Strategy vault / Machine learning
Loft Stop Strategy
This strategy uses a Kalman filter to track prices and dynamically adjusts the stop loss point with a stop loss line to achieve a sliding stop loss.
Overview
This strategy uses a Kalman filter to track prices and dynamically adjusts the stop loss point with a stop loss line to achieve a sliding stop loss.
Principle
This strategy uses a Kalman filter to track prices in real time. The Kalman filter contains two equations:
Prediction equation:
smooth = kf[1] + dk * sqrt(gain / 10000 * 2)
Update equation:
kf = smooth + velo
where dk is the prediction error, gain is the Kalman gain that determines tracking sensitivity.
In addition, the strategy uses a sliding stop loss line to lock in profits. The initial stop loss distance is the stop loss percentage setting, such as 2%.
When long, if the price rises, the stop loss line also moves up gradually approaching the Kalman line, with a step size of downStep, such as 0.5%. If the price falls to the stop loss, reopen the position and set the initial stop loss distance.
Short is similar.
Thus, the strategy can gradually lock in profits according to the trend, with good risk management.
Advantages
Use Kalman filter to track prices in real time with fast response.
Lock in profits with sliding stop loss line, achieving good risk management. Customizable stop loss distance.
Flexibly choose long/short or only long/short.
Actively or conservatively stop loss based on trend.
Flexibly set take profit and stop loss as needed.
Risks
Improper parameter settings of Kalman filter may lead to unstable tracking.
Slippage may trigger stop loss point prematurely. Widen stop loss distance appropriately.
Sliding stop loss is not suitable for strong trending markets, should follow trend.
Stop loss may trigger frequently in ranging markets. Widen stop loss distance or don't use sliding stop loss.
Optimization
Incorporate more indicators to optimize entry timing.
Adjust stop loss line movement step based on market volatility.
Use machine learning to train optimal stop loss parameters.
Incorporate more risk indicators to dynamically adjust position sizing.
Conclusion
The loft stop strategy uses a Kalman filter to track price changes and lock in profits with a sliding stop loss line, ensuring profitability while controlling risks. It is a reliable and easily optimized strategy. Combining it with trend judgment and dynamic position sizing can achieve even better strategy performance.
Parameters
| Name | Default | What it controls |
|---|---|---|
| v_input_float_1 | true | Kalman Gain: |
| v_input_1_close | 0 | Source:: close | high | low | open | hl2 | hlc3 | hlcc4 | ohlc4 |
| v_input_float_2 | 2 | Beginning Approach(%): |
| v_input_float_3 | 0.5 | Final Approach(%): |
| v_input_float_4 | 0.005 | Approach Decrease Step: |
| v_input_float_5 | 1.5 | Take Profit: |
| v_input_float_6 | false | Stop Loss: |
| v_input_bool_1 | true | Long Entry |
| v_input_bool_2 | true | Short Entry |
| v_input_int_1 | true | From Day |
| v_input_int_2 | true | From Month |
| v_input_int_3 | 2021 | From Year |
| v_input_int_4 | 30 | To Day |
| v_input_int_5 | 12 | To Month |
| v_input_int_6 | 2022 | To Year |
Source Pine Script
/*backtest
start: 2023-09-06 00:00:00
end: 2023-10-06 00:00:00
period: 2h
basePeriod: 15m
exchanges: [{"eid":"Futures_Binance","currency":"BTC_USDT"}]
*/
// This source code is subject to the terms of the Mozilla Public License 2.0 at https://mozilla.org/MPL/2.0/
// © BigCoinHunter
//@version=5
// strategy(title='Loft Strategy V1', overlay=true,
// pyramiding=0, default_qty_type=strategy.fixed,
// default_qty_value=100, initial_capital=100000,
// currency=currency.USD, commission_value=0.05,
// commission_type=strategy.commission.percent,
// process_orders_on_close=true)
//-------------- fetch user inputs ------------------
gain = input.float(title="Kalman Gain:", defval=1.0, minval=1.0, maxval=5000.0, step=100.0)
src = input(defval=close, title='Source:')
stopPercentMax = input.float(title='Beginning Approach(%):', defval=2.0, minval=0.1, maxval=30.0, step=0.1)
stopPercentMin = input.float(title='Final Approach(%): ', defval=0.5, minval=0.1, maxval=30.0, step=0.1)
downStep = input.float(title='Approach Decrease Step:', defval=0.005, minval=0.0, maxval = 5, step=0.005)
tp = input.float(title="Take Profit:", defval=1.5, minval=0.0, maxval=100.0, step=0.1) * 0.01
sl = input.float(title="Stop Loss: ", defval=0.0, minval=0.0, maxval=100.0, step=0.1) * 0.01
longEntry = input.bool(defval=true, title= 'Long Entry', inline="11")
shortEntry = input.bool(defval=true, title='Short Entry', inline="11")
//---------- backtest range setup ------------
fromDay = input.int(defval = 1, title = "From Day", minval = 1, maxval = 31)
fromMonth = input.int(defval = 1, title = "From Month", minval = 1, maxval = 12)
fromYear = input.int(defval = 2021, title = "From Year", minval = 2010)
toDay = input.int(defval = 30, title = "To Day", minval = 1, maxval = 31)
toMonth = input.int(defval = 12, title = "To Month", minval = 1, maxval = 12)
toYear = input.int(defval = 2022, title = "To Year", minval = 2010)
//------------ time interval setup -----------
