Gambling Harm Open Data
Five datasets on gambling harm in the United States. Free, machine-readable, CC BY 4.0, with a permanent DOI. No registration, no email gate, no rate limit, no API key. If you are on deadline, the raw CSV links are in the table below and you do not need to ask us first.
Every source these files draw on is a PDF. If this were also a PDF it would have added nothing.
The five datasets
| Dataset | Rows | What it answers | Refreshed | Download |
|---|---|---|---|---|
| Gambling Harm Index | 51 | What does each state actually publish about gambling harm — and how many people contacted its helpline? | Monthly, the 15th | CSV · Schema |
| Sportsbook Responsible-Gambling Tool Matrix | 14 | Which responsible-gambling tools does each major operator publish, and how hard are they to find? | Quarterly | CSV · Schema |
| Gambling Recovery Glossary | 125 | What does this term mean, and what should I write instead? | As needed | CSV · Schema |
| Quit-Gambling App Panel | 107 | How many quit-gambling apps exist, how new are they, and how many have been abandoned? | Monthly | CSV · Schema |
| State and Multi-State Self-Exclusion Programs | 12 | How does someone in this state actually enroll in self-exclusion, and what does it cover? | Quarterly | CSV · Schema |
Everything at once: the GitHub repository or the Zenodo archive.
Gambling Harm Index
Fifty states and the District of Columbia, one row each. Two layers, deliberately kept apart.
The helpline layer comes from one source for all fifty-one jurisdictions: the National Council on Problem Gambling's annual helpline report. Because it is one publisher applying one method everywhere, these are the only figures in the file that are comparable across states. We reproduce NCPG's published rate per 100,000 residents rather than recomputing it, so our number and theirs always agree, and we reproduce NCPG's own significance test on the year-over-year change rather than running our own.
The state-published layer is whatever each state chooses to publish about itself, with the period, the publisher and the source URL attached to every value. These are not comparable across states and we never rank on them. A self-exclusion count from one state and a treatment count from another are counts of different things over different periods under different rules, and adding them up would be wrong.
Thirty-two of the fifty-one jurisdictions publish no recurring, current gambling-harm count; eighteen of them publish nothing at all. That is the largest single finding in the file, and it is a finding rather than a defect. The file uses four distinct markers for absence, and none of them means zero: a state that publishes nothing, a series that exists but could not be retrieved, a number produced but not published in readable form, and a figure that exists but is not comparable. N/A here never means zero harm and never means we did not look.
The full monthly edition, with the narrative, lives at the Gambling Harm Index. The methodology is published in full.
Sportsbook Responsible-Gambling Tool Matrix
Fourteen operators, checked against their own public pages. Deposit, loss, wager and time limits; cool-off and timeout; in-product self-exclusion and whether it can be reversed; whether a written self-exclusion request is accepted; the account-closure path as distinct from self-exclusion; and the documented navigation to reach any of it.
The most useful column is the one that keeps saying "not published." No operator in the set publishes its per-state waiting period for raising a deposit limit. Almost none publishes the duration menu for a cool-off. Almost none carries a last-modified date on its responsible-gambling page, which is precisely why this file is dated and versioned quarterly instead.
There is no score column, no ranking and no recommendation, and there will not be one. A published tool may be hard to find, easy to raise, or narrow in scope, and reducing that to a number would hide the thing worth knowing. The per-operator page is at sportsbook responsible-gambling tools.
Gambling Recovery Glossary
One hundred and twenty-five terms across four sections: recovery language, industry and product terms, regulators and clinical instruments, and stigma-preferred language pairs. Each entry has a plain definition, a phrasing note, and a source or an explicit marker saying no authoritative source was found.
Thirty-two entries carry that marker, twenty-one of them sports-betting jargon. There is no standards body that defines "tilt" or "reload bonus," and the honest move is to say so rather than attach a citation that does not support the sentence. Every term has a stable slug, so the CSV joins cleanly against the published glossary.
Quit-Gambling App Panel
One hundred and seven apps surfaced by three fixed searches on the U.S. iOS App Store; sixty-three of them gambling-specific after hand review.
It is a panel, not a census. Apple's search endpoint returns at most fifty results per term, so every count from this file is a floor and never a total. What the panel shows is churn rather than choice: two-thirds of the gambling-specific apps launched within the last twelve months, a fifth have not been updated in six months or more, two-thirds have fewer than ten ratings, and the median app has three. Do not rank on the average rating; a five from two ratings is not a rating.
Betttr's own app is a row in this file, counted by the same rule as every other app, with no adjustment and no favorable placement. We are in our own dataset and we say so every time. The reader-facing version is at quit-gambling apps compared.
State and Multi-State Self-Exclusion Programs
Who runs the program, how to enroll, the direct URL, and — the column that matters most — what the exclusion actually covers. One state runs four separate tracks and a person has to enroll in each one they need. Another lets the person choose online-only or all gambling. A third holds casinos and internet gaming on entirely separate lists. Two rows that both say "self-exclusion" can mean very different things, so there is no column that reduces scope to a flag.
