PatternKilled

Buy the dip, counted: thirteen thousand opportunities, priced

+65% vs +166%

the best buy-the-dip rule on the best asset, against simply holding that same asset over the same months

In plain English

Everyone believes they could just buy the dips and take the small bounces, and after seven months of watching charts it feels obviously true. So we counted every dip in seven months of minute-by-minute prices, thirteen thousand of them, and made the computer trade them all under fixed rules. About half bounced. The eye remembers those; the account also trades the other half, which kept falling. And on the one coin where dip-buying did make money, buying at completely random moments made money too, because the coin itself was rising: the dip-spotting added almost nothing, and the whole clever exercise earned less than half of what doing nothing would have.

What was claimed

Just buy the dips and take the small wins, a human can see them all over the chart.

The bar it had to clear

dip entries beat random entries under identical exits, by enough to cover round-trip costs

What the data said

a +0.14% per-trade timing edge at best, negative on ETH, and the winning configuration still made less than half of holding

Why it's dead

The operator asked the fairest version of the chart-reading question: with seven months of minute data, how many times did 'that's down, buy the dip, take the bounce' actually occur? So we counted, with the two guards hindsight never applies. First, count every dip the trigger fires on, not just the ones that bounced. Second, run the same exits from random entry times, because in an asset that tripled, any entry rule 'works' on drift alone. Dips occurred constantly, about half bounced, and the eye remembers those. The account also trades the other half. After both guards, the dip signal itself was worth a fraction of a basis point of skill on the one asset where it made money at all, and that asset paid two and a half times more for doing nothing.

The detail

Fixed mechanical rule, no tuning: buy when price sits D% below its rolling 24h high, exit at +T% or −T% (symmetric) or 24 hours, one position at a time, 6.5 bps round-trip cost, D in {2,3,5}%, T in {1,2,3}%. Roughly 13,000 episodes across BTC, ETH, SOL and HYPE on 1.1 million minute-bars, Feb 07 to Sep 01 2026. Hit rates 46-60% everywhere. HYPE, 3% dip chasing a 2% bounce: 545 episodes (84 a month), 54% bounced, +0.088% net per trade, +48% summed, against +165.6% for holding HYPE over the window; the best HYPE cell summed +65.5%. The matched random-entry control (same exits, matched trade count, 50 seeds) also made money on HYPE, so the dip trigger's own contribution was +0.06% to +0.14% per trade. On ETH the timing edge was consistently negative: dips fell further than random entry points at nearly every parameter. On BTC the only positive cells had 25-75 episodes, anecdote-sized. Exploratory census across 36 cells with no pre-registration, so the best cells are selection-biased and quoted only as a ceiling. Tool honesty: the first run reported zero dips on an asset that tripled, because this environment's pandas rolling-max silently returns not-a-number on clean data; a positive control caught it before a false 'no opportunities existed' shipped.

Kill date: 2026-09-01
Sample: ~13,000 dip episodes, 4 assets, 1.1M minute-bars, 36 rule variants, 50-seed random-entry control
Method: Documented kill
Verdict: the eye counts the half that bounced