What it sounds like before it goes wrong.
That is the question we set out to answer, and the material answers it with a silence: 49 posts of 3,299 are about the time before the loss. Almost nobody writes on the evening it is still going well. What is written down is afterwards — the money, the register, the days counted one by one. Here is all of it, with every denominator, every rule and every number the material cannot carry.
B4GAMBLE earns commission from licensed casinos. Nothing in these figures links to one; the site menu does.
What was read
Three open forums: a Swedish debt forum (threads tagged gambling addiction, gambling debt and casino), a Swedish parenting forum's sensitive-topics room, and five subreddits. A post counts only when it stands on its own words — a reply that consists of a quoted earlier post says nothing of its own, and 2,505 such posts were dropped before coding so the same sentences were not counted twice.
| Swedish debt forum | 1,695 coded | 1,695 posts that stand on their own words, of 4,211 collected |
|---|---|---|
| Swedish parenting forum | 604 coded | 604 posts that stand on their own words, of 605 collected |
| 1,000 coded | a fixed selection from 3,178 posts that stand on their own words, of 3,295 collected | |
| Coded in total | 3,299 posts | 3,299 coded records, 1 per post, every field allowed to be left unknown |
The two Swedish forums are coded in full, so their figures are counts, not estimates. The Reddit third is a fixed 1,000-post selection and not a representative sample: the draw follows the posts' running numbers, and two of the five subreddits fell outside it entirely. Of the Reddit posts coded, 918 are Swedish and 82 Danish.
Who writes
Most posts in a thread are answers to somebody else. That is why the figures below are split: a share over all 3,299 posts and a share over the 925 posts where a person writes about their own gambling are two different claims, and both denominators are printed everywhere on this page.
| Writes about their own gambling | 925 | 3,299 posts (28.0 %) |
|---|---|---|
| Writes about someone else's | 127 | 3,299 posts (3.9 %) |
| An answer or a comment with no gambling of their own | 2,210 | 3,299 posts (67.0 %) |
The register
A self-exclusion register — Sweden's Spelpaus, Denmark's ROFUS or a block set with one operator — is named in 174 posts. Those 174 posts are the denominator of every line that follows, and they are the clearest thing in the whole material: when the register comes up, a way round it or a way back to gambling comes up with it in more than a third of the posts.
| Names a self-exclusion register | 174 | 3,299 posts (5.3 %) |
|---|---|---|
| Describes signing up | 130 | 174 posts naming a register (74.7 %) |
| A way round it or a way back, all three together | 65 | 174 posts naming a register (37.4 %) |
| — played somewhere else during the pause | 36 | 174 posts naming a register (20.7 %) |
| — played again once the period ended | 23 | 174 posts naming a register (13.2 %) |
| — tried to have the pause lifted early | 15 | 174 posts naming a register (8.6 %) |
| Describes sitting the period out | 2 | 174 posts naming a register |
| Describes extending it | 4 | 174 posts naming a register |
| Stands at the end of a pause, the day the accounts reopen | 8 | 3,299 posts |
The last three rows are the most easily misread in the study. Two posts about sitting the period out does not mean few people do it: a quiet month gives nobody a reason to write a forum post. The same holds for the reopening day — eight posts in 3,299 is not a measure of how much that day matters, it is a measure of how little is said about it anywhere.
The money
These four rows have the 925 posts about a person's own gambling as their denominator. One of the three forums is a debt forum, where loans and collection are the subject of the room, so none of this can be read as how often a player borrows.
| Quick credit: a short-term loan, a credit card or borrowing from family | 161 | 925 posts about their own gambling (17.4 %) |
|---|---|---|
| Names a sum | 322 | 925 posts about their own gambling (34.8 %) |
| — of those, more than 100,000 kronor | 209 | 322 posts naming a sum (64.9 %) |
| Enforcement or debt collection | 62 | 925 posts about their own gambling (6.7 %) |
Whether a sum is named at all was coded reliably (two coders agreed on 95 % of the re-coded posts); which band it falls into was not (71 %). So the figure published is the coarse one — named or not named — and above or below 100,000 kronor. People write the number down when it is big; this is not the typical loss.
