How the index works

Sources, search rules, tagging and known gaps, so the index can be checked and improved by anyone.

Coverage

The index tracks new peer-reviewed articles, reviews, commentaries and preprints about gambling and closely related activities: sports betting, online gambling, casino and machine gambling, lotteries, loot boxes and simulated gambling in video games, prediction markets and event contracts, gambling-like trading, and the policy, treatment, prevention and measurement research around them. It is not limited by country or language, though English-language journals dominate the sources.

It does not try to cover gaming disorder in the video-game sense, gambling as a metaphor, decision science that uses lotteries or gambling tasks only as a method, or industry market reports.

The index reaches back to January 2022: studies published since then were collected in one pass from PubMed, OpenAlex and Crossref, month by month, and are searchable alongside everything added since. Studies found that way are dated by their publication, not by the day they were collected. Older work is not covered; for anything before 2022, search PubMed or Google Scholar directly.

Sources

Each night a job queries:

• PubMed (NCBI E-utilities): a saved query over titles, abstracts and MeSH headings, by entry date. PubMed brings the strongest metadata (abstracts, affiliations, publication types, PubMed Central links).
• OpenAlex: a boolean search over titles and abstracts, by publication date, limited to journal sources (its repository records, from Zenodo and the like, are mostly promotional spam). OpenAlex brings journals outside MEDLINE, including law, economics and non-English titles, plus author country codes and free-full-text links. Because its search is loose, a study that only OpenAlex supplies must name the subject in its title, or mention it at least three times in the abstract with the first mention in the opening sentences.
• Crossref: everything registered by the journals devoted to gambling (Journal of Gambling Studies, International Gambling Studies, Journal of Gambling Issues, Journal of Behavioral Addictions, Critical Gambling Studies, Gaming Law Review), by DOI registration date.
• Journal feeds: the tables of contents of the gambling journals and of the addiction journals that publish gambling work most often, plus medRxiv's addiction-medicine preprints.

The exact queries and feed addresses are in config/sources.yml. The table at the bottom of this page shows what each source returned on the last run and which ones failed.

What gets kept

A record is kept when its title or abstract contains at least one strong gambling term (gambling, bettors, sports betting, casino, slot machine, lottery tickets, loot box, prediction market, Kalshi, and the words for gambling in the other languages the index sees most). A few terms count only alongside a strong one, because on their own they usually mean something else: self-exclusion (also a term in social psychology and machine learning), wager, lotteries (the choice options of decision science), gamble used as a metaphor, poker and roulette. Records that mention only those, or a bare "lottery" or "bet", are kept but marked as possibly related and left out of the main lists. Mentions of the Iowa Gambling Task and similar laboratory tasks, "testing by betting" in statistics, and "gambling for re-election" in politics are ignored when deciding relevance. Journals whose papers reach the index only through stray mentions (optics, microbiology, chemistry) are skipped.

Duplicates are merged by DOI, then by PubMed ID, then by normalised title. When sources disagree, PubMed's fields win, then Crossref's, then OpenAlex's; the free-full-text link comes from whichever source has one.

Dates

Two dates appear on every entry. The date in the margin is when the study entered this index, which is the useful "new" date: journals publish online first, then in an issue, and databases index weeks later. The publication date after the journal name is the earliest date the sources report. The first build of the index used the sources' own dates to backfill, so early entries carry an approximate "added" date.

Summaries

Each entry shows the first sentence or two of the abstract, capped at about 360 characters, as a pointer. Full abstracts are not republished; the title links to the publisher's record and the "free full text" tag links to an open copy (PubMed Central or the location OpenAlex reports) when one exists.

Tagging

Every study carries topics, study designs and a setting, assigned in three layers. The first that has an answer wins.

1. Hand corrections. A study corrected by the editor, in config/corrections.yml, keeps that correction through every nightly run and every rebuild. Reader reports end up here.
2. The databases' own indexing. For PubMed records, the publication types assigned by NLM's indexers (randomized controlled trial, systematic review, meta-analysis, clinical trial protocol, editorial, letter, case report) decide those designs, and MeSH geographic headings (Finland, Ontario, Iran) give where the study was done.
3. Keyword rules, listed in full at the bottom of this page, for everything else. They are kept simple so that anyone can read them and propose a fix. A topic needs a match in the title or repeated matches in the abstract, so a study that mentions "policy implications" once is not tagged as a policy study. Design rules look for authors describing their own study ("this randomised controlled trial", "we conducted a systematic review") rather than studies they cite, and a protocol is never also shown as a trial.

The rules still make mistakes; the hand-checked scores below say how often. One caution about those scores: the rules were adjusted while part of that set was being checked, so the set flatters them. On a separate random sample of 59 studies, checked before any further change, the rules placed 95% correctly in or out of the index, gave the right setting for 91% and the right key topics for 86%, but gave the complete, correct set of designs for only 59%. Most design misses were designs left blank, not wrong ones; a blank design means "not recorded", not "none".

Setting and author location are different things. The setting shown on each entry is where the study was done, from the MeSH headings, the countries the title names, or, failing those, the countries the abstract names when there are few enough to be the setting rather than a comparison. Author location is where the authors work, from their affiliations (OpenAlex country codes, or country names in PubMed affiliation strings). An Iranian survey with one co-author in Australia has its setting in Iran and its authors in Iran and Australia. Both can be filtered on; only the setting is shown on the entry.

When the rules change, the next nightly run re-reads every study in the index from its sources and applies the new rules to all of them, so old entries are tagged the same way as new ones.

