Skip to content
Against Certainty
Notes on probability, panic & better guesses

File 01 — Field Guide

The core ideas, in plain words.

Thinking in odds, base rates, Bayes, signal versus noise, and calibration. No equations until you ask for them.

The method

Four moves before you panic.

The whole guide collapses into this — a correction routine for the moment a single data point tries to write your whole story.

  1. 01

    Name the belief

    What do I actually think is true right now? Say it in one sentence.

  2. 02

    Check the base rate

    How often does this kind of thing happen anyway, before I look at today?

  3. 03

    Weigh the evidence

    Is this a strong signal, or just loud noise that feels like one?

  4. 04

    Update, don't collapse

    Move your estimate. A little. Not all the way to the worst case.

Errata: common errors.

The specific ways a brain talks itself out of the base rate.

  • Base-Rate NeglectYou judged the odds by how vivid the story felt, not by how often it actually happens.Your brain loves a specific story and forgets to ask how often that story is even true. The base rate is the background frequency of a thing before you add any details, and it is usually the single most useful number you have. Skip it, and one striking detail can talk you into something the odds quietly rule out.e.g. A prospect goes quiet after a great call and you decide the deal is dead. But most deals go quiet because people are busy, not because they hate you. 'Silent, then signs' is far more common than the panic admits.
  • Availability BiasWhatever springs to mind fastest gets treated as whatever happens most.The brain estimates how often something happens by how quickly examples come to mind, and dramatic or recent events come fastest. So the memorable crowds out the typical, and your sense of risk drifts toward whatever you last read or felt. Ease of recall is a fact about your memory, not about the world.e.g. Two indie apps fail loudly on your timeline this week and you conclude the market is dead. You never saw the hundreds that quietly did fine, because nobody posts a thread titled 'business as usual.'
  • Small-Sample PanicYou read a trend into three data points that are mostly noise.Small samples swing wildly by pure chance, but the brain pattern-matches anyway and calls the swing a direction. The fewer observations you have, the louder randomness is relative to any real signal. A short streak feels like proof; usually it is just variance in a costume.e.g. Two bad sales days in a row and you start rewriting the pricing page. Two days is not a trend. It is two days, and one decent Wednesday would erase the whole story.
  • Emotional EvidenceYou mistook the loudness of a feeling for the strength of the case behind it.Feelings arrive fast and certain because they are built for speed, not accuracy. The body reacts before the facts are in, so dread can feel exactly like knowledge while pointing at nothing. A bad feeling is data about your state; it is not a conclusion about reality.e.g. You wake up sure the relationship is failing because you feel hollow. The feeling is real, but it might be a bad night's sleep talking, not the relationship. Panic is not a probability estimate.
  • False CertaintyYou rounded a strong hunch up to a known fact because 'unsure' felt unbearable.The mind treats 'unsure' as a problem to solve rather than a state to hold, so it quietly converts 'probably' into 'definitely.' Confidence feels like competence, which makes overstatement seductive. But certainty you did not earn is just a guess with better posture.e.g. You decide you know exactly why a customer churned and build the whole next quarter around it. You have a theory, not a finding. The honest version starts with 'my best guess is,' and keeps looking.
  • Confirmation BiasYou went looking for proof you were right and stepped over everything that said otherwise.Once a belief is in place, the brain treats agreeable facts as obvious and disagreeable ones as suspect. It is cheaper to defend a position than to revisit it, so you collect support instead of testing the claim. The result feels like research and works like a closing argument.e.g. You think a feature is beloved, so you reread the five glowing reviews and write off the twenty 'meh' ones as people who don't get it. You were tallying applause, not checking whether the room liked it.
  • Negativity BiasOne bad signal outweighs five good ones, and you call the result a fair read.Threats were once expensive to miss, so the brain gives bad news extra volume and a longer shelf life. One criticism can outshout ten compliments and feel like the truer one. That weighting is a survival reflex, not an accurate scale.e.g. Nineteen people say the launch was great and one says it was overrated, and guess which sentence you replay at 2 a.m. The lone critic is not more honest. He is just louder in your head.
  • Regression-to-the-Mean BlindnessYou took credit for a rebound that was going to happen with or without you.Extreme results are partly luck, and luck does not repeat, so after an unusually bad or good stretch things tend to drift back toward normal on their own. The brain misses this and hands the recovery to whatever it happened to be doing. You learn a false lesson and trust it next time.e.g. Your worst-ever traffic week is followed by a normal one, and you decide the panicky midnight tweak fixed it. More likely the bad week was a fluke that corrected itself, and the tweak just took the bow.