The MA5/10/30/60 bundle: convergence, expansion and crosses

The observation

Traders who watch the 5, 10, 30 and 60-period moving averages together know the picture: for a while the four lines fan out and trend, then they drift toward each other until they are almost one line. Shortly after that the market makes a decisive move and the lines fan out again. In the Chinese trading literature this is called adhesion then divergence of the averages; in the West the same thing goes by moving-average compression, the ribbon squeeze or Bill Williams' "sleeping alligator". The mechanism is not magic: four averages of different lengths can only sit on top of each other when price has gone sideways for long enough that the short-term and the medium-term mean agree. Sideways markets do not last; when the range resolves, every average turns the same way at once, and traders who use any of them as a trigger act together. That is why the move out of a bundle is often sharp.

The crosses are the second half of the picture. When MA5 crosses MA10 while MA30 is already above MA60 (or below for a short), the short-term impulse is in the direction of the medium-term trend, which is the classic continuation trigger. A cross against the slow pair is weaker and often a pullback.

What the engine measures

For every timeframe the feature engine now computes:

• bundle width: the distance between the highest and the lowest of MA5, MA10, MA30 and MA60, in ATR units;
• bundle percentile: that width ranked against the last 200 bars. Below the 15th percentile the four lines count as converged (the bundle flag on the technicals snapshot);
• expansion: if the bundle was converged within the last 12 bars, the width is now growing and price closes above all four averages with MA5 above MA10, the vote is +1 (long). The mirror image below all four with MA5 below MA10 votes -1 (short). Anything else votes 0: a converged bundle on its own has no direction yet;
• MA5/MA10 cross: a cross in the last 3 bars votes ±1 when MA30 and MA60 point the same way, ±0.5 against them.

These two techniques ("MA5/10/30/60 bundle breakout" and "MA5/MA10 cross with MA30/60 trend") join the vote vector on all nine timeframes. The existing moving-average ribbon technique (the order of MA5 > MA10 > MA30 > MA60 > MA200, the bullish or bearish alignment) was already there.

Is it real on your pair?

Nobody should take a rule on faith, including this one. The deep analysis replays the whole history of the pair bar by bar, records every bundle breakout and every cross, and measures with the triple-barrier label whether the move paid before the stop was hit, net of fees and funding. From there the combination search tests the bundle alone and together with other techniques (for example "4h bundle breakout + 1h order flow + funding not crowded"). Only combinations that beat the edge bar on the pair's own history are allowed to fire on the consensus card, and the card names them with their win rate and case count. The Techniques tab on the pair page shows the two new rows with their accuracy per horizon as soon as the next deep analysis has run (models retrain automatically because the technique list changed).

So the answer to "does the bundle predict the move" is per pair and per timeframe, and the tool gives it to you in numbers rather than as a belief.