News

News comes from the feeds of regulators, courts, the press and trade outlets, and from Google News searches for a few phrases. A headline is kept when it matches gambling terms, categorised by keyword (public health and treatment, research in the news, prediction markets, regulation and law, industry), and capped at ten items per source per day. The home page shows everything except pure industry news; the news page shows everything. Headlines link to their sources; nothing is rewritten.

Trials and preregistrations

The trials page lists registered studies rather than published ones. Each night the index queries ClinicalTrials.gov (API v2: studies whose conditions mention gambling, plus studies that mention sports betting, wagering, loot boxes or prediction markets anywhere), ISRCTN (the registry's query API, for "gambling") and OSF Registries (public registrations whose title or description mentions gambling, betting, wagering, loot boxes or prediction markets). Records pass the same relevance rules as studies, so a Parkinson's trial that uses the Iowa Gambling Task is dropped. An ISRCTN record that quotes a ClinicalTrials.gov number is treated as the same trial.

Status is the registry's own (ClinicalTrials.gov's "unknown" means the sponsor has not updated the record for two years; ISRCTN's is derived from its recruitment and end dates), and "results posted" means results appear in the registry, not that a paper exists. When a trial's status changes between runs, the previous status is shown for a while. Not covered: ANZCTR, PROSPERO, the EU register and AsPredicted, which have no usable public interface.

Bills

The bills page tracks legislation with action in the past year. State bills come from Open States (API v3): a full-text search for each of a short list of terms (sports betting, sports wagering, gambling, casino, lottery, wagering, prediction market, event contract, self-exclusion, problem gambling, daily fantasy, sweepstakes, video gaming terminal, internet gaming) over bills updated since the last run. Federal bills come from GovInfo, whose listing of bill texts modified in the window is matched by official title, and Congress.gov, which supplies the latest action; known federal bills are re-checked a few at a time so their stage stays current between text versions.

A bill is kept when its title, abstract or subject names gambling, betting, a lottery, casino, gaming regulation, prediction markets or self-exclusion; the rules are bills_include in the taxonomy file. Appropriations and omnibus bills that mention gambling only in their text are usually missed, which is deliberate. The stage (introduced, in committee, reported, passed one chamber, passed both, enacted, vetoed, failed) is derived from the legislature's recorded actions, or from the wording of the latest action for federal bills, and can lag the legislature's site by a day or two.

Courts

The courts page comes from CourtListener's search API: opinions filed in the window and federal dockets (RECAP) with filings in the window, for a query built from gambling terms (sports betting, gambling, casino, lottery, gaming compact, Indian Gaming, event contracts, prediction market, Kalshi, Polymarket, sweepstakes casino, daily fantasy, loot box, self-exclusion). A record is kept when its case name, snippet or docket entries match the cases_include rules. The court level comes from CourtListener's jurisdiction codes. Criminal cases that mention gambling in passing do get through, and state trial courts and most state dockets are not in CourtListener at all, so this is a watch list, not a complete record.

The digest

The weekly digest is written by the editor from a ranked shortlist the index produces each Monday (data/weekly/latest.json). The shortlist scores studies by journal, design, free availability and topic, which is a convenience, not a quality judgment. Choosing what to write about and what to say is editorial.

Search

Search covers every study in the main index: titles, authors, journals, the short summaries and the index's tags, with filters for topic, design, setting, author location, year, language and free full text. It runs in the reader's browser from a static index built each night, so no query is sent to a server or logged anywhere. It does not search full text, and accents are ignored.

What this index is not

It is not a systematic review, and its coverage has not been formally validated against a gold standard. Sources fail, queries miss synonyms, and some journals have no feed. Use it to see what is new, then search properly before claiming that something has not been studied.

Corrections

Write to editor@gambling-research.com or open an issue in the repository. For a misfiled study, include the DOI and what is wrong. For a missing journal or feed, include the feed address. Confirmed fixes go into the corrections file, where they stay fixed.

Data and code

The whole index is published at /data/papers.json (where settings holds the study's countries and countries its authors'), /data/news.json, /data/trials.json, /data/bills.json and /data/cases.json under CC0. The code is MIT-licensed. Anyone can run it for another field by changing the configuration files.

Source health

Last run started 2026-11-01 13:16:34 UTC (nightly). New: 13 studies, 40 news items, 0 trials or registrations, 5 bills, 0 court records.

Recent runs

Checking the rules

138 studies were checked by hand and their correct tags recorded in tests/gold/labels.yml. After every nightly run the index is scored against them; if a change to the rules makes the scores fall below the thresholds, the run stops and nothing is published. Latest scores (2026-11-01 13:30:58 UTC):

7 studies carry a hand correction in config/corrections.yml. The whole index was last re-read and re-tagged under the current rules (version 3) on October 25, 2026.

Topic rules

A study is tagged with a topic when its title or abstract matches the patterns below (case-insensitive regular expressions) often enough: a match in the title counts twice, a match in the abstract once, and the topic needs the number of matches in the second column. The same rules are in config/taxonomy.yml in the repository, where anyone can propose a change.

Design rules

For PubMed records, the publication types assigned by NLM's indexers decide whether a study is a trial, a review, a protocol, a commentary or a case report; the patterns below decide the rest, and decide everything for studies outside PubMed. The title patterns apply to the title; the abstract patterns are phrased to catch authors describing their own study rather than studies they cite. Where two designs cannot both be true, the first one listed wins: a protocol is not also shown as a trial, and a review or commentary is not also shown as a survey.

PubMed publication types