[Development Log] An EA that ignores the “laws of nature” of the market will break. The reality of liquidity and volatility
In the development of systematic trading, a system that shows an extremely stable profit factor in backtests frequently causes sudden and devastating drawdowns during specific late-night or early-morning time bands in actual operation.
Many developers treat this as “mere noise” or the manifestation of unexpected tail risk and focus on superficial fine-tuning of parameters. However, from the perspective of a process design engineer, this breakdown is by no means incidental. It is a inevitability caused by a structural architecture that does not incorporate the market’s inherent “variations in liquidity and volatility” — the natural law of the market — into the system internally.
This article dissects the fragility of a logic that presumes constant expectations while ignoring the real market’s reality, and logically proposes verification processes to measure its adaptability to real environments.
1. The “Natural Law” of Liquidity Depletion
Market infrastructure does not operate in a 24-hour uniform state. Liquidity (liquidity) and volatility (price fluctuation) are constantly and dynamically changing depending on the number of market participants and the amount of trading.
During late night to early morning rollover when major markets close and the number of market participants declines sharply, market liquidity becomes extremely depleted. In this environment, even small orders can cause prices to spike (emergence of sudden volatility), and the quotes from liquidity providers (LPs) become unstable, causing spreads to widen several times over the normal level.
This is the market’s “natural law.” However, in typical backtesting environments, the realistic frictions such as spread widening and slippage due to liquidity depletion are often not accurately reproduced. The system is optimized under the fictitious premise that liquidity and spreads are always constant in tests.
2. Structural Defects Leading to Breakdowns and “Asymmetry of Risk”
When a logic that assumes a certain expectation is tossed into the actual market where liquidity is depleted, the architecture’s vulnerability is exposed to the utmost. In particular, if this vulnerability is tied to drawdown-resistant design philosophies like averaging down or martingale, the outcome becomes catastrophic.
Below are the structural differences between a drawdown-type system that ignores natural law and a process-design, environment-adaptive (single-position) system based on process design.
| Comparison Item | A drawdown-type architecture assuming a constant environment | A process-design type assuming real-world environment adaptation |
| Behavior under liquidity decline | Believes in the rebound as in testing, and forcibly enters recklessly (averaging down) in the worst environment | Detects frictions such as spread widening and structurally refrains from entering |
| Response to volatility | Continues to carry unrealized losses against sudden price spikes (noise), infinitely expanding risk of ruin | At the moment advantage collapses, exits with a small loss via a stop-loss, avoiding a fatal blow |
| Assumptions | The market is always constant and moves according to its own logic — a theoretical construct | The market is uncertain and liquidity fluctuates — the natural law |
| Survivability | Very low (collapse to total loss in liquidity-depletion shocks) | Very high (tolerates and avoids real-world frictions, prioritizes physical survivability) |
In a market with no liquidity, enduring unrealized losses amounts to suicide. A system that doubles positions in a space without rebound energy is a flawed architecture that defers ruin and simply harbors a time-bomb of enormous leverage.
3. Process Design Engineer-Defined “Real-World Adaptation”
I am not a programmer. I do not value the aesthetic beauty of a programming language’s syntax or flashy reports that only exist in test spaces. What I pursue as a process design engineer is robustness (reliability) — whether the system can physically survive a harsh and uncertain market environment.
A system that truly adapts to real markets does not resist the market’s natural law.
Eliminate drawdown-resisting design from the system architecture and hold only a single position. At entry, mandate a clear exit line (SL) at the system level. If unforeseen noise from liquidity depletion occurs, promptly execute the exit line to limit risk to the present.
Furthermore, as a third verification axis to extract edge, optimize operating hours. In times when spreads widen and liquidity falls, structurally refuse entries with a filter, and run the system only in environments where the real liquidity that allows your logic to function properly is guaranteed.
Conclusion: Eliminate theoretical guidance and guarantee absolute defense
A system that assumes a fictitious constant expectation and ignores the natural law of markets is a hollow shell with no defensive power in real-world infrastructure.
The true aim of system trading is not to compete with numbers in a virtual space. In the real market, which changes mercilessly in real time and where liquidity ebbs and flows, the goal is to protect your capital thoroughly and continuously reap the edge over the long term.
We at Semura Lab. do not advocate running a black-box system with weak logical foundations. We constantly audit the honesty of trading logic and continue to prioritize process design to survive in harsh markets.
To conquer unknown territories, we must redesign the system from its core into a robust structure that excludes optimistic test expectations, enforces single-position and strict exit management, and assumes adaptation to the environment.