To tell a craving pattern from a coincidence, treat every suspected repeat as a hypothesis and run three checks: compare it against your base rate, count the times the same situation passed quietly, and use it to make one specific prediction about the future. A pattern that survives all three deserves a plan. One that fails was probably noise, and learning that is genuinely useful, because it stops you preparing for the wrong thing.

Anyone who logs cravings meets this problem within a few weeks. Two gambling urges on consecutive Fridays feel like a discovery. Three restless, vape-heavy mornings feel like a diagnosis. Sometimes they are the start of real insight. Sometimes they are random events clumping together, the way random events do. The skill is telling which, before you reorganise your life around either conclusion.

Why coincidences look so convincing

Short runs of randomness are not smooth. If urges arrive irregularly, some weeks will contain clusters and some will contain gaps, with no deeper cause at all. Three late-night scrolling sessions in one week can happen by chance inside a month of ordinary variation.

Memory adds a second distortion. Hits are vivid and easy to recall; quiet passes leave no mark. You remember the two Fridays when the craving arrived at 9pm. The Fridays when nothing happened do not feel like data, so they get dropped from the story. What remains looks like a law: "Friday nights always beat me."

Researchers who study craving treat memory with the same caution. Because craving fluctuates across ordinary daily life, the field relies on repeated in-the-moment ratings, known as ecological momentary assessment, rather than asking people to reconstruct their urges from memory later. Your log deserves the same discipline: what you happen to remember is not the same as what actually happened.

There is a real cost to believing a false pattern. You prepare for the wrong hour, blame the wrong cue, and collect a piece of evidence against yourself that was never true. There is also a cost to the opposite mistake, dismissing a genuine pattern as random and walking into the same difficult situation unprepared. The three tests below exist to protect you from both errors.

Real cue patterns do exist

Scepticism about your own log should not slide into scepticism about patterns altogether. Learned associations between a habit and its context are well established. WithYou, a UK drug and alcohol support charity, describes triggers as reminders formed by association with past use, and notes they can be external, such as people, places, and times of day or week, or internal, such as moods and physical sensations. Smokefree.gov groups smoking triggers around routines, emotions, social situations, and withdrawal, and recommends tracking which situations produce urges.

The research base is substantial. A 2022 systematic review and meta-analysis in JAMA Psychiatry, covering 237 studies and more than 51,000 participants, found that cue exposure and craving were prospectively associated with later drug use and relapse. Among the findings, cue-induced craving showed some of the strongest associations, and same-day links were stronger than longer lags. In plain terms: what surrounds an urge, close in time to the urge, genuinely matters.

So the question is never "do patterns exist?" They do. The question is narrower and more honest: is this specific repeat in my log one of them, or am I staring at clumped randomness? That is a question you can test.

Test one: compare the pattern against your base rate

Before concluding that Fridays are your problem, count two numbers: how many Fridays had an urge, and how many urges you had in total.

If you logged nine urges in a month and three of them landed on Fridays, Friday carries real weight. If you logged nine urges spread across six different days and three happened to fall on Fridays, "Friday" explains very little, because urges were happening anyway. A pattern should stand out from your ordinary rate, not just sit on top of it.

The same logic applies to times, places, and feelings. "I always crave when I'm stressed" means little if you are stressed most days and crave on a fraction of them. "Three of my four strongest urges this month followed an argument" is a ratio you can work with. Write the two numbers down before you let the story form. If you are unsure how much history is enough for this kind of comparison, How Many Logs Do You Need Before a Pattern Means Anything? covers the sample-size side in detail.

Test two: count the times nothing happened

This is the check most logs skip, because quiet events feel like non-events. Go back through the same period and find the occasions when the suspected cue was present but the craving was weak or absent.

Suppose your hypothesis is "being alone after a night shift triggers the urge to drink." Look for every post-shift evening, not just the bad ones. If you find two difficult nights and four quiet ones, ask what differed. Perhaps on the quiet nights you ate before driving home, took a different route past the shop, or had arranged to call someone. Those exceptions are not failures of the pattern. They are the pattern's most informative part, because they point at the conditions that protect you.

A coincidence tends to dissolve under this test: the misses turn out to be just as common as the hits, and you had simply never counted them. A genuine pattern survives it with texture, failing sometimes, but failing for reasons you can name.

Test three: let the pattern make one prediction

Retrospective explanations are cheap. Given any log, a tired mind can construct a story that fits it perfectly. Prediction is the discipline that separates storytelling from evidence.

