0bps

Strategy vault / Breakout

5 day High Low Breakout Price Channel Strategy

This strategy is named "5-day High-Low Breakout Price Channel Strategy". It constructs a price channel using the highest high and lowest low prices over the past 5 trading days, and trades breakouts of the channel…

Author
ChaoZhang
Complexity
Intermediate · 117 lines
Parameters
5
Backtest bar
4h · intraday
Tested on
BTC/USDT · futures

This strategy is named "5-day High-Low Breakout Price Channel Strategy". It constructs a price channel using the highest high and lowest low prices over the past 5 trading days, and trades breakouts of the channel.

The steps are:

  1. Calculate the highest high and lowest low prices over the past 5 trading days.

  2. Take the top 2 highest high prices to build the upper rail, and the bottom 2 lowest low prices for the lower rail.

  3. When price rises above the upper rail by a certain percentage (e.g. 0.3%), a buy signal is generated.

  4. When price drops below the lower rail by a certain percentage, a sell signal is generated.

  5. After entry, the second highest/lowest price is used as the stop loss, or take profit at a certain percentage (e.g. 0.5%) is tracked before exiting.

The advantage is using key high/low price breakouts to determine trend reversal points. Breaking the channel represents concentrated price-volume thrusts. But over-trading in ranging markets should be prevented.

In summary, watching key price area breakouts is a classical trend following approach. But traders still need confirmation with other indicators and parameter optimization to maximize the strategy's utility.

Parameters

NameDefaultWhat it controls
v_input_10.3Entry - % point above high/low
v_input_20.5Profit Target %
v_input_3BOTHTrade Type: LONG, SHORT, BOTH ( case sensitive )
v_input_42High/Low line width (Enter 0 to hide )
v_input_5NODisplay sorted low/high: YES, NO ( case sensitive )

Source Pine Script

/*backtest
start: 2023-08-13 00:00:00
end: 2023-09-12 00:00:00
period: 4h
basePeriod: 15m
exchanges: [{"eid":"Futures_Binance","currency":"BTC_USDT"}]
*/

// @version=4
// Based on Sort pseudo-array v2 by apozdnyakov https://www.tradingview.com/script/IUlIoSnA-Sort-pseudo-array-v2/

strategy(title="5 Day high/low breakout strategy", shorttitle="5 days range breakout", overlay=true)
entry_factor = input(title="Entry - % point above high/low", type=input.float, defval=0.3, minval=0, maxval=5)
profit_target = input(title="Profit Target %", type=input.float, defval=0.5, minval=0, maxval=5)
trade_type   = input(defval = "BOTH", title = "Trade Type: LONG, SHORT, BOTH ( case sensitive )", type = input.string)
width = input(defval = 2, title = "High/Low line width (Enter 0 to hide )", type = input.integer, minval=0, maxval=5)
debug = input(defval= "NO", title = "Display sorted low/high: YES, NO ( case sensitive )", type = input.string)

high_day1 = security(syminfo.tickerid, "D", high[1], lookahead = barmerge.lookahead_on)
high_day2 = security(syminfo.tickerid, "D", high[2], lookahead = barmerge.lookahead_on)
high_day3 = security(syminfo.tickerid, "D", high[3], lookahead = barmerge.lookahead_on)
high_day4 = security(syminfo.tickerid, "D", high[4], lookahead = barmerge.lookahead_on)
high_day5 = security(syminfo.tickerid, "D", high[5], lookahead = barmerge.lookahead_on)

low_day1 = security(syminfo.tickerid, "D", low[1], lookahead = barmerge.lookahead_on)
low_day2 = security(syminfo.tickerid, "D", low[2], lookahead = barmerge.lookahead_on)
low_day3 = security(syminfo.tickerid, "D", low[3], lookahead = barmerge.lookahead_on)
low_day4 = security(syminfo.tickerid, "D", low[4], lookahead = barmerge.lookahead_on)
low_day5 = security(syminfo.tickerid, "D", low[5], lookahead = barmerge.lookahead_on)


