What it is
A triangle is a pattern in which price narrows between two converging lines until it breaks one of them. There are three:
- Ascending: the highs stay at the same height (a flat resistance) and the lows rise. The textbook expects an upward breakout.
- Descending: the lows stay at the same height (a flat support) and the highs fall. The textbook expects a downward breakout.
- Symmetrical: the highs fall and the lows rise. The textbook gives no side: you wait for the breakout.
For all three, the textbook reading after the breakout is the same: price should carry on in the direction it broke.
How we detect it
With fixed rules and without looking at the future, the same as the app's tool:
- Swings: the last 2 to 4 swing highs and lows of the previous 60 hours (each the extreme of the 3 candles on either side), with at least 2 of each.
- Lines: the best-fitting line through the highs and the one through the lows. No candle in the pattern closes outside them.
- Shape: a line is flat if it moves less than one average volatility (1 ATR) over the whole pattern. The lines must converge: at the end the pattern is at most 60% as wide as at the start. At least 15 hours.
- Breakout: the first candle that closes outside one of the two lines. That is where you decide.
Then the usual measure: two bands at ±1.5 ATR from the breakout close and which one price touches first within 48 hours, compared with chance in the same data.
Which side did it break?
| Triangle | Cases | Broke on the textbook side |
|---|---|---|
| Ascending (textbook: up) | 258 | 32.6% ±5.7 |
| Descending (textbook: down) | 234 | 37.6% ±6.2 |
It is the clearest result of the whole study: two in three ascending triangles broke downwards, and nearly two in three descending ones, upwards. We also checked it against the trend before the pattern, and that does not explain it (they broke with that trend half the time).
A likely explanation is geometric: in an ascending triangle, the line that moves is the lower one, and it is the one that catches up with price while the upper one stays put. The pattern drawn with the idea of "buying pressure" breaks, in practice, more often through the rising line.
And after the breakout?
| Breakout | Cases | Followed through | Chance |
|---|---|---|---|
| Ascending | 258 | 52.5% ±6.1 | 50.1% |
| Descending | 234 | 51.9% ±6.4 | 49.9% |
| Symmetrical | 308 | 52.8% ±5.6 | 50.0% |
| All | 800 | 52.4% ±3.5 | 50.0% |
| Ascending and descending that broke on the textbook side | 172 | 55.8% ±7.4 | 50.0% |
| Ascending and descending that broke the other way | 320 | 50.3% ±5.5 | 50.0% |
How to read it:
- After the breakout, price followed through slightly more often than chance (52.4%), within the margin: a triangle breakout is not a guarantee, it is a small clue.
- When it broke on the textbook side, it followed through more (55.8%), but that happened in only a third of the cases and you cannot know in advance.
- What you train: not jumping the gun. Entering before the breakout because "ascending triangles break upwards" would have gone wrong two times in three.
By coin and by year
| BTC | ETH | SOL | |
|---|---|---|---|
| Followed through | 48.3% ±5.5 | 52.2% ±5.9 | 59.2% ±6.7 |
| Year | 2021 | 2022 | 2023 | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|---|
| Followed through | 50.4% | 59.0% | 53.9% | 50.0% | 45.1% | 53.4% |
Each year on its own has few cases (102–181) and a margin of ±7 to ±10 points. As with flags, SOL is the one that departs most from chance.