The sentence turns up in every conversation about drugs, usually as the closing argument.
I’ve done it a hundred times and I’m fine.
I have said it. I have believed it. It feels like the strongest evidence a person can offer, because it is the only kind they have collected personally: a hundred data points, all pointing the same way, gathered at some risk. Compared to a stranger telling you it is dangerous, it feels overwhelming.
It is worth almost nothing, and I can show you exactly how much almost nothing is.
What surviving a hundred times actually proves#
Suppose each use carries some probability of killing you. Call it p. You do not know p. That is the entire problem, and it is the thing your hundred uses were supposed to settle.
If each use is independent, the chance of getting through n uses untouched is:
(1 - p)^nNow run it backwards. You survived a hundred times. What does that rule out?
There is a tidy statistical result for exactly this situation, sometimes called the rule of three: if you have run n trials and seen zero events, the upper bound of the 95% confidence interval for the event rate is about 3/n.
A hundred uses with nothing bad happening puts the upper bound at 3/100, or 3% per use.
Sit with that for a second. Your hundred-times-and-fine, your hard-won personal evidence, the thing you would bet against a stranger’s warning, is compatible with a per-use death rate of up to three percent.
Three percent per use is not safe. Three percent per use is a coin you would never flip. At 3%, the chance of getting through a hundred uses is about 5%, which means you would be the unlucky-to-be-lucky one in twenty, and the other nineteen versions of you are not available to comment.
| Uses survived | Upper bound on per-use risk |
|---|---|
| 10 | 26% |
| 50 | 6% |
| 100 | 3% |
| 500 | 0.6% |
| 1 000 | 0.3% |
Even a thousand uses only gets you down to “probably under one in three hundred”. For something you do twice a week, one in three hundred arrives on a schedule.
This works fine in the other direction. If something bad has happened, that is strong evidence the risk is real. One overdose tells you far more than a hundred non-overdoses do. Evidence is not symmetric here, and the cheap evidence is the useless kind.
Why survivors are the only ones talking#
There is a second problem sitting underneath the first, and it is the reason the sentence is so persuasive despite being so weak.
Everyone who says “I’ve done it a hundred times and I’m fine” is, definitionally, someone it did not kill. You are not going to hear the counter-argument from the people who ran the same experiment and lost, because the way you lose is by not being available for interview.
So the pool of available testimony is filtered. Every anecdote you receive about how safe something is comes pre-selected for survival. It is not that people are lying. It is that the dead are not contributing, and the sample looks reassuring for a mechanical reason that has nothing to do with the drug.
This is why “everyone I know does it and they’re fine” is a weaker statement than it sounds. Your social circle is also a survivor pool, and it gets quieter in ways that are easy to attribute to people moving away.
The part that actually breaks the maths#
Everything above assumes each use is an independent roll of the same dice. For contaminated supply, that assumption is wrong, and it is wrong in the direction that hurts.
Fentanyl and its relatives do not arrive evenly distributed through a batch. They arrive in hot spots. Powder does not mix like liquid. A gram that has been cut somewhere upstream by someone with a spoon and no reason to care has regions of very different concentration, which is where the phrase “chocolate chip cookie” comes from in harm reduction circles: the dangerous part is not spread through the dough, it is in lumps.
That changes the model in three specific ways.
Your hundred uses may have been a handful of batches. If you bought from the same source for months, you did not run a hundred independent trials. You ran maybe eight, and repeated each one a dozen times. The effective sample size is not a hundred, it is closer to eight, and the same calculation on eight trials gives an upper bound of over 30%.
A safe test does not certify the rest. Testing a corner of a bag and finding nothing tells you about that corner. With clustered contamination, the negative result carries much less information about the rest of the bag than intuition suggests. This is the single most misunderstood thing about reagent testing, and it is not an argument against testing, it is an argument for testing properly: homogenise the whole batch first, then sample it.
Risk is concentrated in transitions. Not in the hundredth use of a familiar bag, but in the first use of a new one. New source, new batch, new dealer, or the same dealer with new supply. The maths and the anecdotes agree here: the dangerous dose is disproportionately the first from an unfamiliar batch.
What this changes in practice#
If you are going to use anyway, and I have never had much luck telling people otherwise, then your history is not the safety evidence you think it is. So put the caution where the actual risk is.
- Treat every new batch as a first time. Not the first time ever, the first time with this. Start low, wait properly, and do it with someone present. Your tolerance is real; your knowledge of what is in the new bag is not.
- Homogenise before you test or dose. Crush and mix the whole amount. It turns clustered risk back into something closer to average risk, which is the one situation where the naive maths becomes true and works in your favour.
- Naloxone in the room, not in the drawer at home. The cost of being wrong is asymmetric, so the preparation should be too.
- Never alone on a new batch. The entire failure mode of an opioid overdose is that nobody notices in time.
- Count batches, not doses. If you must keep a mental tally of how many times it has been fine, count sources. It is a much smaller and more honest number.
The honest version of the sentence#
I’ve done it a hundred times and I’m fine is not a lie. It is a true statement about the past that people use as a false statement about the future.
The accurate version, and it is much less comfortable, is something like: I have survived a hundred draws from a distribution I have never measured, mostly from a small number of batches, and the strongest thing I can conclude is that the risk is probably under a few percent each time.
Nobody says that at a party. But it is what the evidence supports, and pretending otherwise is how people end up describing a friend in the past tense.
I got out of this. Not through good judgement, which I did not have at the time, and not because I understood the arithmetic, which I certainly did not. I got out because I was lucky for long enough to get help, and I know exactly how thin a thing that is to have depended on. If you are still counting your hundred, the number does not protect you. It just means you have not met the tail yet.
/Henrik
