Texas Sharpshooter
Cherry-picking clusters in data after the fact and treating them as meaningful — like shooting a wall, then painting the target around the holes.
Examples
The name comes from an old joke: a marksman fires a scatter of shots at the side of a barn, then walks up and paints a bullseye around the spot where the most holes landed. Presto, a sharpshooter. The same move happens with real data.
A reader compares their day to their horoscope.
Horoscope: “A conversation today brings unexpected news.” Priya: “My coworker did mention a surprise meeting! This is so accurate.”
The horoscope also said “a financial decision requires patience,” which didn’t happen, and Priya doesn’t count that. Out of a dozen vague predictions, she noticed the one that loosely fit and ignored the rest.
Companies do the same with performance data:
Post: “After our redesign, checkout completion is up 12%! The new design works.” Reply: “Wasn’t average order value down 8% and support tickets up 15% that same month?”
Out of a dashboard full of metrics, one moved in a flattering direction. That metric got the announcement; the others didn’t.
Why the reasoning fails
Large enough datasets always contain some cluster, streak, or coincidence purely by chance — that’s what randomness looks like at scale. The Texas sharpshooter fallacy draws the “target,” the claim of significance, after already seeing where the data landed, instead of predicting the pattern in advance and testing it. Because the target is painted around the hits, it’s guaranteed to match — and that guarantee is exactly what makes it worthless as evidence. A real finding has to be specified before looking at the data, or confirmed on a fresh, separate sample.
How to respond
- Ask what was predicted beforehand: “Was this pattern expected before you saw the results, or spotted afterward?”
- Ask about the whole dataset: “What did the other metrics do? Are we only hearing about the one that improved?”
- Ask for replication: “Does this pattern hold up on a different sample, or just this one?”
- Don’t dismiss every after-the-fact observation. Noticing a pattern and then designing a proper test to check it is exactly how real discoveries start — the fallacy is stopping at the noticing.