This strategy aims to capture strong trends in cryptocurrency markets by using multiple channels and moving averages to identify trend signals, and combines volume indicators to filter false breakouts while adaptively stopping losses to lock in profits, allowing profits to be made in trending markets.
The strategy uses a combination of fast channel, slow channel and fast moving average to identify trends. The fast channel parameters are more sensitive for capturing short-term price fluctuations; the slow channel parameters are more moderate for judging the major trend; the fast moving average parameters are in between, generating trading signals when it breaks through the channel.
Specifically, it first calculates the upper and lower rails of the fast channel, and the moving average. When the price breaks through the upper rail, if the lower rail of the slow channel is also above the moving average, a long signal is generated; conversely, when it breaks through the lower rail, it checks if the upper rail of the slow channel is below the moving average, generating a short signal.
In addition, it detects the K-line pattern, requiring multiple K-lines to be arranged in sequence to filter false breakouts; and calculates the price change rate indicator to determine if it has entered a consolidation to avoid missing reversal opportunities; and incorporates volume indicators to ensure volume follows price at breakout.
For stop loss, the strategy uses adaptive stops loss. Based on recent volatility, it dynamically adjusts the stop loss percentage. This allows locking in as much of the trend profit as possible while ensuring effective stop loss.
The biggest advantage of this strategy is that the criteria for generating trading signals are relatively strict, which can effectively filter out non-trend false breakouts and truly capture turning points in market trends. Specifically, there are several main aspects:
The combination of multiple channels and moving averages has stricter criteria and can reduce misjudgment probability.
K-line sequence validation avoids wrong signals from a single aberrant K-line.
Incorporating price change rate indicator can determine if it has entered consolidation to avoid missing reversal opportunities.
Adding volume indicator judgment ensures signals are generated only when volume follows price, avoiding ineffective breakout.
Adaptive stop loss mechanism can maximize locking in trend profits while ensuring stop loss.
So in general, this strategy has the characteristics of optimized configuration, prudent decision making, adaptive stop loss, making it very suitable for capturing trending opportunities.
Although this strategy has done a lot of optimization in filtering false breakouts and capturing trends, there are still some risks to note:
The complex parameter settings can lead to large differences between parameter combinations, requiring extensive testing to find the optimal parameters, otherwise it may generate too many false signals.
When the gap between fast moving average and channel is too small, it tends to generate frequent entries and exits, which is not conducive to persistently tracking trends.
The stop loss percentage calculation in the adaptive stop loss mechanism relies on simple standard deviation, which may lead to insufficient stop loss in extreme market conditions.
It relies heavily on technical indicators and may fail to respond to major fundamental changes.
As a trend following strategy, it underperforms in range-bound choppy markets.
To control these risks, the following measures are recommended:
Do sufficient backtesting to determine optimal parameter combinations, or consider using machine learning for parameter optimization.
Moderately widen channel intervals, lengthen moving average periods to reduce unnecessary entries.
Consider introducing more advanced volatility models like hedge fund methods.
Refer to fundamental information in a timely manner to avoid purely technical trading.
Increase judgment on market states and pause trading in choppy markets.
The strategy can be further optimized in the following ways:
Introduce machine learning algorithms to achieve automatic parameter optimization, by recording parameter performance in different market environments to build a lookup table for dynamic optimization.
Add judgment on market states, such as adding modules to determine if the market is trending or choppy, and pause trading in choppy markets to avoid unnecessary losses.
Optimize stop loss strategies, such as trailing stop loss, proportional stop loss etc.
Incorporate fundamental factors to send out alerts when major fundamental events occur, avoiding losses purely based on technical indicators.
Conduct portfolio optimization, combining this strategy with other unrelated strategies to further diversify risks.
Introduce quantitative trading framework for automated signal execution and strict risk control.
In summary, this strategy is very suitable for capturing trending opportunities in cryptocurrency markets. It uses multiple channels and moving averages to generate trading signals, and effectively filters out false breakout noise and successfully locks in trend profits. But parameters optimization, stop loss methods, market state judgement etc still need attention. With continuous improvement, it has potential for steady investment returns. It provides a great example for quantitative strategy design.
