Trading Overlapping Cycles and Interference

Volatility across a trading day isn’t one continuous, evenly distributed signal. It’s the sum of several distinct cycles running on their own schedules, layered on top of each other. The Asian session has its own volume and range profile. London has a different one. New York has another. None of these are random noise sitting on top of a flat baseline — each is a fairly repeatable daily cycle in its own right, and what you actually observe on a chart is the superposition of all of them happening at once. Treating the trading day as a single undifferentiated block of “the market” throws away most of the structure that’s actually useful, because the interesting behavior lives in how these cycles combine, not in any one of them alone.

Why “overlap” is the correct word, not just a scheduling term

The London/New York overlap, 13:00-16:00 UTC, gets treated in most trading content as simply “the busiest hours,” which is true but incomplete. It’s busiest specifically because it’s the window where two separate, largely independent sources of liquidity and participation are active simultaneously. London’s institutional flow hasn’t wound down for the day, and New York’s has just ramped up. These aren’t two halves of one cycle, they’re two distinct cycles whose active periods happen to intersect for three hours, and the volatility during that intersection isn’t simply additive in a boring way — it reflects genuine interaction between two separate pools of participants reacting to, and trading against, each other’s activity in real time.

This is worth taking literally as an interference concept rather than just a metaphor. In wave physics, constructive interference happens when two waves’ peaks align and produce an amplitude larger than either wave alone. Destructive interference happens when a peak in one aligns with a trough in the other, and they partially cancel. Session volatility behaves in a genuinely analogous way: the overlap window is close to a constructive case, where both sessions’ active periods reinforce each other into an amplitude neither session produces alone. But the transition zones — the tail end of the Asian session bleeding into the London open, for instance — can behave more like a partial cancellation, where Asian session participants are already thinning out before London volume has really built up, producing a comparatively quiet stretch that doesn’t fit either session’s “typical” profile cleanly.

What this means for a strategy validated on one cycle

A pattern validated using data drawn heavily from the London/New York overlap has effectively been validated against the constructive-interference case — the reinforcing combination of two active sessions. Applying that same pattern logic to the Asian session, where there’s only one cycle active and no second session reinforcing it, is not a like-for-like test even if you’re using the identical entry rule. The underlying volatility regime the pattern was shaped by simply isn’t present outside the overlap, and expecting the same win rate outside that window is expecting the pattern to perform in a wave condition it was never actually validated under.

This is a more specific version of why session-scoped configs matter. The start_hour and end_hour fields aren’t just a filter to avoid trading at inconvenient times, they’re an acknowledgment that the JSON config as a whole is only valid for the specific interference condition it was built and tested against. A single strategy config genuinely can’t be expected to perform consistently across a single-cycle session and a two-cycle overlap window without separate validation for each, because the two conditions aren’t variations on the same signal, they’re structurally different combinations of underlying cycles.

Reading the transition zones correctly

The quiet stretch between the end of the Asian session and the ramp-up of London activity is where the interference framing earns its keep the most, because it’s easy to misread as simply “low volatility” when it’s more precisely “a period where one cycle is fading out faster than the other is building in.” Spread tends to widen disproportionately during these transition windows relative to the actual realized volatility, because liquidity providers are themselves adjusting their own risk exposure across the same handoff, and that widened spread is a real cost that a backtest built primarily on overlap-window data is likely to underweight, since the overlap window rarely experiences that same handoff-driven spread behavior.

Window	Cycles active	Typical character
Asian session (00:00-08:00 UTC)	One	Lower amplitude, narrower range, thinner liquidity
Asian/London transition (~06:00-08:00 UTC)	Fading + building	Often the quietest stretch of the day, spread can widen disproportionately
London session (08:00-16:00 UTC)	One, then two	Amplitude builds through the session
London/NY overlap (13:00-16:00 UTC)	Two	Highest amplitude, constructive combination of both cycles

Why this isn’t the same as just tracking volatility

It’s tempting to collapse all of this into “just measure realized volatility per hour and size accordingly,” which captures some of the effect but misses the compositional part. Two hours with identical measured volatility can still represent very different underlying conditions if one is a single cycle running at unusually high amplitude for its own session, and the other is two cycles in ordinary superposition. Those two situations can look the same in a simple volatility measurement while implying very different things about how likely the current move is to continue, reverse, or simply reflect one session’s participants briefly overreacting before the other session’s flow steps in and corrects it. A pattern that depends on continuation is a different bet in each of those cases, even at matched volatility.

Keeping this from turning into an excuse for complexity

None of this is an argument for building an elaborate wave-decomposition model before you’re allowed to trade a session. The practical takeaway is much simpler: validate separately for genuinely distinct interference conditions rather than pooling all hours into one dataset and hoping the resulting average behavior transfers evenly across a day that isn’t actually uniform. A strategy meant for the overlap should be validated on overlap data. A strategy meant to catch the quieter single-cycle stretches needs its own validation, with its own expectation for win rate, spread cost, and healthy performance range, because it’s genuinely a different market condition, not a lower-volume version of the same one. The interference isn’t a curiosity about market physics. It’s the actual reason session-scoping a config was never an arbitrary convenience in the first place.