A long shelf, then a break upward

The idea: price sits in a well-worn range for months, and leaving it means a run to the next such range. When we tested it, breaking out of the shelf paid exactly as much as a random day on the same coin.

Why it sounded right

A shelf is a range a coin has traded inside for months. You can see it on a chart without any indicator: a dense zone where almost all the turnover happened. There is nothing to do inside it, price just walks back and forth. And this part is not imaginary: plot where turnover actually piled up over the last six months and most coins really do have a shelf, and it really is narrow.

From there the logic feels inevitable. While price is inside the shelf, everyone who bought and sold there is sitting above it. The moment a daily close goes above the upper edge, all that weight is below. There is nobody left to sell: higher up, history is thin, so there is no resistance, so price flies to the next dense zone. Runs like that are rare, but they are what draws the whole trend. Look at any big chart: a long sideways stretch, then one move, and the coin lives at a new level.

What made it even more convincing is how easily the rule can be written down. The shelf can be computed, not eyeballed. The break counts on a daily close, not on a wick. The target sits at the lower edge of the next dense zone — known in advance. The stop uses the coin's ordinary daily range. Nothing is fitted, everything has a cause. A rule like that is pleasant to test, because it feels like there is not much left to test.

How we tested it

We took large coins, daily candles over many years, and found every upward break by that definition: the coin sat inside a shelf for most of the past month, then closed above the edge with room to spare. Signals came out at roughly one per coin per year — so "these moves are rare" was confirmed immediately.

Then two control runs, and without both of them the result means nothing. First, a random entry on the same coin: pick a random day, use the same stop and the same target, count the result. That answers the question "what would this coin have paid anyway, with no signal at all?" Second, a correction for the market: subtract the typical move of everything else from the coin's move. Without that, large coins "win" simply because over the years we measured they beat the typical altcoin, and any rule applied to them would look brilliant.

How it ended

The rule as a whole is a zero. The raw result looks decent until you look at the random entry: it pays the same. On mid and small coins it is worse — there the typical trade lags the market, and the average is held up by one rare but huge case, which makes it a lottery, not a rule. A bull market does not save it: the rule makes money there, but a random day in the same market makes just as much. Market conditions describe the market, they are not a signal. Breaks to the downside gave nothing in any conditions.

The story about "a months-long move from shelf to shelf" fell apart as well. The typical trade resolved within days: either it gets there right away or it never gets there. And half the trades first go against the entry by roughly as much as the whole intended move — meaning a tight stop is fundamentally incompatible with this rule.

One narrow slice survived: a large coin breaking into territory with no history at all — new highs, where the next shelf simply does not exist. There the result clearly beats a random entry. But we looked at about ten slices, one is alive, and it holds few trades. That is what a finding produced by digging through slices looks like, not a result. It is a candidate for a separate test, not a rule.

What this teaches

• An observation can be true and the trade still empty. "These moves are rare and strong" — true. "You can catch them on the break out of a shelf" — no.
• A random entry on the same coin is mandatory. Without it, any rule looks like it works in a rising market.
• Compare against the market over the same window. Otherwise you are not measuring the rule, you are measuring the fact that large coins went up in those years.
• If one slice out of ten works, that is most likely digging, not a finding. Test it again on other data instead of building it into a strategy.
• Test the rule before you take the trade, not after. Then a "six-month move" that actually resolves in five days shows up for free.