Strategy: Donchian / Opening-Range Breakout

Some of the most durable systems in trading history are breakouts, and the logic is disarmingly simple: when price pushes past a level it could not cross before, momentum has likely shifted, so you join the move. The Turtles built a fortune on exactly this idea. In this final playbook chapter we code two breakout classics - the Donchian channel, which trades a break of the highest high and lowest low over a lookback window, and the opening-range breakout (ORB), which trades a break of the first few minutes' range of the session.

They are cousins. Donchian is a swing-trading workhorse that works on daily bars and never needs a clock. ORB is its intraday sibling, anchored to the morning and governed by TimeNum(). Build both and you will own the two ends of the breakout family.

The Donchian channel

Richard Donchian's channel is just two lines: the highest high and the lowest low over the last N bars. A close - or here, a high - that crosses the upper line is a fresh breakout to the upside; a break of the lower line ends the trade.

The one subtlety that beginners miss: you must take the channel from the previous bar with Ref(..., -1). The current bar's own high is part of HHV(High, N), so without the shift the price could never "break out" of a level that always includes itself. Shifting back one bar fixes that and makes the signal non-repainting in the bargain.

_SECTION_BEGIN("Donchian Breakout System");

SetChartOptions(0, chartShowArrows | chartShowDates);
_N(Title = StrFormat("{{NAME}} - {{INTERVAL}} {{DATE}}  O %g  H %g  L %g  C %g  (%.1f%%)",
    O, H, L, C, SelectedValue(ROC(C, 1))));
Plot(Close, "Close", colorDefault, styleCandle | styleNoTitle | GetPriceStyle());

Length = Param("Channel Length", 20, 5, 100, 1);

// Channel from the PREVIOUS bar - so a breakout is a genuine fresh event
upper = Ref(HHV(High, Length), -1);
lower = Ref(LLV(Low,  Length), -1);
Plot(upper, "Upper", colorGreen, styleThick);
Plot(lower, "Lower", colorRed,   styleThick);

// Long on a break of the upper channel; exit on a break of the lower
Buy  = Cross(High, upper);
Sell = Cross(lower, Low);

Buy  = ExRem(Buy,  Sell);
Sell = ExRem(Sell, Buy);

// Fill at the breakout level, gap-safe (never better than the open)
SetTradeDelays(0, 0, 0, 0);
BuyPrice  = Max(Open, upper);
SellPrice = Min(Open, lower);
SetPositionSize(100, spsPercentOfEquity);

PlotShapes(IIf(Buy,  shapeUpArrow,   shapeNone), colorLime,   0, Low,  -20);
PlotShapes(IIf(Sell, shapeDownArrow, shapeNone), colorOrange, 0, High,  20);

_SECTION_END();

The opening-range breakout

The ORB captures the high and low of the first few minutes of the session, then trades the first break of that range. The intuition is that the opening auction sets the day's battle lines; a decisive break of them often runs. Everything hinges on the same session-anchoring trick from the VWAP chapter, plus a frozen snapshot of the range.

_SECTION_BEGIN("Opening Range Breakout");

SetChartOptions(0, chartShowArrows | chartShowDates);
Plot(Close, "Close", colorDefault, styleCandle | styleNoTitle | GetPriceStyle());

// First N bars of the session define the range. On 5-min bars, 3 bars = 15 min.
ORBars = Param("Opening Range Bars", 3, 1, 12, 1);

newDay = Day() != Ref(Day(), -1);
barNum = 1 + BarsSince(newDay);            // bar number within the session

// Freeze the range high/low once the opening range completes
ORHigh = ValueWhen(barNum == ORBars, HighestSince(newDay, High));
ORLow  = ValueWhen(barNum == ORBars, LowestSince(newDay, Low));

Plot(IIf(barNum > ORBars, ORHigh, Null), "OR High", colorGreen, styleThick);
Plot(IIf(barNum > ORBars, ORLow,  Null), "OR Low",  colorRed,   styleThick);

tn        = TimeNum();
tradeWin  = barNum > ORBars AND tn <= 143000;  // break out after the range, before 14:30
squareOff = tn >= 151500;

Buy  = Cross(High, ORHigh) AND tradeWin;
Sell = Cross(ORLow, Low) OR squareOff;

Buy  = ExRem(Buy,  Sell);
Sell = ExRem(Sell, Buy);

SetTradeDelays(0, 0, 0, 0);
BuyPrice  = Max(Open, ORHigh);
SellPrice = Min(Open, ORLow);
SetPositionSize(100, spsPercentOfEquity);

PlotShapes(IIf(Buy,  shapeUpArrow,   shapeNone), colorLime,   0, Low,  -20);
PlotShapes(IIf(Sell, shapeDownArrow, shapeNone), colorOrange, 0, High,  20);

_SECTION_END();

HighestSince(newDay, High) gives the highest high since the session opened; sampled exactly on the bar where barNum == ORBars, that is the opening-range high. ValueWhen then holds that frozen value across the rest of the day, so ORHigh stays put as a level to break - it does not creep upward as the day extends.

Backtesting breakouts honestly

Breakouts are the systems where sloppy fills flatter a backtest most. By definition you are entering at the exact moment price is moving fastest, so the gap between the level and your real fill - slippage - is wider than for a mean-reversion entry. Two defences:

First, the Max(Open, upper) / Min(Open, lower) fills above already stop the backtest from pretending you bought at a level the market gapped past. Second, set a realistic slippage and brokerage in the Analysis settings, then re-read the report. A breakout system that only works at zero cost is not a system.

Try it yourself

• Run the Donchian system on NIFTY daily data with a 20-bar channel, then try 55 bars (the classic Turtle length) and compare.
• Add a 100-day EMA filter so you only take Donchian longs when price is above it - does the trend filter cut the false breaks?
• Run the ORB on BANKNIFTY 5-minute data with ORBars = 3, then with ORBars = 6, and see which range the instrument respects.
• Add a fixed stop below ORLow for longs and measure the effect on drawdown.

Recap

• A Donchian channel is the highest high and lowest low over N bars; take it from the previous bar with Ref(..., -1) so a breakout is a genuine event.
• An opening-range breakout freezes the first few minutes' high and low with BarsSince, HighestSince and ValueWhen, then trades a break within a time window governed by TimeNum().
• Fill breakouts gap-safely with Max(Open, level) and include realistic slippage, because breakout entries suffer the worst fills.
• Expect false breaks; manage them with a stop and a trend filter, and validate everything in sandbox first.

That completes the strategy playbook - five systems you can run, tune and trust. Next we take a system live, starting with reading a higher timeframe on a lower-timeframe chart.