On-tick vs on-close

Every condition has an evaluate field. The default is on_close: the engine waits for the strategy's timeframe bar to close, then checks the condition against that bar's values. The alternative is on_tick: the engine checks the condition every minute against the live 1-minute close.

On-close (default)

Use this for any condition that reads an indicator value: RSI, ADX, MACD, EMA, Bollinger Bands, Williams %R, anything else. Indicator values are only meaningful once the bar that produced them closes. On-close also matches backtest semantics exactly, so the live behavior mirrors what you measured.

On-tick

Use this for price-anchored conditions that need to fire intra-bar:

• "Live price entered the demand zone"
• "Price crossed the breakout level"
• "Price is in a specific range" (limit-order-like)

The engine swaps the condition's price columns (open/high/low/close) for the latest 1-minute close. Indicator values stay at their last-closed-bar reading.

What you give up

Faster reaction is rarely free. In our internal data, identical strategies switched from on-close to on-tick caught roughly 2x the entries but the win rate dropped from 56% to 40%. The extra entries were the noisy intra-bar wicks that would have been gone by the next close.

Reach for on-tick when the strategy is fundamentally about price touching a level - not as a generic "more signals" knob.

Backtest fidelity

When the engine sees an on-tick condition, the backtest walks each bar as four synthetic sub-bars (open -> low -> high -> close) and probes the entry rules at each step. The first sub-bar that satisfies all conditions fires the entry, and the entry price equals that sub-bar's price.

This gives backtest fills that match the live runtime's intra-bar behavior. The approximation is faithful for zone-touch and level-break strategies. It's less precise for momentum strategies with tight intra-bar stops where the exact path within the bar matters - the engine doesn't know whether price went up-then-down or down-then-up within a bar, it just visits both extremes.

Mixing both modes

Conditions inside the same rule can use different modes. Common pattern: indicator filter on-close, price-touch trigger on-tick.

{
  "long": {
    "logic": "AND",
    "conditions": [
      { "type": "comparison", "indicator": "rsi", "operator": "<", "value": 30 },
      { "type": "price_in_supply_demand_zone", "zone": "demand", "evaluate": "on_tick" }
    ]
  }
}

The RSI check runs once per closed bar; the zone-touch runs every minute. The entry fires the minute both are true.