What is said to have helped
Counted only where a post names something as having helped. Most posts name nothing, so every share here is small by construction, and the order matters more than the size.
| Treatment or a helpline | 109 | 925 posts about their own gambling (11.8 %) |
|---|---|---|
| Time and routine | 81 | 925 posts about their own gambling (8.8 %) |
| A register or a block with an operator | 69 | 925 posts about their own gambling (7.5 %) |
| Telling someone | 55 | 925 posts about their own gambling (5.9 %) |
| Handing the money to someone else | 54 | 925 posts about their own gambling (5.8 %) |
| Blocking software or a block at the bank | 22 | 925 posts about their own gambling (2.4 %) |
| A debt plan | 22 | 925 posts about their own gambling (2.4 %) |
| Says plainly that nothing helped | 11 | 925 posts about their own gambling (1.2 %) |
This is what people mention. It is not evidence of what works, and it is not a ranking of treatments.
The days people count
102 posts carry the writer's own count of days without gambling. Weeks, months and years were converted to days. A counter is posted as a milestone, which is exactly why the first month is almost missing: it is lived, not published.
| Posts with a count of their own | 102 | 3,299 posts |
|---|---|---|
| Middle value of the count | 365 days | 102 posts with a count |
| A year or more | 52 | 102 posts with a count |
| Under 30 days | 7 | 102 posts with a count |
What this material cannot show
Every item below is a claim we could have made from these numbers and do not make.
- Not shares of people. A post is not a person, the same person can write many posts, and nobody here was asked a question — they all chose to write.
- Not how often anything happens. Silence is not absence: whoever sits out a pause quietly has nothing to post about, which is why two posts about waiting it out cannot be turned into a rate.
- No payday story. Payday or the small hours are named in 25 of the 925 posts about a person's own gambling (2.7 %). That will not carry a claim about the week the salary arrives, and we publish none.
- No change over the years. The share moves up and down between years with no direction, and the mix of forums changes from year to year, which is enough on its own to produce a false trend. No year-on-year figure is published here.
- Nothing Danish. 82 Danish posts were coded and ROFUS is named 7 times in the whole material. Danish figures need Danish material, collected on its own.
- The Reddit third is a fixed selection of 1,000 posts out of 3,178, not a representative sample, so no figure here describes Reddit, or Sweden, as a population. The two Swedish forums are coded in full.
- Two thirds of the posts are answers and comments. A share over 3,299 posts is therefore far smaller than the same share over the 925 posts about a person's own gambling, and both are printed.
- The question in the title is not answered here. 49 posts of 3,299 place anyone before the trouble; the one forum where people write while still playing does not serve this collection.
How it was done
- One unit is one post. Each is judged on its own, without the thread around it, against a closed scheme of twelve fields in which every field may be left unknown.
- No quotations anywhere. Everything is paraphrased: no usernames, no thread addresses, no quoted sentences.
- A light language model coded all 3,299 posts in batches of at most 50. Larger batches were thrown away: at 80 and at 150 posts the coder began inventing triggers and reasons the posts did not contain. Coverage was checked post by post, by identifier.
- 165 posts (5 %) were coded again, blind, by a stronger model. Agreement runs from 0.71 to 0.96 by field, and from 0.90 to 0.99 on the coarse questions this page actually asks. Nothing fell below 0.70, so no field was dropped; the one weak field is published only in its coarse form.
- No share is reported on fewer than 30 posts. Every share in the file carries a 95 % interval.
- The posts themselves are not published. Researchers can have the collection under their own ethical approval.
The numbers as a file
Every figure on this page, and the counts behind them, in one CSV: section, name, scope, count, denominator, share, and the 95 % interval where there is one. Reuse it, check it, or disagree with it.
Free to use with attribution (CC BY 4.0). The file holds aggregates only — rows, denominators and shares. No post text, no users, no thread addresses.
What the numbers are for
The register appears in this material mostly at its two edges: the day a person signs up, and the day it ends. The second one is almost unwritten — eight posts in 3,299. That day is what the plan below is built around, and the ten steps are free and need no account.
Sources
The forums are open and public; they are described here in words and never linked to a thread, because no person in this material asked to be quoted. The regulators' own figures about Spelpaus and ROFUS, with their sources and files, are on their own page.