// sorts a list of up to the fixed length
sort_all(type) =>
	float s0 = na
	float s1 = na
	float s2 = na
	float s3 = na
	float s4 = na

    h_val = security(syminfo.tickerid, "D", high, false)
	
	float min = na
	float last = na

	for i = 0 to 4
		float min_local = na
		float last_local = na
		float val = na
		
		for l = 0 to 4
    		if type == "high"
    		    val := l == 0 ? high_day1 : val
                val := l == 1 ? high_day2 : val
		        val := l == 2 ? high_day3 : val
    		    val := l == 3 ? high_day4 : val
	    	    val := l == 4 ? high_day5 : val
            else
    		    val := l == 0 ? low_day1 : val
	    	    val := l == 1 ? low_day2 : val
		        val := l == 2 ? low_day3 : val
    		    val := l == 3 ? low_day4 : val
	    	    val := l == 4 ? low_day5 : val

			if(na(min) or val > min or (val == min and l > last))
				new_min_local = na(min_local) ? val : min(min_local, na(min) ? val : max(min, val))
				if(na(min_local) or new_min_local != min_local)
					last_local := l
					min_local := new_min_local
		
		min := min_local
		last := last_local
		
		s0 := i == 0 ? min : s0
		s1 := i == 1 ? min : s1
		s2 := i == 2 ? min : s2
		s3 := i == 3 ? min : s3
		s4 := i == 4 ? min : s4

	[s0, s1, s2, s3, s4]


[high5, high4, high3, high2, high1] = sort_all("high")
[low1, low2, low3, low4, low5] = sort_all("low")

plot(high1, color = color.blue, style=plot.style_circles, linewidth=width)
plot(high2, color = color.red, style=plot.style_circles, linewidth=width)
plot(low1, color = color.blue, style=plot.style_circles, linewidth=width)
plot(low2, color = color.red, style=plot.style_circles, linewidth=width)


if close >= (high1 * (1 + entry_factor/100)) and strategy.position_size == 0 and hour <= 12
    strategy.entry(id = "long_entry", long = true, qty = 1, stop = high2)

strategy.close(id = "long_entry", when = strategy.position_size != 0 and (close < high2  or close > high1 * (1 + (entry_factor + profit_target)/100)))


if close <= (low1 * (1 - entry_factor/100)) and strategy.position_size == 0 and hour <= 12
    strategy.entry(id = "short_entry", long = false, qty = 1, stop = low2)

strategy.close(id = "short_entry", when = strategy.position_size != 0 and (close > low2  or close < low1 * (1 - (entry_factor + profit_target)/100)))
    

if (hour == 14)
    strategy.close_all()

//No more than 1 order per day
// strategy.risk.max_intraday_filled_orders(2)


//Check whether this is the first bar of the day? If yes, display highs for last 5 days
// t = time("1440", session.regular)
// is_first = na(t[1]) and not na(t) or t[1] < t

// if (is_first and debug != "NO")  
//     label.new(bar_index, na, tostring(high1) + ", " + tostring(high2) + ", " + tostring(high3) + ", " + tostring(high4) + ", " + tostring(high5), style=label.style_cross, yloc=yloc.abovebar)
//     label.new(bar_index, na, tostring(low1) + ", " + tostring(low2) + ", " + tostring(low3) + ", " + tostring(low4) + ", " + tostring(low5), style=label.style_cross, yloc=yloc.belowbar)

Published by ChaoZhang on the FMZ strategy square · last modified 2023-09-13. Collected in the Quant Trading Vault. Unvalidated research: re-test with your own data and costs before trusting it.

Build this with AgenKit

A spec is not a trading system. Paste this into your coding agent and AgenKit runs a specialist team — architecture, test-first build, risk gates, review — with paper trading on by default.

in Claude Code, Cursor, Codex & 4 more
/agenkit build a production-grade trading bot with a backtest engine, risk controls and paper trading by default from strategies/5-day-High-Low-Breakout-Price-Channel-Strategy.md
  1. npx agenkit install engineering-kit in your project
  2. Paste the command into your agent and approve each phase
  3. Get tests, backtest, risk controls and a paper-trading switch

Understand it first

Before you run this, read the lesson behind the idea and how to backtest it honestly. Specs like this one are the survivors of an unknown number of trials; deflate their results accordingly.

Related specs

esc

Try , , or .

↑↓ move↵ open