Global

Methods

caEma(vs, len, optopt) → {Array}

Description:

• 計算指數移動平均EMA(Exponential Moving Average)序列

  平滑係數alpha=2/(len+1)；種子ema0為輸入前len筆value之簡單算術平均，對應輸出第0筆，其time為vs[len-1].time 第0筆物件僅有{time,param}(無value/diff/ratio，為既有現象)；第1筆起(對應輸入索引len起)物件為{time,value,diff,ratio,param}，diff為ema減現值，ratio為diff除以現值(現值為0時給0) 輸入vs筆數不足len時回傳空陣列；len非數值時拋出例外

  Unit Test: Github

Source:

• caEma.mjs, line 39

Example

let vs = [
    { time: '2020-01-01T00:00:00', value: 1 },
    { time: '2020-01-01T04:00:00', value: 2 },
    { time: '2020-01-01T08:00:00', value: 3 },
    { time: '2020-01-01T12:00:00', value: 4 },
]

let rs = caEma(vs, 2)
console.log(rs)
// => [
//   { time: '2020-01-01T04:00:00', param: 1.5 },
//   { time: '2020-01-01T08:00:00', value: 2.5, diff: -0.5, ratio: -0.16666666666666666, param: 2.5 },
//   { time: '2020-01-01T12:00:00', value: 3.5, diff: -0.5, ratio: -0.125, param: 3.5 }
// ]

Parameters:

Returns:

Type

Array

calcAdx(arr, key, optopt) → {Promise}

Description:

• 計算各週期之ADX(Average Directional Index，動向指標，含+DI/-DI/DIcross)

  自輸入索引1起算+DM、-DM與TR(True Range)，經兩層Wilder's平滑後得+DI、-DI、DX與ADX 第二層平滑之ADX種子對應輸入索引2len-1起算，其後逐根遞推；資料筆數n小於2len+1時該期vs回空陣列 第一筆ADXslope因無前值可比故為0

  Unit Test: Github

Source:

• calcAdx.mjs, line 45

Example

let arr = [
    { time: '2020-01-01T00:00:00', Open: 100, High: 101, Low: 99, Close: 100 },
    { time: '2020-01-01T04:00:00', Open: 100, High: 103, Low: 99, Close: 102 },
    { time: '2020-01-01T08:00:00', Open: 102, High: 103, Low: 100, Close: 101 },
    { time: '2020-01-01T12:00:00', Open: 101, High: 106, Low: 100, Close: 105 },
    { time: '2020-01-01T16:00:00', Open: 105, High: 106, Low: 102, Close: 103 },
    { time: '2020-01-01T20:00:00', Open: 103, High: 108, Low: 102, Close: 107 },
    { time: '2020-01-02T00:00:00', Open: 107, High: 108, Low: 105, Close: 106 },
]

calcAdx(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [
        //     { time: '2020-01-01T20:00:00', ADX: 100, plusDI: 29.23076923076923, minusDI: 0, DIcross: 29.23076923076923, ADXslope: 0 },
        //     { time: '2020-01-02T00:00:00', ADX: 100, plusDI: 22.287390029325515, minusDI: 0, DIcross: 22.287390029325515, ADXslope: 0 }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcAtr(arr, key, optopt) → {Promise}

Description:

• 計算各週期之ATR(Average True Range，真實區間均值)

  TR(i)=max(|High-Low|,|High-前一筆Close|,|Low-前一筆Close|)，自輸入索引1起算 ATR種子取索引1至len之TR簡單平均，其後以Wilder's平滑法(ATR=(ATR前一筆*(len-1)+TR)/len)遞推，第一筆對應輸入索引len，資料筆數n小於len+1時該期vs回空陣列 atrRatio於close為0時回0；atrRatioMod以diviProt保護分母(close+opt.plusClose)避免除以趨近0之值；atrChange第一筆無前值故為0；trRank為當根TR於前len根中嚴格小於之比例

  Unit Test: Github

Source:

