SMC Bot Forex: How to Build a Smart Money Concepts EA That Follows Real Rules

Smart Money Concepts (SMC) and the ICT method are among the most popular ways traders read the forex market. They are also among the hardest to automate, because they are usually taught by eye. Two traders look at the same chart and mark different order blocks. A bot cannot work like that. To build an SMC bot or Expert Advisor, every concept must first become a precise rule.

Why SMC is worth automating

Some core SMC ideas match what researchers know about currency markets. Federal Reserve economist Carol Osler studied real stop-loss and take-profit orders at a dealing bank and found, in research published in 2003 and 2005, that stop orders cluster just beyond round numbers and can trigger self-feeding price cascades. That is the institutional reality behind what traders call liquidity sweeps. Trend structure, pullback entries and session timing are also well grounded.

Automation adds what discretionary SMC lacks: consistency, and the ability to test.

SMC concepts as bot rules

| Concept | Rule a bot can follow | |---|---| | Swing high / low | A candle whose high (low) exceeds the N candles on each side | | Uptrend | The last two swing highs and lows are both rising | | Break of structure | A close beyond the latest swing in the trend direction | | Change of character | A close beyond the latest swing against the trend | | Dealing range | Lowest low to highest high of the last N candles | | Premium / discount | Above / below the 50% level of the range | | OTE zone | 62% to 79% retracement of the range | | Bullish fair value gap | Candle 3's low is above candle 1's high | | Liquidity pool | Two or more swing highs (lows) within about 0.1 x ATR | | Sweep | Price trades beyond a pool and closes back inside |

Once written this way, an AI assistant can code each concept, and you can count how often each one actually leads to a profitable trade.

Top-down: foundation and structure

The most robust SMC bots work top-down. The monthly and weekly charts are the foundation: they decide whether only buys, only sells or nothing is allowed. The daily chart defines the area of interest. The 4-hour and 1-hour charts are the structure that times the entry, typically a sweep of nearby liquidity followed by a break of structure in the higher-timeframe direction inside the discount or premium zone.

A simple "bias panel" that prints BUYS ONLY, SELLS ONLY or STAND ASIDE from the weekly and monthly structure will stop more bad trades than any entry signal.

The traps specific to SMC bots

• Repainting swings. A swing high is only confirmed after the following candles close. An EA that treats it as known earlier is cheating in its backtest.
• Too many filters. Every extra confluence removes some losing trades from the past and makes the bot more overfitted to it.
• Order blocks without definitions. "The last opposite candle before a strong move" needs exact numbers: how strong, how many candles, which body or wick.
• Ignoring costs. SMC entries on low timeframes often have tight stops, so spreads are a large fraction of the risk.

We learned these lessons building the TebotechSignals engines: rules that looked excellent on six months of data were slightly negative over nearly four years. The full story, and the testing protocol we now use, is in The Hybrid Trader.

Add the safety engine

An SMC EA should never place an order without passing a risk module: fixed risk per trade, daily and maximum loss limits, exposure limits and a kill switch. This matters even more for traders taking prop firm challenges.

Build it step by step

Use an AI assistant to code one concept at a time, verify each on twenty examples, then combine them. Our guides to building with ChatGPT and Claude give the prompts. Start at the assistant level, where the bot marks setups and you decide, and only automate execution once your records justify it.