Strategi Dagangan Pembalikan Purata Bergerak Bidirectional adalah strategi perdagangan kuantitatif yang dibina berdasarkan teori pembalikan purata harga. Strategi ini menangkap peluang pembalikan harga dengan menubuhkan beberapa purata bergerak dan memasuki pasaran apabila harga menyimpang secara ketara dari purata bergerak, dan keluar apabila ia kembali.
Idea teras strategi ini adalah pembalikan purata harga, yang menunjukkan bahawa harga cenderung turun naik di sekitar nilai purata, dan mempunyai peluang yang lebih tinggi untuk kembali apabila mereka menyimpang terlalu jauh dari purata. Khususnya, strategi ini menetapkan tiga kumpulan purata bergerak: purata bergerak kemasukan, purata bergerak keluar, dan purata bergerak berhenti-kerugian. Ia akan membuka kedudukan panjang atau pendek yang sesuai apabila harga mencapai purata bergerak kemasukan; kedudukan dekat apabila harga mencapai purata bergerak keluar; dan kawalan kerugian dengan purata bergerak berhenti-kerugian sekiranya harga terus trend tanpa kembali.
Dari perspektif kod logik, terdapat dua purata bergerak kemasukan - panjang dan pendek - yang terdiri daripada purata bergerak pantas dan perlahan masing-masing. Penyimpangan antara mereka dan harga menentukan saiz kedudukan. Di samping itu, purata bergerak keluar adalah purata bergerak berasingan yang menandakan kapan untuk menutup kedudukan. Apabila harga mencapai garis ini, kedudukan yang ada akan rata.
Kelebihan utama strategi pembalikan purata bergerak dua arah termasuk:
Strategi ini berfungsi dengan baik dengan instrumen turun naik yang rendah yang mempunyai turun naik harga yang agak kecil, terutamanya apabila memasuki kitaran terikat julat. Ia dapat menangkap peluang dengan berkesan dari pembalikan harga sementara. Sementara itu, langkah-langkah kawalan risiko agak komprehensif, menutup kerugian dalam julat yang munasabah walaupun harga tidak kembali.
Terdapat juga beberapa risiko yang berkaitan dengan strategi ini:
Beberapa cara untuk mengurangkan risiko di atas termasuk:
Terdapat juga ruang yang cukup untuk mengoptimumkan lagi strategi ini:
Strategi perdagangan pembalikan purata bergerak dua hala bertujuan untuk mendapat keuntungan daripada pembalikan harga selepas penyimpangan yang ketara dari paras purata bergerak. Dengan langkah kawalan risiko yang betul, ia dapat mencapai keuntungan yang konsisten melalui penyesuaian parameter. Walaupun risiko seperti mengejar trend dan turun naik yang berlebihan masih wujud, mereka boleh ditangani dengan meningkatkan logik kemasukan, mengurangkan saiz kedudukan dan banyak lagi. Strategi yang mudah difahami ini layak penyelidikan dan pengoptimuman lanjut dari peniaga kuantitatif.
