How to read a signal study
Every study page answers one question: after this signal fired on this coin, did the price behave differently from an ordinary week? Five numbers carry the answer.
A live example
Take Bitcoin after RSI falling below 30. It has happened 30 times with a full week of history after it. The price was higher 7 days later in 57% of those cases.
On its own that figure says little. The comparison is what matters.
Win rate against an ordinary week
Bitcoin was higher after any random 7 days 53% of the time. So the signal's edge is not 57%; it is the gap between the two, here 4 points. A coin that has risen for years makes almost every signal look good until you subtract its ordinary drift.
Median move
The median 7-day move after the signal was +1.7%, against +0.5% for an ordinary week. The median is used instead of the average because one extreme week can move an average a long way.
Could it be luck?
We draw 10,000 random sets of days of the same size and count how often they produce a gap at least this large. After correcting for the 3,565 studies tested together, the answer for this example is 79%. The lower that figure, the harder the result is to explain by chance.
Proof score and verdict
Sample size, the luck test and whether the pattern held in both halves of the history are combined into a proof score from 0 to 100. This example scores 19, which reads as "No better than chance". A direction is only called at 55 or more.
Fewer than about 20 cases is an anecdote, whatever the win rate.