/*backtest
start: 2022-09-21 00:00:00
end: 2023-09-27 00:00:00
period: 1d
basePeriod: 1h
exchanges: [{"eid":"Futures_Binance","currency":"BTC_USDT"}]
*/
//@version=4
strategy("Extremely Overfit", overlay=true, commission_type=strategy.commission.percent, commission_value=.16, default_qty_type=strategy.percent_of_equity, default_qty_value=100, pyramiding = 1)
price = close
goLong = input(title="go long?", type=input.bool, defval=true)
goShort = input(title="go short?", type=input.bool, defval=true)
//trendRestrict = input(title="basic trend restriction?", type=input.bool, defval=false)
dynamicRestrict = true //input(title="dynamic trend restriction?", type=input.bool, defval=true)
longtrendimpt = true //input(title="additional weight on long-term trends?", type=input.bool, defval=true)
volRestrict = true //input(title="volume restriction?", type=input.bool, defval=true)
conservativeClose = false //input(title="conservative order closing?", type=input.bool, defval=false)
Restrictiveness = input ( -40,step=10,title ="Restrictiveness (higher = make fewer trades)")
volatilityImportance = 3.2 //input( 3.2, step = 0.1, minval = 0)
fastChannelLength = input( 6 )
fastChannelMargin = input ( 3.2, step = 0.1, minval = 0)
slowChannelLength = input ( 6, step = 1, minval = 0)
slowChannelMargin = input ( 1.5, step = 0.1, minval = 0)
fastHMAlength = input (4, step = 1, minval = 0)
stopLoss = input( 3, step = 0.1, minval = 0)
//altClosePeriod = input( 27, step = 1, minval = 1)
//altCloseFactor = input( 4.9, step = 0.1)
stopLossFlexibility = 50 //input(50, step=10, title="effect of volatility on SL?")
volumeMAlength = 14 //input ( 14, step = 1, minval = 1)
volumeVolatilityCutoff = 3.8 // ( 3.8, step = 1, minval = 0)
trendSensitivity = 3.8 //input ( 3.8, step = 0.1)
obvLookback = 10 //input(10, step = 10, minval = 10)
obvCorrThreshold = 0.89 //input(0.89, step = 0.01)
ROClength = 80 //input( 80, step = 10)
ROCcutoff = 5.6 //input( 5.6, step=0.1)
trendRestrict = false
//trendLookback = input ( 360, step = 10, minval = 10)
//longTrendLookback = input(720, step = 10, minval = 10)
//longTrendImportance = input(1.5, step = 0.05)
trendLookback = 360
longTrendLookback = 720
longTrendImportance = 1.5
//conservativeness = input( 2.4, step = 0.1)
conservativeness = 0
//trendPower = input( 0, step=1)
trendPower = 0
//conservativenessLookback = input( 650, step = 10, minval = 0)
conservativenessLookback = 10
//consAffectFactor = input( 0.85,step=0.01)
consAffectFactor = 0.85
//volatilityLookback = input(50, step=1, minval=2)
volatilityLookback = int(50)
recentVol = stdev(price,volatilityLookback)/sqrt(volatilityLookback)
//price channel
fastChannel = ema(price, fastChannelLength)
fastChannelUB = fastChannel * (1 + (float(fastChannelMargin) / 1000)) + (recentVol * (float(volatilityImportance) * (1 + (Restrictiveness/100))))
fastChannelLB = fastChannel * (1 - (float(fastChannelMargin) / 1000)) - (recentVol * (float(volatilityImportance) * (1 + (Restrictiveness/100))))
fchU = ((fastChannelUB < open) and (fastChannelUB < close))
fchL = ((fastChannelLB > open) and (fastChannelLB > close))
//plot(fastChannelUB)
//plot(fastChannelLB)
//slow channel
//slowChannelLBmargin = input ( 2, step = 0.1, minval = 0 )
slowChannel = ema(ema(price,slowChannelLength),slowChannelLength)
