How we test a setup

A setup is worth an alert only if it does better than buying at a random time. Here is how that's checked, in four steps.

1. The rule comes first

Each setup is written down as one exact rule before it's tested: when the alert comes, the line it's wrong under (a 15-minute close under a price, not a touch) and its target. Nothing is adjusted on the year it's tested on.

2. Random timing is the bar

For every setup found, 30 random moments of the same coin are picked at the same hour of day, on equally calm or wild days, in the same month, with the target and the stop at the same distances. A setup beats random timing when it reaches its target before its stop more often than those random buys do. Most dips don't: "down 2%" on its own does no better than random.

3. A year of data, and a fresh window

• A year of Binance 15-minute candles for BTC, ETH and ZEC (8/2025 to 9/2026), with the rules fixed. These are the records the pages quote.
• 90 days of Hyperliquid 15-minute candles for BTC, ETH, HYPE, ZEC and SPCX (7/25 to 10/15/2026), where the rules were built, with a last 30 days left untouched to check them.
• An independent rewrite of one test, from the written rules alone, found the same setups and the same edge.

4. Recorded live

Every setup since 10/16, where the test data ends, is recorded the moment its candle closes and followed for 24 hours: it reached the target, closed under the line, or neither. Every one.

What the results mean

• Proven: it beat random timing over the year on that coin. (On BTC the flush and reclaim is proven with a smaller edge.)
• Still being tested: no clear edge yet, or no full year of data (HYPE and SPCX). Recorded live; you can still pick it, and its alert says it's still being tested.
• Not offered: it did worse than random over the year on that coin.

What it isn't

Beating random timing a few more times in 100 is a real edge but a small one: it helps with timing, not with size. A strong trend swamps it either way. Nothing here is advice for anyone's own positions.