Bitcoin: 76% win rate, +100% avg. Planning for +26%Bitcoin / U.S. dollarBITSTAMP:BTCUSDSpokoStocksA Bull Bear Power(13) zero-line cross fired on the Bitcoin weekly chart three bars ago. Tested across every prior occurrence on this chart, it won 76.5% of the time over the following 20 weeks, with an average return of +100.74%. Those two numbers are the reason I do not trade off win rates. ──────────────────────────────── THE NUMBER THAT IS ALWAYS MISSING Over an average 20-week window in this sample, Bitcoin returned +62.71% with no signal, no timing, and no analysis at all. That is the baseline. It is the bar any bullish signal has to clear before it has told you anything, because on an asset that rose this far, almost every entry looked brilliant. A win rate quoted without its baseline hasn't been measured. It's been decorated. Against that bar, this signal does clear it: Win rate 76.5% Average return +100.74% Baseline +62.71% Edge +38.03% t-stat 2.39 Occurrences 34 ──────────────────────────────── NOW THE PART THAT SHOULD LOWER YOUR EXPECTATIONS 1. THE AVERAGE IS A LIE OF COMPOSITION Average return +100.74% Median return +26.44% Std deviation 246% The average is nearly four times the median. A handful of these occurrences landed inside the 2017 and 2020 vertical phases and dragged the mean up with them. The typical occurrence returned about 26%. The average is what happened to a few signals. The median is what probably happens to you. I size for the median. 2. THE EDGE HAS AN EXPIRY DATE 5 weeks: +6.14% edge 10 weeks: +4.99% edge 20 weeks: +38.03% edge 40 weeks: -8.52% edge By 40 weeks the signal underperformed simply holding. Whatever this pattern captures appears finished inside roughly 20 weeks. Holding longer historically gave the advantage back. That is a holding period, not a suggestion. 3. YOU HAVE TO SURVIVE THE MIDDLE Average maximum favourable move +161.67% Average maximum adverse move -19.82% On average, price moved roughly 20% against the entry at some point before the horizon closed, including on the occurrences that ultimately worked. The reward-to-risk is attractive, but only for a position built to withstand a drawdown of that size. A tight stop would have cut out most of the winners. ──────────────────────────────── THE SIGNAL THAT RANKED FIRST IS NOT THE ONE I TRUST MOST The leaderboard on this chart ranks every built-in Technicals signal by edge over baseline, with a minimum of 30 occurrences. Two rows matter: Bull Bear Power(13) cross above 0 34 occurrences · +38.03% edge · t-stat 2.39 HMA(9) turns up 62 occurrences · +27.84% edge · t-stat 3.15 The first has the bigger edge. The second has nearly double the occurrences and a stronger t-stat. A large edge on a small sample and a moderate edge on a large sample are not the same discovery. The second is the more reliable finding, even though it looks less impressive. Ranking by headline number alone would have put them in the wrong order. That is the actual skill in this work: knowing which of two good-looking rows is the one that will still be good-looking next year. ──────────────────────────────── ONE ASYMMETRY WORTH KNOWING The same indicator crossing BELOW zero shows an edge of -10.8% over the same horizon across 36 occurrences. The bullish cross carries information. The bearish cross does not. That asymmetry appears repeatedly on assets in a strong secular uptrend, and it is why a symmetric long-short system built on one oscillator usually disappoints. ──────────────────────────────── HOW I AM ACTUALLY TREATING IT Bias: long, with a working horizon near 20 weeks rather than open ended. Sizing: based on the +26% median and an assumed ~20% adverse excursion. Invalidation: Bull Bear Power(13) crossing back below zero, which is the same condition that produced the signal, reversing. ──────────────────────────────── WHERE THIS ANALYSIS IS WEAK 34 occurrences is the entire history available on Bitcoin weekly. It is a small sample, and it covers one secular regime. Every long signal in that sample inherits the same upward drift, which is exactly why I read the edge column and not the raw return. The 20-week windows overlap, so the t-stat is optimistic. I treat 2.39 as a reason to look closer, not as proof. And this is one symbol on one timeframe. The same test on other markets reorders the leaderboard completely. Signal edges appear to be local to the market and period they were measured on, and they decay. ──────────────────────────────── RUN THIS ON YOUR OWN CHART The workflow is three steps and works on anything: 1. Pick the signal you actually use. 2. Find the baseline return of that instrument over the same horizon. 3. Subtract. If the difference is near zero, your win rate is measuring the market, not your signal. ──────────────────────────────── DISCLAIMER This is analysis of historical data, not investment advice. How a pattern behaved in the past does not determine how it will behave next. The sample here is small and drawn from a single market regime. I have no idea what Bitcoin will do over the coming weeks, and nobody posting a chart does. Manage your own risk and do your own work.