Cherry-picking is choosing which facts to present, or which window of time to measure them over, specifically because that choice supports a conclusion you already wanted to reach, while an equally true but less convenient window or fact gets left out. It rarely requires stating anything false. A game’s player count really can be up 40% “since launch week” while also being down 60% from its peak the month before, both numbers accurate, both describing the same series, and a reader given only one of them walks away with close to the opposite impression of how the game is actually doing.
Financial disclosure has a documented version of this exact move, and the SEC’s own guidance on non-GAAP financial measures names it directly by describing what isn’t allowed. Adjusting a reported figure only in one direction, “adjusting only for charges while excluding similar gains from the same period,” is a violation the SEC’s guidance calls out by name, not a borderline judgment call. A company that strips out a one-time cost to make a quarter look better, while leaving in a one-time gain of similar size from the same quarter that would have made the comparison look worse, is applying the adjustment selectively rather than consistently, and the guidance treats that asymmetry as the tell. The same document flags giving a non-GAAP figure “undue prominence” over the comparable GAAP measure, presenting the friendlier number first, in a larger font, or in a headline, while the less flattering official number appears later or not at all, as a separate but related problem: not cherry-picking the number itself, but cherry-picking which number the reader sees first.
Time windows are cherry-picked the same way numbers are, by choosing the start and end points that produce the story you want rather than the ones a neutral observer would pick. That’s exactly why formal definitions of market conditions specify a minimum window rather than leaving it open. Investor.gov’s own glossary defines a bear market as a broad index falling “20% or more over at least a two-month period,” and the two-month floor is doing real work in that definition: without it, any single bad trading day inside a longer uptrend could be relabeled a bear market by simply measuring from that day’s open to its close. Requiring a minimum duration before a label applies is a structural defense against cherry-picking the shortest window that produces the scariest, or the most flattering, number.
The corrective isn’t refusing to look at any short window at all; a single strong week is real information, not fabricated. It’s asking, deliberately, what the same number looks like over a longer window, or against a different starting point, before treating one window’s story as the whole story. A chart showing a game’s player count over its single best week is true. So is the same chart widened to a full year, or started from the game’s launch instead of its peak. Reading a claim built on a chart means asking which window was chosen and why, checking whether a different, equally defensible window changes the conclusion, and treating a claim that only survives one specific window with the same skepticism the SEC’s own guidance applies to a company that only shows investors the version of its numbers that got adjusted in one direction.