SMT liquidity divergence

By the AKQX QuantumTrade team · Educational content, not investment advice. This describes a strategy family as a set of rules. It is not a recommendation, makes no claim that it is profitable, and its example numbers are made up.

What is the idea?

Markets that usually move together tend to make their highs and lows together. When one sets a new high and the other does not, they have diverged. The strategy reads that disagreement as a sign that the move lacks broad participation, or that liquidity above the old high has been taken rather than a trend extending.

Everything depends on definitions. Which two markets, how long the correlation has to have held, how a swing is identified, and what you do once the divergence appears are all choices, and different choices produce different strategies under the same name. Two traders can look at the same day and see different divergences.

That is why the idea belongs in a backtest and not on a chart. A rule that finds the divergence mechanically, without anyone looking at the outcome first, can be counted. A divergence spotted by eye can always be justified afterwards.

The rules, in order

1. Choose a pair of correlated markets and a timeframe, for example two large crypto pairs on a 4-hour chart.
2. Identify swing highs and lows on both with one fixed definition, such as a number of candles on each side.
3. Flag a divergence when one market takes the prior swing high or low and the other does not.
4. Wait for a confirmation step before acting, so the divergence alone is not the trigger.
5. Enter on a faster timeframe in the direction the divergence implies, with a stop beyond the swing and a defined target.
6. Cancel the setup if it is not triggered within a set number of candles.

The settings the rules hide

Every setting below is a choice, and each one changes which trades the strategy takes. They have to be fixed before a test, and the more of them there are, the more room there is to fit the past.

What AKQX provides for this

AKQX detects SMT divergence as a swing-style comparison across a configurable pair, and lets you choose whether both markets are traded independently or only one, for example the one that was swept or the one that was not.

Ready starting strategies exist for crypto pairs: wide-anchor-zone variants driven by a single SMT or by an SMT plus a minor-SMT confirmation, across timeframes from 3 hours to 1 day with 5-minute entries, and variants that keep the higher-timeframe reference current as the session develops.

There are also research variants with an extra confirmation stage in the entry chain. These are starting points for your own testing. They are not recommendations and carry no performance claim.

How to backtest it honestly

1. Fix every definition in the table above before you run anything, and write them down.
2. Check the correlation of your pair over the test period instead of assuming it. A pair that decoupled for six months in the middle of the test is two different experiments.
3. Count divergences first, then outcomes. If there are only a few dozen over years, the sample is too small to conclude anything, whatever it shows.
4. Run with realistic fees and slippage. Entries on a fast timeframe with a stop behind a swing can have tight stops where costs matter.
5. Nudge the swing length and the confirmation, and check the result degrades gradually. Then test on a period you did not tune on.
6. Paper trade it. Divergence setups that need two feeds to agree are exactly where live data timing differs from history.

The full method is in how to backtest a crypto strategy, and the way results are scored is on the methodology page.

A worked example, with made-up numbers

Illustrative only: on a 4-hour chart, market A makes a higher high at 5,010 while correlated market B stops at a lower high of 4,990. Under a rule using a 5-candle swing, that is a divergence. The rule then waits for a 5-minute candle to close back through a level before entering, and cancels the setup after 6 candles. Run across a long history, the count of such events, not any single one, is what the test reports.

Where it goes wrong

• Hindsight. Every past reversal has a divergence in retrospect. Only a rule that fires before the outcome counts.
• A pair that is not really correlated, so the divergence is noise.
• Too many settings. Pair, swing, timeframe, confirmation, entry and expiry give plenty of room to fit the past.
• Trading both markets as if they were independent trades when they are one idea twice.

Questions about smt liquidity divergence

Keep reading