Transitioning from Blueprint to Board : Seeking Structural Integrity for a Multi-Tier Trend Following System

New comment
 

Hello everyone 😊 ,

I am developing my first EA, which is built for myself. My background is an architect, so I designed it using my experience as an architect + trader + my requirements to create this EA. However, I have no background as a developer and still have a lot to learn. I am here to share my ideas, ask for advice, and gather thoughts to find blind spots that I might have overlooked. It is like bringing my pre-final thesis blueprint to open for the committee (community) to critique and check the design before the final version, in order to move into the forward test and live processes respectively.

(Note: I am a left-handed visual thinker, So I am not good at using professional terms and have issues with them. So I choose to use terms that I can understand myself. I hope everyone understands.)


Main Architecture

The EA (House893) is designed as a multi-engine EA working together. Each engine has different duties but shares the Shared State System and risk control. Mainly, it uses a structure-based concept, which is a structure I designed myself (can automatically update according to market conditions) + price action (No Indicators).

Multi-Engine System : The main system is divided into 3 engines:

  1. Core Engine 4H (Shift = 1) : The main engine responsible for finding the main direction and entering main positions to control the cycle. It works well when a trend occurs, loses frequently during sideways markets, and causes high drawdown (DD) because it holds trends without trailing SL.

  2. Adaptive Engine 1H (Shift = 1): An auxiliary engine to collect short-term profits. It works to solve problems when the market goes sideways, including entering trades following Core Engine 4H to balance the DD of Core Engine 4H. However, it can enter independently when there are no Core Engine 4H orders, and execute counter entries when special conditions are met.

  3. Storm Engine (Shift = 1): A special engine that is currently disabled. It will be automatically activated when the market has high volatility, causing Core Engine 4H and Adaptive Engine 1H to stop working, and will hand off back to Core Engine 4H when the market returns to normal conditions.

(Note: This Multi-Engine framework also serves as the main blueprint for my Manual Hybrid Mode, allowing me to manually execute visual scalping entries alongside the EA seamlessly without messy indicators. And I might add Scalping Mode in the next version).


SL / Trailing SL / Reverse Exit

The system uses SL to cut losses, setting the EA to choose the best values including:

  • SL at the structure line

  • SL at the candlestick wick end + buffer %

  • SL capped according to risk %

It mainly uses Trailing Stop and Reverse Exit to close profits. There is no TP setting (TP is set only in Storm mode). I designed the DD Control layer as a risk management mode to control TP placement, consisting of soft DD control, medium DD escalation, Circuit Breaker → freeze mode, and Emergency Exit.


DD Control Layer Concept

EA (House893) is designed to have DD Control as one of the main layers to control each engine's behavior. I have prepared some structure and code to support it, but it is not activated yet because I expect many unseen edge cases and bugs. Therefore, the DD Control is still turned off (false) so that I can inspect and develop other parts of the system for stability first, then test the DD Layer step by step.

The concept of DD Control is currently just a concept and architecture. It is not considered a finished system yet, and I might not succeed in making it. I welcome critiques regarding the architecture and risk assessment on whether this concept has any weaknesses, blind spots, or risks that should be considered before actual use.

  1. Soft DD Control : Located inside the Medium DD escalation, it always chooses to close and set TP for the first 1H order first (maximum of 3 pending orders).

    • SOFT → Lock Profit in a staircase pattern according to 1, 3, 5, 7, 9, 11, 13, 15 principles continuously :

      • Profit 5% (5x1) —> Lock Profit 0%

      • Profit 15% (5x3) —> Lock Profit 10%

      • Profit 25% (5x5) —> Lock Profit 20%

      • Profit 35% (5x7) —> Lock Profit 30%

      • Profit 45% (5x9) —> Lock Profit 40%

      • Profit 55% (5x11) —> Lock Profit 50%

      • ...and so on.

    • It resets values when Medium shifts to a new Tier.

  2. Medium DD Escalation : Chooses to close and set TP for the first 4H order first (maximum of 5 pending orders).

    • Medium → Lock DD according to capital in each Tier :

      • Tier 1: Capital <= $50,000 – DD max 20%, Flexible 1% (Soft Control runs continuously inside Tier 1)

      • Tier 2: Capital >= $52,500 ($50,000 + Holding 5%) – DD max 15%, Flexible 1% (Soft Control runs continuously inside Tier 2)

      • Tier 3: Capital >= $105,000 ($100,000 + Holding 5%) – DD max 10%, Flexible 1% (Soft Control runs continuously inside Tier 3)

      • Tier 4: Capital >= $157,500 ($150,000 + Holding 5%) – DD max 5%, Flexible 1% (Soft Control runs continuously inside Tier 4)

  3. Circuit Breaker : Forbids opening all new orders, but allows choosing to place TP for existing pending orders to return to Medium Control.

    • Tier 1 —> DD max 35%

    • Tier 2 —> DD max 30%

    • Tier 3 —> DD max 25%

    • Tier 4 —> DD max 20%

  4. Emergency Exit :

    • Every Tier → DD Max 100% → If this event occurs, the user will decide to open/close the system to enter the DD Control Layer again.


Filter Layer

  • Entry / Market Filter : Daily Range Filter, Spread Filter, Slippage Filter, Direction Filter

  • Entry Validation Filter : Zone Occupied Filter, Zone Lock Filter, Max Open Orders Filter, Allow Only One Buy / Sell Filter

  • Candle / Structure Filter : Body Filter, Structure Validation Filter

  • Risk Filter : Entry Capacity Filter, Drawdown Control Filter

  • Safety Filter : Execution Validation Filter, Freeze Mode Filter, Global Pause Filter

  • Counter / Engine Filter : H4 Opened This Candle Filter, Counter Direction Filter, Engine Activation Filter


System Utilities & Infrastructure

  • AutoSave / State Persistence

  • Semi-Modular Architecture

  • Execution & Safety Layer

  • Error Handling System

  • Zone Memory & Lock System

  • Trailing Lifecycle System

  • Engine Synchronization Logic

  • Comment / Tracking System

  • Monitoring & Operations


I have attached the backtest results, entry overview, initial SL, trailing SL, and reverse exit below. The lot size is fixed at 0.01 (Spread 50), closing profits via trailing and reverse exit, with no TP placement (TP only in Storm mode). Position sizing is not applied yet, and the DD Control layer mode is turned off because I expect many unseen edge cases and bugs.

I have also attached the Candlestick and Price Behavior summary that I made based on my understanding, which serves as an important condition for entries to answer any potential questions regarding overfitting.


What I need:

I want perspectives from people experienced in EA development to critique this architecture from your point of view, especially regarding :

  1. Potential weaknesses or technical debt.

  2. Edge cases or worst-case scenarios that could cause the system to lose real money.

  3. Risks of using Multi-Engine + Shared State.

  4. Risks of State Persistence / Recovery.

  5. Conceptual weaknesses in the DD Control Layer concept.

  6. If you were to continue developing this, which part would you worry about the most and why?


My Goal : I want to design and develop House893 to be as strong as a building and able to survive in diverse market conditions, even though I still have a lot to learn.


Thank you for your help 😊
 
For Sideway 😊
New comment