Before the next occurrence of your suspected pattern, write down exactly what you expect:

  • When: this Saturday, between 9pm and 11pm
  • Situation: alone at home after the match finishes
  • Early sign: refreshing the scores app without a reason
  • What I expect: an urge strong enough to notice, around 6 out of 10 or more

Then wait, and check honestly. One correct specific prediction is worth more than three explanations fitted after the fact. If the prediction fails twice, downgrade the hypothesis rather than quietly revising it to match what happened. A pattern that only ever explains the past and never anticipates the future is not yet a pattern you can plan around.

Keep the prediction specific. "I will feel bad at some point this weekend" can always be confirmed. "The urge will arrive within an hour of the final whistle" can be wrong, and that is exactly why it is useful.

Give the evidence time to grow

Even a pattern that passes all three tests should be held lightly at first. A handful of entries can suggest a question; it takes longer to establish an answer. This is the same restraint built into Hard Bit's on-device forecast: it makes no prediction at all below five entries and leans fully on your own logged history only at around twenty. You can read how that fits alongside logging and milestones in the Hard Bit overview. Whether or not you use an app, adopt the same rule of thumb. Let confidence grow with observations instead of arriving all at once.

Watch for temporary clusters too. Illness, a holiday week, payday, a shift change, or a run of bad sleep can manufacture a convincing-looking pattern that evaporates when life returns to normal. A cluster with an obvious one-off cause is not a stable pattern, so note the cause and keep logging through ordinary weeks before you commit to a plan.

A structured weekly look at the entries makes this much easier than glancing at the log in anxious moments. How to Review a Week of Craving Logs walks through a calm review routine, and How to Identify Your Craving Triggers and Difficult Times explains which fields to capture so the tests above have something to work with.

When a pattern survives, act on it once

A confirmed pattern earns exactly one practical preparation, not a new identity. "I am the kind of person who always fails on Saturdays" is a story. "Saturdays after the match are difficult, so I will charge my phone in the kitchen and arrange to eat with my brother" is a plan.

Choose one change that intervenes early in the sequence, before the point where willpower usually gets the bill. For a ready-made structure, A Five-Minute Plan for the Start of a Craving helps you script the first move, the bridge action, and the support contact in advance.

Then keep testing. Real patterns shift as your routines, seasons, and relationships change, and a preparation that worked in October may need adjusting by spring. The log is never finished, and neither is the hypothesis.

When the log points beyond self-tracking

Sometimes the three tests confirm something more serious than an inconvenient hour. Frequency climbing week on week, amounts escalating, withdrawal symptoms when you stop, money problems from gambling, or repeated failure of plans that used to work, all of these are reasons to bring in more support, not reasons to hide the record.

If you may be physically dependent on alcohol or sedatives, do not stop abruptly on the strength of a pattern in a log; withdrawal can be dangerous, so seek medical guidance first, and urgent help if severe symptoms or immediate danger appear. For smoking or vaping, official quit services can explain evidence-based support and medicines rather than leaving you with distraction tips alone. For gambling, specialist support and practical financial barriers matter more than any diary. A tracker can help you describe what has been happening to a clinician or support worker. It cannot assess risk, diagnose anything, or replace them. Our editorial policy explains how we keep that boundary clear.

Frequently asked questions

How many times must something repeat before it counts as a pattern?

There is no universal number, which is why the three tests matter more than a count. Two repeats with a strong base-rate advantage, visible counter-examples you can explain, and one successful prediction carry more weight than six repeats you never examined. As a rough discipline, hold any conclusion loosely until it has survived several ordinary weeks of logging.

Can a false pattern still teach me something?

Yes. A hypothesis that fails testing still tells you what you were watching for, and why. If "Mondays are dangerous" dissolved under the counter-example check, you have learned that your difficult moments are scattered rather than scheduled, which points toward a general in-the-moment plan instead of a Monday-specific one. A disproved pattern is a log doing its job.

What if the pattern changes once I start watching it?

That is expected, and it is not cheating. Monitoring changes behaviour: once you know the final whistle is your danger point, you prepare, and the pattern softens. Record the change rather than abandoning the log, and treat the weakening as progress, while staying honest if the same cue reappears in a new disguise.

Should I trust my own impression or an app's forecast?

Treat both as hypotheses. An impression is fast but forgets the quiet misses. A forecast built only from your logged entries is immune to mood but blind to anything you did not record. When they agree, plan confidently. When they disagree, that disagreement is itself worth a closer look at what the last few weeks actually contained.