slowChannelUB = slowChannel * (1 + (float(slowChannelMargin) / 2000)) + (recentVol * (float(volatilityImportance) * (1 + (Restrictiveness/100))))
slowChannelLB = slowChannel * (1 - (float(slowChannelMargin) / 2000)) - (recentVol * (float(volatilityImportance) * (1 + (Restrictiveness/100))))
schU = ((slowChannelUB < close))
schL = ((slowChannelLB > close))
cschU = (((slowChannelUB * (1 + conservativeness)) < close))
cschL = (((slowChannelUB * (1 - conservativeness)) > close))
//plot(slowChannel,color = #00FF00)
//plot(slowChannelUB,color = #00FF00)
//plot(slowChannelLB,color = #00FF00)
fastHMA = hma(price,fastHMAlength)
fastAboveUB = (fastHMA > slowChannelUB)
fastBelowLB = (fastHMA < slowChannelLB)
//plot(fastHMA, color = #FF0000, linewidth = 2)
//consecutive candles
//consecutiveCandlesReq = input(1, step = 1, minval = 1, maxval = 4)
consecutiveCandlesReq = 1
consecutiveBullReq = float(consecutiveCandlesReq)
consecutiveBearReq = float(consecutiveCandlesReq)
cbull = ((close[0] > close[1]) and (consecutiveBullReq == 1)) or (((close[0] > close[1]) and (close[1] > close[2])) and consecutiveBullReq == 2) or (((close[0] > close[1]) and (close[1] > close[2]) and (close[2] > close[3])) and consecutiveBullReq == 3) or (((close[0] > close[1]) and (close[1] > close[2]) and (close[2] > close[3]) and (close[3] > close[4])) and consecutiveBullReq == 4)
cbear = ((close[0] < close[1]) and (consecutiveBearReq == 1)) or (((close[0] < close[1]) and (close[1] < close[2])) and consecutiveBearReq == 2) or (((close[0] < close[1]) and (close[1] < close[2]) and (close[2] < close[3])) and consecutiveBearReq == 3) or (((close[0] < close[1]) and (close[1] < close[2]) and (close[2] < close[3]) and (close[3] < close[4])) and consecutiveBearReq == 4)
//trend detection
//trendCutoff = input(0, step = 0.1)
trendCutoff = 0
trendDetectionPct = float(trendCutoff/100)
trendVal = float((close[0] - close[trendLookback])/close[0])
trendUp = (trendVal > (0 + trendDetectionPct))
trendDown = (trendVal < (0 - trendDetectionPct))
//plot(trendVal+36.5,linewidth=2)
// peak indicators
peakHigh = ((fastHMA > fastChannelUB) and (fastChannelLB > slowChannelUB))
peakLow = ((fastHMA < fastChannelLB) and (fastChannelUB < slowChannelLB))
TpeakHigh = (fastHMA > fastChannelUB) and (fastChannelUB > slowChannelUB)
TpeakLow = (fastHMA < fastChannelUB) and (fastChannelLB < slowChannelLB)
//TpeakHigh = (fastHMA > fastChannelUB) and (fastChannelLB > avg(slowChannelUB,slowChannelLB))
//TpeakLow = (fastHMA < fastChannelUB) and (fastChannelUB < avg(slowChannelLB,slowChannelUB))
//TpeakHigh = ((crossover(fastHMA,fastChannelUB)) and (fastChannelLB > slowChannelUB))
//TpeakLow = ((crossover(fastChannelLB,fastHMA)) and (fastChannelUB < slowChannelLB))
//TpeakHigh = (fastHMA > (fastChannelUB * (1 + (trendPower/800)))) and (fastChannelUB > (slowChannelUB * (1 + (trendPower/800))))
//TpeakLow = (fastHMA < (fastChannelUB * (1 - (trendPower/800)))) and (fastChannelLB < (slowChannelLB * (1 - (trendPower/800))))
//TpeakHigh = (fastHMA > (fastChannelUB * (1 + (trendPower/800)))) and (avg(fastChannelUB,fastChannelLB) > (slowChannelUB * (1 + (trendPower/800))))
//TpeakLow = (fastHMA < (fastChannelUB * (1 - (trendPower/800)))) and (avg(fastChannelLB,fastChannelUB) < (slowChannelLB * (1 - (trendPower/800))))
//plot(fastChannelUB * (1 + (trendPower/700)), color=#FF69B4)
// and for closing...