This is the least complete file in the set: eleven jurisdictions plus the multi-state program. Absence of a state here is absence of verification, not absence of a program, so do not compute a share of states from it. Extending it to full coverage is the top of the queue. The reader-facing version is the self-exclusion directory.
How to cite
For the dataset in general, cite the concept DOI — it always resolves to the newest version:
For a specific release, so your numbers stay reproducible, cite the version DOI:
BibTeX:
In a news story this is enough: Gambling Harm Open Data, Betttr LLC, September 2026, with a link to this page. Name the individual file if your figure comes from one, so a reader can find the same number. Journalists do not need permission and do not need to email us first.
License
Creative Commons Attribution 4.0 International. Free to use, including commercially, including inside a product, including by a competitor. The only condition is credit and a link.
Most of what is in these files is factual data, and in the United States facts are not copyrightable, so a data license here is closer to a norm and a courtesy than an enforceable claim. Given that, we picked the option every reuser, repository and journal data policy already recognizes without needing a paragraph of explanation. We want the data used. Credit comes from the DOI and the citation block above far more than from the license text. If attribution is inconvenient for you, take the data anyway and cite us when you can.
How these files are made, and what we refuse to do
No estimating, modelling, imputing or interpolating. Every number is a published number with a URL.
No composite score. There is no single harm number for a state and there will not be one, because the inputs are not comparable enough to earn it.
No ranking on figures that are not comparable.
No news articles as sources. Press-only figures with no primary source are excluded from the published-value columns entirely. We do not launder an article into a dataset.
Missing means missing, and says which kind. Each file documents its own markers in its column dictionary, and none of those markers means zero.
Every release is permanent. Corrections are appended with a dated note, never silently substituted, and the previous release stays live at its own tag and its own DOI. A method change is dated, listed, and never applied retroactively to a release that has already shipped.
Who makes this, and who pays for it
Gambling Harm Open Data is produced by Betttr LLC, maker of a recovery app; no operator or affiliate money.
Betttr LLC is a Delaware limited liability company based in Sherman Oaks, California. Its revenue comes from people who use its app, and from nothing else. It has never accepted money, sponsorship, data, advertising, placement or in-kind support from a sportsbook, a casino, a lottery, a gambling operator, an affiliate marketer, a gambling trade association, or anyone acting for them, and it will not. If that ever changes it will be disclosed on this page before the next release, and the change will be dated.
We are in our own data: the Betttr app is a row in the app panel, counted the same way as every other app.
We have a stake in gambling harm being taken seriously. That is a reason to check our arithmetic, and it is why every number here carries the URL it came from.
Corrections
Corrections go to support@betttr.net. State agencies, regulators, councils, operators and app developers who believe we have their number, scope or period wrong are the most valuable senders, and we correct on their word plus a source. We publish the corrections we find ourselves exactly the way we publish the ones we are sent. If a figure turns out to be unsupportable rather than merely wrong, we remove it and say that we removed it — we would rather have a hole with an explanation than a number we cannot stand behind.
The running log is in the repository changelog, newest first.
Frequently asked questions
Can I use this commercially, or in a product I sell?
Yes. The license is CC BY 4.0, which permits commercial use, redistribution and derivative work. Credit Betttr LLC and link to this page or to the DOI. You do not need to ask, and you do not need to tell us — though we would like to know, as a request rather than a condition.
Is there an API?
Not yet. The files are small enough that the raw CSV links on this page work as one: they are plain HTTPS, uncached-friendly, served directly from the repository, and stable across releases. The main branch always points at the newest release; a tagged URL or the version DOI gives you a frozen copy that will not move under a published analysis.
How is this different from the National Council on Problem Gambling's report?
NCPG publishes the helpline series, annually, as an appendix inside a PDF, and it is the single best source in the field — we reproduce their figures as published, without recomputing them. What did not exist before this is the pooling: the state-by-state layer of what each jurisdiction publishes about itself, in one machine-readable place, with the period and the source URL attached to every value and an explicit marker where a state publishes nothing. The Index is a normalization layer on top of their work, not a replacement for it.
You make a recovery app. Why should I trust your gambling-harm data?
You should not trust it, you should check it — which is exactly why every figure in every file carries the URL it came from, why the methodology is published in full, why the corrections policy is published before the data rather than after, and why our own app appears in our own app panel counted by the same rule as everyone else's. We take no operator or affiliate money. The design of these files is an argument that we are checkable: no estimates, no model, no composite score, no ranking on numbers that are not comparable.
If gambling is a problem for you or someone you know, the National Problem Gambling Helpline™ is 1-800-MY-RESET (1-800-697-3738) — call or text, 24 hours a day, free and confidential — or chat at 1800myreset.org. It is not a crisis line; for a crisis, call 911 or 988.