start = timestamp(fromYear, fromMonth, fromDay, 00, 00) // backtest start window
finish = timestamp(toYear, toMonth, toDay, 23, 59) // backtest finish window
window() => true // create function "within window of time"
//------- define the global variables ------
enterLongComment = "ENTER LONG"
exitLongComment = "EXIT LONG"
enterShortComment = "ENTER SHORT"
exitShortComment = "EXIT SHORT"
longTPSL = "Long TP/SL"
longTP = "Long TP"
longSL = "Long SL"
shortTPSL = "Short TP/SL"
shortTP = "Short TP"
shortSL = "Short SL"
var bool long = true
var bool stoppedOutLong = false
var bool stoppedOutShort = false
var float kf = 0.0
var float velo = 0.0
//------ kalman filter calculation --------
dk = src - nz(kf[1], src)
smooth = nz(kf[1], src) + dk * math.sqrt(gain / 10000 * 2)
velo := nz(velo[1], 0) + gain / 10000 * dk
kf := smooth + velo
//--------- calculate the loft stopLoss line ---------
var stopPercent = stopPercentMax
var stopLoss = kf - kf * (stopPercent /100)
if long == true
stopLoss := kf - (kf * (stopPercent / 100))
if long[1] == true and stopLoss <= stopLoss[1]
stopLoss := stopLoss[1]
else if (long[1] == true)
stopPercent := stopPercent - downStep
if(stopPercent < stopPercentMin)
stopPercent := stopPercentMin
if(kf < stopLoss)
long := false
stopPercent := stopPercentMax
stopLoss := kf + (kf * (stopPercent / 100))
else
stopLoss := kf + (kf * (stopPercent / 100))
if long[1] == false and stopLoss >= stopLoss[1]
stopLoss := stopLoss[1]
else if(long[1] == false)
stopPercent := stopPercent - downStep
if(stopPercent < stopPercentMin)
stopPercent := stopPercentMin
if(kf > stopLoss)
long := true
stopPercent := stopPercentMax
stopLoss := kf - (kf * (stopPercent / 100))
//--------- calculate the input/output points -----------
longProfitPrice = strategy.position_avg_price * (1 + tp) // tp -> take profit percentage
longStopPrice = strategy.position_avg_price * (1 - sl) // sl -> stop loss percentage
shortProfitPrice = strategy.position_avg_price * (1 - tp)
shortStopPrice = strategy.position_avg_price * (1 + sl)
//------------------- determine buy and sell points ---------------------
buySignall = window() and long and (not stoppedOutLong)
sellSignall = window() and (not long) and (not stoppedOutShort)
//---------- execute the strategy -----------------
if(longEntry and shortEntry)
if long
strategy.entry("LONG", strategy.long, when = buySignall, comment = enterLongComment)
stoppedOutLong := true
stoppedOutShort := false
else
strategy.entry("SHORT", strategy.short, when = sellSignall, comment = enterShortComment)
stoppedOutLong := false
stoppedOutShort := true
else if(longEntry)
strategy.entry("LONG", strategy.long, when = buySignall, comment = enterLongComment)
strategy.close("LONG", when = sellSignall, comment = exitLongComment)
if long
stoppedOutLong := true
else
stoppedOutLong := false
else if(shortEntry)
strategy.entry("SHORT", strategy.short, when = sellSignall, comment = enterShortComment)
strategy.close("SHORT", when = buySignall, comment = exitShortComment)
if not long
stoppedOutShort := true
else
stoppedOutShort := false
//----------------- take profit and stop loss -----------------
if(tp>0.0 and sl>0.0)
if ( strategy.position_size > 0 )
strategy.exit(id="LONG", limit=longProfitPrice, stop=longStopPrice, comment = longTPSL)
else if ( strategy.position_size < 0 )
strategy.exit(id="SHORT", limit=shortProfitPrice, stop=shortStopPrice, comment = shortTPSL)
else if(tp>0.0)
if ( strategy.position_size > 0 )
strategy.exit(id="LONG", limit=longProfitPrice, comment = longTP)
else if ( strategy.position_size < 0 )
strategy.exit(id="SHORT", limit=shortProfitPrice, comment = shortTP)
else if(sl>0.0)
if ( strategy.position_size > 0 )
strategy.exit(id="LONG", stop=longStopPrice, comment = longSL)
else if ( strategy.position_size < 0 )
strategy.exit(id="SHORT", stop=shortStopPrice, comment = shortSL)
//------------- plot charts ---------------------
lineColor1 = long ? color.green : color.red
lineColor2 = long ? color.aqua : color.fuchsia
kalmanLine = plot(kf, color=lineColor1, linewidth=3, title = "Kalman Filter")
stopLine = plot(stopLoss, color=lineColor2, linewidth=2, title = "Stop Loss Line")
Published by ChaoZhang on the FMZ strategy square · last modified 2023-10-07. 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.
/agenkit build a production-grade trading bot with a backtest engine, risk controls and paper trading by default from strategies/Loft-Stop-Strategy.md- npx agenkit install engineering-kit in your project
- Paste the command into your agent and approve each phase
- 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
- AI Driven Dynamic Volatility Adaptive Trend Breakout Trading StrategyPine Script · Machine learning · 2d
- AI Powered Volatility Price System Divergence Trading StrategyPine Script · Machine learning · 1h
- Configurable BBRSIAroon Strategy BacktestPine Script · Machine learning · 5m
- Double Decker RSI Trading StrategyPine Script · Machine learning · 1h
- Dual Moving Average Matching Strategy Based on Bollinger BandsPine Script · Machine learning · 1h
- Dynamic Weighted Moving Average Trading StrategyPine Script · Machine learning · 1d