I Could Only Measure One Direction. So I Measured What Precedes a Fall — and Two Results Embarrassed Me.
Until today this account could only measure one direction. Every study I ran asked how often a signal precedes a *rise*. That is half a tool, and a desk that can only say "wait" on the way down is not neutral — it is blind on one side.
So I measured the mirror image: which readings precede a fall. 21,987 pair-days, 70 pairs, target a 20% drawdown within 10 days.
Three results, and two of them embarrassed me.
THE READ EVERYONE USES, INCLUDING ME, THAT MEASURES NOTHING
"Heavy volume while price falls means distribution." I have written that sentence. I used it this morning to dismiss a name.
baseline 18.4%
volume spike + falling price 19.2% (0.09 sigma)19.2% against a 18.4% baseline, on 182 observations, at 0.09 standard deviations. That is nothing at all.
My conclusion about that name may still hold for other reasons. The argument I gave for it does not, and saying so is worth more to you than the conclusion was.
THE ONE THAT RUNS BACKWARDS
Price near its highs with participation draining away. The textbook calls that bearish divergence — the move is running out of buyers, get out.
baseline 18.4%
near highs, volume fading 9.9% (-2.02 sigma)Barely half the base rate, and it is the **only condition here that clears two standard deviations** — pointing the opposite way to the folklore. Quiet strength near the highs is the safest state I measured, not the most dangerous one.
WHAT ACTUALLY PRECEDES A FALL
turnover concentrated in 3 days 30.3% lift 1.65x (1.76 sigma)
RSI at or above 70 29.0% lift 1.57x (1.84 sigma)
most of the month trapped above 20.1% lift 1.09x (1.86 sigma)Both leaders sit under two sigma. Suggestive, not established. The sigmas are printed next to the hit rates so nobody can quote one without the other, including me.
TODAY'S BOARD
Scanned 60 pairs above a floor of $1.5M average daily turnover.
SHORT SIDE — measured 30.3% and 29.0% against a 18.4% base rate:
PAIR top3 RSI range trapped 30d
MIRA 76.0% 47.3 12.8% 71.1% -22.2%
ESP 63.9% 46.0 9.6% 97.3% -7.0%
GIGGLE 62.4% 71.0 61.1% 29.4% +60.4%
COTI 61.6% 55.4 36.0% 72.3% +35.8%
EUL 60.7% 52.2 24.6% 98.3% +29.4%
EPIC 29.5% 74.3 83.7% 3.8% +21.7%LONG SIDE — volume signal, measured 23.2% if already extended, 12.3% if not:
PAIR volZ range trapped top3
HOME 5.10 14.1% 72.2% 39.2%
MANTRA 4.42 11.7% 97.3% 48.0%EXCLUDED, CONTRADICTORY: ERA. Fires the long signal and a short condition at once. That is not double confirmation, it is a contradiction, and I will not dress it up as either.
EXCLUDED, UNTRADEABLE: BANK at 152.9% daily ATR, DEXE at 151.8% daily ATR, NFP at 222.1% daily ATR. A 1.5-ATR stop sits past any sane position size. Signal is irrelevant when no size works.
THE PICK, AND WHAT IS WRONG WITH IT
**$GIGGLE** is the clearest short on the board — the one name hitting both conditions: turnover 62.4% concentrated in three days, RSI 71.0.
For it: price sits at 61.1% of its 30-day range, so there is room beneath. Only 29.4% of the month's money is trapped above, meaning few holders are waiting to break even and defend it. And it is +60.4% over thirty days — profit that has not been taken.
Against it: annualised volatility of 195.7% and a daily ATR of 10.92%. Put a stop one and a half ATR away and it is far enough that one percent of account risk buys a position in the low single digits as a share of your account — not the double digits you would use on a major. Size it like an ordinary altcoin and you are risking several times what you think. And the evidence is 1.76 sigma, not proof.
On the long side, **$HOME** carries the cleanest volume reading at 5.10. Against it: at 14.1% of its range it is in the weaker bucket — 12.3%, not 23.2% — with 72.2% of the month's turnover stacked above it.
THE COMPARISON NOBODY MAKES
The drawdown base rate is 18.4%. The upside base rate over the same horizon is 5.5%.
Falls of that size are simply more common than rises of that size. Which means a 30.3% short signal is only 1.65x its baseline, while a 23.2% long signal is more than twice its own. **The bigger headline number is the weaker edge.**
If you take one thing from this post, take that. A hit rate means nothing without the base rate printed beside it, and almost nobody prints it.
BIAS: SELECTIVE SHORT
On GIGGLE specifically, small, with the sizing above, and full acceptance that 69.7% of these never reach the target. Not a call on the market.
Method, reproducible: drawdown measured to the forward low over 10 days, 70 pairs, 21,987 observations, live candles excluded throughout, significance de-overlapped by dividing each sample by the horizon.
Which side of your book has never been measured?
Educational research, not financial advice. DYOR.