该策略综合利用Voss预测滤波器和Ehlers瞬时趋势线指标,以识别市场周期性转折点,实现量化交易。Voss滤波器可提前发出买入/卖出信号,而瞬时趋势线指标则用于判断整体趋势方向,减少Voss滤波器在趋势市场中的误导。该策略可用于比特币等周期性比较明显的品种,在回测中表现较好。
Voss预测滤波器来自约翰·F·埃勒斯的文章《A Peek Into The Future》。该滤波器的计算公式如下:
_filt = 0.5 * _s3 * _x1 + _f1 * _s2 * _filt[1] - _s1 * _filt[2]
_voss = _x2 * _filt - _sumC
其中,_x1为价格的一阶差分;_x2为平滑因子;_s1、_s2、_s3为滤波参数;_f1为周期参数;_filt为滤波结果;_voss为最终输出。
该滤波器可看作是一种平滑滤波,它强调当前和过去几个周期的信息,从而提前发出买入/卖出信号。由于内在的组 delays,它就像“看进了未来”一样,可在其他指标之前发出预测性信号。
瞬时趋势线指标由以下公式计算:
_it = (_a-((_a*_a)/4.0))*_src+0.5*_a*_a*_src[1]-(_a-0.75*_a*_a)*_src[2]+2*(1-_a)*nz(_it[1])+-(1-_a)*(1-_a)*nz(_it[2])
该指标实时绘制一条与价格最符合的趋势线,可准确判断趋势方向和强弱。
当Voss由负转正,并上穿滤波结果时产生买入信号。
当Voss由正转负,并下穿滤波结果时产生卖出信号。
同时,只有当瞬时趋势线指标确认趋势方向时,才会发出交易信号。这能过滤掉Voss滤波器在趋势市场中可能发出的错误信号。
可通过以下方法降低风险:
该策略可从以下方面进行优化:
该策略综合Voss滤波器和趋势指标,可有效识别市场的周期性反转点。通过优化参数,控制风险,该策略可实现稳定的量化交易系统。它可广泛应用于具有明显周期性的品种,在回测中已展现出良好的交易效果。总体而言,该策略具有独特的预测能力,且可通过多方面优化,具有广阔的应用前景。
/*backtest
start: 2023-08-19 00:00:00
end: 2023-09-18 00:00:00
period: 1h
basePeriod: 15m
exchanges: [{"eid":"Futures_Binance","currency":"BTC_USDT"}]
*/
// A Peek Into the Future
// John F. Ehlers
// TASC Aug 2019
// Created by e2e4mfck for tradingview.com
// Modified by © Bitduke
//@version=4
//strategy("Voss Strategy (Filter + IT)", overlay=false, calc_on_every_tick=false,pyramiding=0, default_qty_type=strategy.cash,default_qty_value=1000, currency=currency.USD, initial_capital=1000,commission_type=strategy.commission.percent, commission_value=0.075)
// voss filter
source = input(close, type = input.source)
period = input(20, type = input.integer)
predict = input(4, type = input.integer)
bandwidth = input(0.25, type = input.float)
// it trendline
src = input(hl2, title="Source IT")
a = input(0.07, title="Alpha", step=0.01)
fr = input(false, title="Fill Trend Region")
ebc = input(false, title="Enable barcolors")
hr = input(false, title="Hide Ribbon")
voss_filter (_period, _predict, _bandwidth, _source) =>
float _filt = 0, float _sumC = 0, float _voss = 0
_PI = 2 * asin(1)
_order = 3 * _predict
_f1 = cos(2 * _PI / _period)
_g1 = cos(_bandwidth * 2 * _PI / _period)
_s1 = 1 / _g1 - sqrt(1 / (_g1 * _g1) - 1)
_s2 = 1 + _s1
_s3 = 1 - _s1
_x1 = _source - _source[2]
_x2 = (3 + _order) / 2
for _i = 0 to (_order - 1)
_sumC := _sumC + ((_i + 1) / _order) * _voss[_order - _i]
if bar_index <= _order
_filt := 0
_voss := 0
else
_filt := 0.5 * _s3 * _x1 + _f1 * _s2 * _filt[1] - _s1 * _filt[2]
_voss := _x2 * _filt - _sumC
[_voss, _filt]
[Voss, Filt] = voss_filter(period, predict, bandwidth, source)
instantaneous_trendline (_src, _a, _freq, _ebc, _hr) =>
_it = 0.0
_it := (_a-((_a*_a)/4.0))*_src+0.5*_a*_a*_src[1]-(_a-0.75*_a*_a)*_src[2]+2*(1-_a )*nz(_it[1], ((_src+2*_src[1]+_src[2])/4.0))-(1-_a)*(1-_a)*nz(_it[2], ((_src+2*_src[1]+_src[2])/4.0))
_lag = 2.0*_it-nz(_it[2])
[_it, _lag]
[it, lag] = instantaneous_trendline(src, a, fr, ebc, hr)
// - - - - - - - - - - //
plot(Filt, title = "Filter", style = plot.style_line, color = color.red, linewidth = 2)
plot(Voss, title = "Voss", style = plot.style_line, color = color.blue, linewidth = 2)
hline(0.0, title = "Zero", linestyle = hline.style_dashed, color = color.black, linewidth = 1)
plot(hr? na:it, title="IT Trend", color= fr? color.gray : color.red, linewidth=1)
plot(hr? na:lag, title="IT Trigger", color=fr? color.gray : color.blue, linewidth=1)
// Strategy Logic
longCondition = lag < it and crossover(Voss,Filt)
shortCondition = it > lag and crossover(Filt,Voss)
strategy.entry("Voss_Short", strategy.short, when=shortCondition)
strategy.entry("Voss_Long", strategy.long, when=longCondition)
// === Backtesting Dates === thanks to Trost
testPeriodSwitch = input(true, "Custom Backtesting Dates")
testStartYear = input(2019, "Backtest Start Year")
testStartMonth = input(1, "Backtest Start Month")
testStartDay = input(1, "Backtest Start Day")
testStartHour = input(0, "Backtest Start Hour")
testPeriodStart = timestamp(testStartYear, testStartMonth, testStartDay, testStartHour, 0)
testStopYear = input(2020, "Backtest Stop Year")
testStopMonth = input(2, "Backtest Stop Month")
testStopDay = input(29, "Backtest Stop Day")
testStopHour = input(0, "Backtest Stop Hour")
testPeriodStop = timestamp(testStopYear, testStopMonth, testStopDay, testStopHour, 0)
testPeriod() =>
time >= testPeriodStart and time <= testPeriodStop ? true : false
testPeriod_1 = testPeriod()
isPeriod = true
// === /END
if not isPeriod
strategy.cancel_all()
strategy.close_all()