Build Your First Strategy

This walks through creating one strategy end to end. It assumes you have read the Technical Scanner overview and know what the four layers are.

The example: M5 entries on gold, only in the direction of the H1 trend.

1. Create the strategy

Technical tab → New strategy.

Give it a name. The ID is generated from the name and locked afterwards — accounts and historical setups reference a strategy by ID, so renaming one would break those links. The display name stays editable.

Add XAUUSD under Symbols. Leaving Symbols empty means “every symbol my terminal streams”, which is rarely what you want for a first strategy.

2. Set up the engine profile

Scroll to Engine and open the picker. Start from the standard watcher-zone engine.

You want a trend timeframe, and the standard engine does not have one — so click Fork & edit first. This copies the profile under a new name and points only this strategy at the copy, leaving every other strategy on the original.

In the forked profile’s Core section you will see the timeframe slots:

Role: Entry      Timeframe: M5

Click + Add trend timeframe and set it to H1. You now have:

Role: Entry      Timeframe: M5
Role: Trend      Timeframe: H1

That is the whole change. The H1 series now feeds the trend filter — but nothing uses it yet, which is step 4.

Leave the detection settings alone for now. Swing lookbacks, ATR length and zone sizing are worth tuning once you have a baseline to compare against, not before.

3. Pick an entry profile

Back on the strategy, open the Entry picker.

The scan mode is the most consequential field in the whole builder: it decides what pattern the scanner hunts for, and which of the other settings have any effect. Fields that cannot reach the runtime for your scan mode are greyed out, and anything you have set that cannot apply is listed under Set but ignored. Read that list — a value sitting in an inert field is the most common reason a strategy behaves differently from how it reads.

Prefer the ATR-denominated size gates over the pip ones where both exist. A pip is a broker reporting convention, not a property of the market: the same index arrives as point 0.1 from one broker and point 0.01 from another, so a pip threshold is a 10x different filter on each. ATR multiples mean the same thing everywhere.

4. Pick a strategy profile

This layer owns stops, targets and everything that can reject a signal.

To use the H1 trend slot you added, find HTF direction and set it to align. That is the only wiring needed — the filter reads whatever trend source the engine provides, and now that source is real H1 candles instead of a long lookback on M5.

Set a sensible daily cap while you are here. A strategy that fires forty times on a choppy day will tell you more about your caps than about your edge.

5. Read the validation panel

Before saving, check the panel at the bottom:

  • Schema errors — a value is the wrong type or out of range. Must be fixed.
  • Composition errors — the layers are incompatible. For example a stop-loss mode that needs pattern legs paired with a scan mode that does not produce them.
  • Realism warnings — legal, but will behave differently live than it looks. Warm-up delays from higher timeframes show up here, as do order types that fill far more readily in a backtest than in the market.

What this strategy does below the warnings is generated from your actual configuration. If it does not describe what you meant to build, the configuration is wrong, not the description.

6. Backtest it

Backtest tab → Backtest section. Pick your strategy and a date range, and run it.

You get trades, win rate, average R, max drawdown and net R, with spread, slippage, commission and swap modelled. There is no timeframe field: the run uses the strategy’s own timeframes, so a result you like is a result you can reproduce.

Judge a run on net R against max drawdown, not on net R alone. A strategy that makes 300R with a 200R drawdown is not better than one that makes 60R with a 15R drawdown.

7. Simulate a day

Backtest tab → Day Simulation. Pick a symbol, a date and your strategy.

This replays one day slice by slice and produces chart images at each step. Use it to answer “why did it not fire here?” — a backtest gives you totals, a simulation shows you the bars.

Once you are happy, link the strategy from the account’s settings and pair it with a risk preset. The preset owns position sizing; the strategy owns detection and trade management.

Start it linked to a demo account. The first thing you learn from live running is not whether the edge is real — it is whether your warm-up, symbol mapping and caps behave the way you assumed.

What to change first when it underperforms

In rough order of impact:

  1. Scan mode — the wrong pattern is not fixable by tuning thresholds.
  2. Stop-loss placement — structural versus fixed changes the R distribution more than almost anything else.
  3. Filters — a session window or trend alignment often removes a whole cluster of losses at once.
  4. Detection thresholds — prominence and leg-height gates. Tune these last, and one at a time, or you will not know which one mattered.