/*backtest start: 2023-12-15 00:00:00 end: 2024-01-14 00:00:00 period: 1h basePeriod: 15m exchanges: [{"eid":"Futures_Binance","currency":"BTC_USDT"}] */ //@version=5 strategy(title = "hamster-bot MRS 2", overlay = true, default_qty_type = strategy.percent_of_equity, initial_capital = 100, default_qty_value = 30, pyramiding = 1, commission_value = 0.1, backtest_fill_limits_assumption = 1) info_options = "Options" on_close = input(false, title = "Entry on close", inline=info_options, group=info_options) OFFS = input.int(0, minval = 0, maxval = 1, title = "| Offset View", inline=info_options, group=info_options) trade_offset = input.int(0, minval = 0, maxval = 1, title = "Trade", inline=info_options, group=info_options) use_kalman_filter = input.bool(false, title="Use Kalman filter", group=info_options) //MA Opening info_opening = "MA Opening Long" maopeningtyp_l = input.string("SMA", title="Type", options=["SMA", "EMA", "TEMA", "DEMA", "ZLEMA", "WMA", "Hma", "Thma", "Ehma", "H", "L", "DMA"], title = "", inline=info_opening, group=info_opening) maopeningsrc_l = input.source(ohlc4, title = "", inline=info_opening, group=info_opening) maopeninglen_l = input.int(3, minval = 1, title = "", inline=info_opening, group=info_opening) long1on = input(true, title = "", inline = "long1") long1shift = input.float(0.96, step = 0.005, title = "Long", inline = "long1") long1lot = input.int(10, minval = 0, maxval = 10000, step = 10, title = "Lot 1", inline = "long1") info_opening_s = "MA Opening Short" maopeningtyp_s = input.string("SMA", title="Type", options=["SMA", "EMA", "TEMA", "DEMA", "ZLEMA", "WMA", "Hma", "Thma", "Ehma", "H", "L", "DMA"], title = "", inline=info_opening_s, group=info_opening_s) maopeningsrc_s = input.source(ohlc4, title = "", inline=info_opening_s, group=info_opening_s) maopeninglen_s = input.int(3, minval = 1, title = "", inline=info_opening_s, group=info_opening_s) short1on = input(true, title = "", inline = "short1") short1shift = input.float(1.04, step = 0.005, title = "short", inline = "short1") short1lot = input.int(10, minval = 0, maxval = 10000, step = 10, title = "Lot 1", inline = "short1") //MA Closing info_closing = "MA Closing" maclosingtyp = input.string("SMA", title="Type", options=["SMA", "EMA", "TEMA", "DEMA", "ZLEMA", "WMA", "Hma", "Thma", "Ehma", "H", "L", "DMA"], title = "", inline=info_closing, group=info_closing) maclosingsrc = input.source(ohlc4, title = "", inline=info_closing, group=info_closing) maclosinglen = input.int(3, minval = 1, maxval = 200, title = "", inline=info_closing, group=info_closing) maclosingmul = input.float(1, step = 0.005, title = "mul", inline=info_closing, group=info_closing) startTime = input(timestamp("01 Jan 2010 00:00 +0000"), "Start date", inline = "period") finalTime = input(timestamp("31 Dec 2030 23:59 +0000"), "Final date", inline = "period") HMA(_src, _length) => ta.wma(2 * ta.wma(_src, _length / 2) - ta.wma(_src, _length), math.round(math.sqrt(_length))) EHMA(_src, _length) => ta.ema(2 * ta.ema(_src, _length / 2) - ta.ema(_src, _length), math.round(math.sqrt(_length))) THMA(_src, _length) => ta.wma(ta.wma(_src,_length / 3) * 3 - ta.wma(_src, _length / 2) - ta.wma(_src, _length), _length) tema(sec, length)=> tema1= ta.ema(sec, length) tema2= ta.ema(tema1, length) tema3= ta.ema(tema2, length) tema_r = 3*tema1-3*tema2+tema3 donchian(len) => math.avg(ta.lowest(len), ta.highest(len)) ATR_func(_src, _len)=> atrLow = low - ta.atr(_len) trailAtrLow = atrLow trailAtrLow := na(trailAtrLow[1]) ? trailAtrLow : atrLow >= trailAtrLow[1] ? atrLow : trailAtrLow[1] supportHit = _src <= trailAtrLow trailAtrLow := supportHit ? atrLow : trailAtrLow trailAtrLow f_dema(src, len)=> EMA1 = ta.ema(src, len) EMA2 = ta.ema(EMA1, len) DEMA = (2*EMA1)-EMA2 f_zlema(src, period) => lag = math.round((period - 1) / 2) ema_data = src + (src - src[lag]) zl= ta.ema(ema_data, period) f_kalman_filter(src) => float value1= na float value2 = na value1 := 0.2 * (src - src[1]) + 0.8 * nz(value1[1]) value2 := 0.1 * (ta.tr) + 0.8 * nz(value2[1]) lambda = math.abs(value1 / value2) alpha = (-math.pow(lambda, 2) + math.sqrt(math.pow(lambda, 4) + 16 * math.pow(lambda, 2)))/8 value3 = float(na) value3 := alpha * src + (1 - alpha) * nz(value3[1]) //SWITCH ma_func(modeSwitch, src, len, use_k_f=true) => modeSwitch == "SMA" ? use_kalman_filter and use_k_f ? f_kalman_filter(ta.sma(src, len)) : ta.sma(src, len) : modeSwitch == "RMA" ? use_kalman_filter and use_k_f ? f_kalman_filter(ta.rma(src, len)) : ta.rma(src, len) : modeSwitch == "EMA" ? use_kalman_filter and use_k_f ? f_kalman_filter(ta.ema(src, len)) : ta.ema(src, len) : modeSwitch == "TEMA" ? use_kalman_filter and use_k_f ? f_kalman_filter(tema(src, len)) : tema(src, len): modeSwitch == "DEMA" ? use_kalman_filter and use_k_f ? f_kalman_filter(f_dema(src, len)) : f_dema(src, len): modeSwitch == "ZLEMA" ? use_kalman_filter and use_k_f ? f_kalman_filter(f_zlema(src, len)) : f_zlema(src, len): modeSwitch == "WMA" ? use_kalman_filter and use_k_f ? f_kalman_filter(ta.wma(src, len)) : ta.wma(src, len): modeSwitch == "VWMA" ? use_kalman_filter and use_k_f ? f_kalman_filter(ta.vwma(src, len)) : ta.vwma(src, len): modeSwitch == "Hma" ? use_kalman_filter and use_k_f ? f_kalman_filter(HMA(src, len)) : HMA(src, len): modeSwitch == "Ehma" ? use_kalman_filter and use_k_f ? f_kalman_filter(EHMA(src, len)) : EHMA(src, len): modeSwitch == "Thma" ? use_kalman_filter and use_k_f ? f_kalman_filter(THMA(src, len/2)) : THMA(src, len/2): modeSwitch == "ATR" ? use_kalman_filter and use_k_f ? f_kalman_filter(ATR_func(src, len)): ATR_func(src, len) : modeSwitch == "L" ? use_kalman_filter and use_k_f ? f_kalman_filter(ta.lowest(len)): ta.lowest(len) : modeSwitch == "H" ? use_kalman_filter and use_k_f ? f_kalman_filter(ta.highest(len)): ta.highest(len) : modeSwitch == "DMA" ? donchian(len) : na //Var sum = 0.0 maopening_l = 0.0 maopening_s = 0.0 maclosing = 0.0 pos = strategy.position_size p = 0.0 p := pos == 0 ? (strategy.equity / 100) / close : p[1] truetime = true loss = 0.0 maxloss = 0.0 equity = 0.0 //MA Opening maopening_l := ma_func(maopeningtyp_l, maopeningsrc_l, maopeninglen_l) maopening_s := ma_func(maopeningtyp_s, maopeningsrc_s, maopeninglen_s) //MA Closing maclosing := ma_func(maclosingtyp, maclosingsrc, maclosinglen) * maclosingmul long1 = long1on == false ? 0 : long1shift == 0 ? 0 : long1lot == 0 ? 0 : maopening_l == 0 ? 0 : maopening_l * long1shift short1 = short1on == false ? 0 : short1shift == 0 ? 0 : short1lot == 0 ? 0 : maopening_s == 0 ? 0 : maopening_s * short1shift //Colors long1col = long1 == 0 ? na : color.green short1col = short1 == 0 ? na : color.red //Lines // plot(maopening_l, offset = OFFS, color = color.new(color.green, 50)) // plot(maopening_s, offset = OFFS, color = color.new(color.red, 50)) plot(maclosing, offset = OFFS, color = color.fuchsia) long1line = long1 == 0 ? close : long1 short1line = short1 == 0 ? close : short1 plot(long1line, offset = OFFS, color = long1col) plot(short1line, offset = OFFS, color = short1col) //Lots lotlong1 = p * long1lot lotshort1 = p * short1lot //Entry if truetime //Long sum := 0 strategy.entry("L", strategy.long, lotlong1, limit = on_close ? na : long1, when = long1 > 0 and pos <= sum and (on_close ? close <= long1[trade_offset] : true)) sum := lotlong1 //Short sum := 0 pos := -1 * pos strategy.entry("S", strategy.short, lotshort1, limit = on_close ? na : short1, when = short1 > 0 and pos <= sum and (on_close ? close >= short1[trade_offset] : true)) sum := lotshort1 strategy.exit("Exit", na, limit = maclosing) if time > finalTime strategy.close_all()