A common read on a chart goes like this. When a 4h candle closes strongly green near its high, and MACD has not crossed yet or has only just crossed, the next few hours should hold near that top or give back part of the move to familiar levels. We wanted to know which part of that read the data supports.
The test
We called a candle strong when its body was at least one ATR (14 periods on the 4h) and it closed in the top 20% of its range. MACD used the standard 12, 26 and 9 settings on the 4h. We kept the cases where MACD had not crossed yet or had only just crossed, meaning the histogram was still at or below zero, or had turned positive on that candle or the one before.
We measured what followed on the 1h candles that start at the 4h close: where price closed one, two and three hours later, how far it pulled back into the 4h candle, and whether it made a new high above it. The data is Kraken BTC/USD in two windows, 2018 to 2023 and January 2024 to June 2026.
| After a strong green 4h close | 2018–2023 | 2024–Jun 2026 |
|---|---|---|
| Higher after 3 hours | 48% | 47% |
| New high above the 4h high within 3 hours | 74% | 76% |
| Pulled back at least 25% of the candle | 36% | 51% |
| Pulled back at least 50% of the candle | 12% | 21% |
The shape holds
The typical path pokes above the high and chops there, and deep pullbacks are uncommon. Part of the new-high rate is mechanical, because the close already sits near the high. Still, price stayed in the top half of the candle over the next three hours in 88% of cases in the first window and 79% in the second.
The direction does not
Three hours later, price was higher about as often as it was lower: 48% and 47% of the time. The strong candle says very little about which way the next hours end.
MACD timing made little difference. In the recent window, candles where MACD had only just crossed pulled back at least a quarter of the candle in 51% of cases, against 41% where it had already crossed. With 135 cases, that gap is within the noise.
What it is good for
The parent candle describes the shape of the next hours: where price is likely to chop and how far it tends to pull back. It does not describe the destination. A band for the next one or two candles can use the parent candle to place its range, but it should not lean its centre upward just because the candle was green.
Narrowing a band to make it look directional does not add information either. It only makes the band wrong more often. If a forecast is going to lean one way, the lean has to come from conditions that have shown an edge on data the model never saw.