closeLong = (crossover(fastHMA,fastChannelUB) and (fastChannelLB > slowChannelUB))
closeShort = (crossover(fastChannelLB,fastHMA) and (fastChannelUB < slowChannelLB))
//closeLong = (crossover(fastHMA,fastChannelUB) and (fastChannelLB > slowChannelUB)) or (roc(price,altClosePeriod) > altCloseFactor)
//closeShort = (crossover(fastChannelLB,fastHMA) and (fastChannelUB < slowChannelLB)) or (roc(price,altClosePeriod) < (altCloseFactor) * -1)
//closeLong = (crossover(fastHMA,fastChannelUB) and (fastChannelLB > slowChannelUB)) or (((price - fastChannelUB) > (altCloseFactor * abs(((fastChannelUB - fastChannelLB)/2) - ((slowChannelUB - slowChannelLB)/2)))) and (fastChannelLB > slowChannelUB))
//closeShort = (crossover(fastChannelLB,fastHMA) and (fastChannelUB < slowChannelLB)) or (((fastChannelLB - price) > (altCloseFactor * abs(((fastChannelUB - fastChannelLB)/2) - ((slowChannelUB - slowChannelLB)/2)))) and (fastChannelUB < slowChannelLB))
//closeLong = crossover(fastHMA,fastChannelUB) and ((fastChannelLB[0] - fastChannelLB[1]) < (slowChannelUB[0] - slowChannelUB[1]))
//closeShort = crossover(fastChannelLB,fastHMA) and ((fastChannelUB[0] - fastChannelUB[1]) > (slowChannelLB[0] - slowChannelLB[1]))
//stop-loss
priceDev = stdev(price,trendLookback) * (1 + stopLossFlexibility/5)
stopLossMod = stopLoss * (1 + (priceDev/price))
//longStopPrice = strategy.position_avg_price * (1 - (stopLoss/100))
//shortStopPrice = strategy.position_avg_price * (1 + (stopLoss/100))
longStopPrice = strategy.position_avg_price * (1 - (stopLossMod/100))
shortStopPrice = strategy.position_avg_price * (1 + (stopLossMod/100))
// volume
volumeMA = ema(volume,volumeMAlength)
volumeDecrease = ((not volRestrict ) or (volumeMA[0] < ema(volumeMA[1] * (1 - (volumeVolatilityCutoff/100)),5)))
volumeCutoff = ema(volumeMA[1] * (1 - (volumeVolatilityCutoff/100)),5)
//plot(volumeMA)
//plot(volumeCutoff)
// detect volatility
//trendinessLookback = input ( 600, step = 10, minval = 0)
trendinessLookback = trendLookback
trendiness = (stdev(price,trendinessLookback)/price) * (1 - (Restrictiveness/100))
longtermTrend = ((price - price[longTrendLookback])/price)
//dynamicTrendDetected = (dynamicRestrict and (abs(trendiness * 100) < trendSensitivity))
dynamicTrendDetected = (longtrendimpt and (dynamicRestrict and (abs(trendiness * 100) < (trendSensitivity+(longtermTrend * longTrendImportance))))) or (not longtrendimpt and ((dynamicRestrict and (abs(trendiness * 100) < trendSensitivity))))
// adapt conservativeness to volatility
//consVal = sma(((stdev(price,conservativenessLookback))/price)*100,25)
consVal = sma(((stdev(price,conservativenessLookback))/price)*100,25)
cVnorm = sma(avg(consVal,3),60)
cVal = consVal - cVnorm
//conservativenessMod = conservativeness * (cVal * consAffectFactor)
conservativenessMod = conservativeness * (consVal * consAffectFactor)
//plot(consVal,linewidth=4)
//plot(cVnorm,color = #00FF00)
//plot(cVal,linewidth=2)
// ROC cutoff (for CLOSING)
//rocCloseLong = (ema(roc(price,ROClength),10) > ROCcutoff)
//rocCloseShort = (ema(roc(price,ROClength),10) < (ROCcutoff * -1))
ROCval = roc(price,ROClength)
ROCema = ema(ROCval,30)
