked as such. Typographical fixes that do not change a number or a claim are not logged.
If you believe something here is wrong, please tell us. A specific challenge to a specific number is the most useful thing anyone can send us, and we will check it against the source data rather than defend it.
Corrections to published studies
None to date.
This section will list them as they occur, and it is not expected to stay empty indefinitely.
Revised during verification
These are not corrections. Nothing below was ever published in its original form. They are recorded because every published figure is re-queried before it appears, and this is what that process caught. Each is a real error that would have reached print without it.
| Study | Draft figure | Published figure | Why it changed |
|---|---|---|---|
| How players choose an online casino | 34.7% of players compare more than one casino | 27.8% | The original figure counted repeat clicks on the same brand as comparisons. Deduplicating by brand cut it by seven points. The published study states the correction rather than quietly using the lower number. |
| The game supply divide | Dissimilarity index of 25.385 | 0.435 | A dissimilarity index must fall between 0 and 1; this one did not, which signalled it was wrong. It had been normalised by the number of brands rather than the number of supply relationships, and brands carry many suppliers each. |
| What re-regulation does to player demand | A Dutch effect roughly three times larger | +2.1pp | Joining the licence table directly onto a click aggregate duplicated rows for brands holding several licences, inflating the licensed share. Pre-aggregating licence flags per brand fixed it. |
| Where new casino brands are licensed | Offshore trio holds 37% of brands and 7.6% of deposit value | 32.1% and 6.6% | Summing across licensing blocs double-counts brands holding licences in more than one jurisdiction. The figures are now deduplicated at brand level. |
| Cryptocurrency acceptance and the licensing boundary | 68% of the change attributable to market composition | 47% composition, 51% offshore adoption, 2% EEA adoption | The original decomposition did not reproduce when the cohort buckets were rebuilt from source. The revised split is the one the query returns. |
Investigated and not published
Findings that looked promising and were dropped because the data could not support them:
- A comparison of deposit-limit tooling across licensing regimes. The underlying data has no responsible-gambling or deposit-limit field; the claim could not be checked and was cut.
- A payment-method comparison involving POLi, which returned zero brands in every market examined. The comparison was removed rather than reported as a finding of absence.
- A study of which individual slot titles appear most widely. The field that appeared to hold game titles holds 27 category values, not titles.
- A scroll-depth figure in the consumer-behaviour study was dropped because the underlying series could not be reproduced to the precision the chart implied.
The methodology sets out how these studies are built and the rules they follow.