How to Pause a NinjaScript Strategy Around News

Why an automated strategy needs to know about news

NinjaTrader 8 has no economic calendar available to NinjaScript. A strategy sees bars, and a release-driven spike is just a large bar — indistinguishable from any other large bar, and often a very convincing one. Momentum entries fire into the spike, mean-reversion entries fire into the fade, and both are filled in a book that has thinned out precisely because everyone else knew the number was coming.

The fix is not clever logic. It is giving the strategy the schedule it is missing, and then writing three or four lines that act on it.

Reading news state from a NinjaScript strategy

The pattern is the standard one for reading any indicator from a strategy: hold a private reference, instantiate it in State.DataLoaded, and index its series in OnBarUpdate. The example below uses News Markers PRO, which publishes its blackout state for exactly this purpose.

public class NewsAwareStrategy : Strategy
{
    private BadBuddha_NewsMarkers_PRO news;

    protected override void OnStateChange()
    {
        if (State == State.SetDefaults)
        {
            Name        = "NewsAwareStrategy";
            Calculate   = Calculate.OnBarClose;
        }
        else if (State == State.DataLoaded)
        {
            // NinjaTrader generates this wrapper from the indicator's own
            // inputs. Calling it bare takes every default; pass them
            // positionally to override. IntelliSense shows the signature
            // once the add-on has been imported.
            news = BadBuddha_NewsMarkers_PRO();
            AddChartIndicator(news);
        }
    }

    protected override void OnBarUpdate()
    {
        if (CurrentBar < BarsRequiredToTrade)
            return;

        // Series are doubles, so compare against a midpoint rather than
        // testing equality with 1.0.
        bool inNewsWindow = news.InNewsWindowSeries[0] > 0.5;

        if (inNewsWindow)
        {
            if (Position.MarketPosition != MarketPosition.Flat)
                Print($"{Time[0]}  holding through: {news.UpcomingEventTitle}");

            return;   // no new entries while the window is open
        }

        // ... normal entry logic below
    }
}

Two details worth keeping. Instantiate in State.DataLoaded, not State.Configure — the bars the indicator needs are not available yet in Configure. And compare the series against a midpoint rather than testing equality with 1.0: NinjaScript series are doubles, and exact equality on a double is a habit that will eventually cost you a session.

Winding down instead of stopping dead

A binary in-window check is blunt. The countdown series let the strategy taper — stop opening new positions well before the print, then decide separately what to do with anything already open:

double minutesToHigh = news.MinutesToNextHighSeries[0];

// Wind down rather than stopping dead.
if (minutesToHigh > 0 && minutesToHigh <= 15)
    return;                                   // stop opening new positions

if (minutesToHigh > 0 && minutesToHigh <= 5 &&
    Position.MarketPosition == MarketPosition.Long)
{
    Print($"{Time[0]}  flattening {minutesToHigh:F1} min before " +
          $"{news.UpcomingEventTitle}");
    ExitLong("NewsFlat", "");
}

MinutesToNextMediumSeries works the same way for medium-impact releases, which most strategies will want to treat differently — visibility and a log line rather than a full stand-down.

What a news filter does to a backtest

It will probably make the equity curve look better, and you should be suspicious of that. A backtest fills at bar prices. It does not model the widened spread, the thinned book, or the stop that sweeps twenty ticks to find size — which means the trades a news filter removes are exactly the trades the backtest was already scoring too generously. The improvement you see is real; it is just smaller than the number on the screen.

There is a second trap. A news indicator can only mark events it has calendar data for, so a backtest over a period the calendar does not cover runs with the filter silently disabled — and a filter that never fires appears to cost nothing. Check how far back your events actually load before reading anything into a multi-year result.

The honest test is a live or sim run with logging on. Run the filter in observation mode first, record every bar where the strategy would have stood aside, and compare those trades against the rest of the session. That comparison is worth more than any amount of historical optimisation.