ROCabs = abs(ROCema)
ROCallow = ROCabs < ROCcutoff
ROCallowLong = (ROCabs < ROCcutoff) or ((ROCabs >= ROCcutoff) and ((fastChannelLB < slowChannelLB) and (fastHMA < fastChannelLB)))
ROCallowShort = (ROCabs < ROCcutoff) or ((ROCabs >= ROCcutoff) and ((fastChannelUB > slowChannelUB) and (fastHMA > fastChannelUB)))
//plot(ROCallow)
// obv
evidence_obv = (correlation(price,obv[0],obvLookback))
obvAllow = evidence_obv > obvCorrThreshold
//if (not na(vrsi))
if trendRestrict or dynamicTrendDetected
//if (strategy.position_size == 0)
if not (strategy.position_size < 0)
if trendUp
//if cbear and schL and fchL and trendUp and goLong
if cbear and TpeakLow and volumeDecrease and ROCallow and goLong and obvAllow
//if cbear and peakLow and rocHigh and volumeDecrease and goLong
strategy.entry("Long", strategy.long, comment="Long")
if not (strategy.position_size > 0)
if trendDown
//if cbull and schU and fchU and trendDown and goShort
if cbull and TpeakHigh and volumeDecrease and ROCallow and goShort and obvAllow
//if cbull and peakHigh and rocLow and volumeDecrease and goShort
strategy.entry("Short", strategy.short, comment="Short")
else
//if (strategy.position_size == 0)
if not (strategy.position_size < 0)
//if cbear and peakLow and goLong
//if cbear and peakLow and volumeDecrease and ROCallow and goLong
if TpeakLow and goLong and obvAllow
strategy.entry("Long", strategy.long, comment="Long")
if not (strategy.position_size > 0)
//if cbull and peakHigh and goShort
//if cbull and peakHigh and volumeDecrease and ROCallow and goShort
if TpeakHigh and goShort and obvAllow
strategy.entry("Short", strategy.short, comment="Short")
if conservativeClose
//pkHigh = ((fastHMA > fastChannelUB) and (fastChannelUB > (slowChannelUB * (1 + conservativeness/1000))))
//pkLow = ((fastHMA < fastChannelLB) and (fastChannelLB < (slowChannelLB * (1 - conservativeness/1000))))
//pkHigh = ((fastHMA > fastChannelUB) and (fastChannelUB > (slowChannelUB * (1 + conservativenessMod/1000))))
//pkLow = ((fastHMA < fastChannelLB) and (fastChannelLB < (slowChannelLB * (1 - conservativenessMod/1000))))
pkHigh = ((fastHMA > fastChannelUB) and (fastChannelUB > (slowChannelUB * (1 + ((conservativenessMod/1000) * (1 - Restrictiveness/100))))))
pkLow = ((fastHMA < fastChannelLB) and (fastChannelLB < (slowChannelLB * (1 - ((conservativenessMod/1000) * (1 - Restrictiveness/100))))))
if (strategy.position_size > 0)
//if fastAboveUB
//if pkHigh and closeLong
if closeLong
strategy.close("Long", comment="closeLong")
if (strategy.position_size < 0)
//if fastBelowLB
//if pkLow and closeShort
if closeShort
strategy.close("Short", comment="closeShort")
else
if (strategy.position_size > 0)
//if fastAboveUB
if peakHigh
strategy.close("Long", comment="closeLong")
if (strategy.position_size < 0)
//if fastBelowLB
if peakLow
strategy.close("Short", comment="closeShort")
if (strategy.position_size > 0)
strategy.exit(id="Long", stop=longStopPrice, comment="stopLong")
if (strategy.position_size < 0)
strategy.exit(id="Short", stop=shortStopPrice, comment="stopShort")
//plot(strategy.equity, title="equity", color=color.red, linewidth=2, style=plot.style_areabr)