• calcAtr.mjs, line 46

Example

let arr = [
    { time: '2020-01-01T00:00:00', Open: 100, High: 101, Low: 99, Close: 100 },
    { time: '2020-01-01T04:00:00', Open: 100, High: 103, Low: 99, Close: 102 },
    { time: '2020-01-01T08:00:00', Open: 102, High: 103, Low: 100, Close: 101 },
    { time: '2020-01-01T12:00:00', Open: 101, High: 106, Low: 100, Close: 105 },
    { time: '2020-01-01T16:00:00', Open: 105, High: 106, Low: 102, Close: 103 },
]

calcAtr(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [
        //     { time: '2020-01-01T12:00:00', atr: 4.333333333333333, atrRatio: 0.04126984126984127, atrRatioMod: 0.04126984126984127, atrChange: 0, trRank: 0.6666666666666666 },
        //     { time: '2020-01-01T16:00:00', atr: 4.222222222222222, atrRatio: 0.040992448759439054, atrRatioMod: 0.040992448759439054, atrChange: -0.025641025641025553, trRank: 0.3333333333333333 }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcAts(arr, key, optopt) → {Promise}

Description:

• 計算各週期之ATS(Average Trade Size，平均單筆成交額)

  ats(i)=NumberOfTrades(i)>0 ? QuoteAssetVolume(i)/NumberOfTrades(i) : 0 atsRatio為當根ats相對最近len根(含當根)簡單平均之倍數，baseline為0時回1；atsChange為當根ats相對前一根之變化率，前一根ats為0時回0 第一筆對應輸入索引len-1，資料筆數n小於len+1時該期vs回空陣列

  Unit Test: Github

Source:

• calcAts.mjs, line 42

Example

let arr = [
    { time: '2020-01-01T00:00:00', QuoteAssetVolume: 33441599, NumberOfTrades: 46046 },
    { time: '2020-01-01T04:00:00', QuoteAssetVolume: 15032938, NumberOfTrades: 29738 },
    { time: '2020-01-01T08:00:00', QuoteAssetVolume: 20445895, NumberOfTrades: 32476 },
    { time: '2020-01-01T12:00:00', QuoteAssetVolume: 14890182, NumberOfTrades: 29991 },
]

calcAts(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [
        //     { time: '2020-01-01T08:00:00', atsRatio: 1.0146995613283087, atsChange: 0.2454075213472537 },
        //     { time: '2020-01-01T12:00:00', atsRatio: 0.9129026732554982, atsChange: -0.21138421472562993 }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcBbpb(arr, key, optopt) → {Promise}

Description:

• 計算各週期之BBpctB(布林通道%B)與BBwidth(布林通道寬度)

  Middle=SMA(Close,len)，Sigma為母體標準差，Upper/Lower=Middle±opt.k*Sigma(opt.k預設2) BBpctB=(Close-Lower)/(Upper-Lower)，通道寬度為0時回0.5；BBwidth=(Upper-Lower)/Middle，Middle為0時回0；BBwidthChange為與前一筆BBwidth之差，首筆回0 BBwidthMod、BBwidthChangeMod為以diviProt保護分母之修正版本，分母為(Middle-opt.subMid)取opt.powMid次方後加opt.plusMid，分子為通道寬度加opt.plusBandWidth，供分母易趨近0之情境(如以指數OHLC計算)使用 第一筆對應輸入索引len-1，資料筆數n小於len時該期vs回空陣列

  Unit Test: Github

Source:

• calcBbpb.mjs, line 50

Example

let arr = [
    { time: '2020-01-01T00:00:00', Close: 100 },
    { time: '2020-01-01T04:00:00', Close: 102 },
    { time: '2020-01-01T08:00:00', Close: 101 },
    { time: '2020-01-01T12:00:00', Close: 105 },
]

calcBbpb(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [
        //     { time: '2020-01-01T08:00:00', BBpctB: 0.5, BBwidth: 0.03233649825454878, BBwidthChange: 0, BBwidthMod: 0.03233649825454878, BBwidthChangeMod: 0 },
        //     { time: '2020-01-01T12:00:00', BBpctB: 0.8432032364918327, BBwidth: 0.06622103264405994, BBwidthChange: 0.03388453438951116, BBwidthMod: 0.06622103264405994, BBwidthChangeMod: 0.03388453438951116 }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcCandle(arr, key, optopt) → {Promise}

Description:

• 計算各根K線之型態特徵量(Candle Pattern，含實體、影線比例與吞噬強度)

  本指標僅反映當下該根K線自身形狀，無週期累積概念，kp僅含'4hr':1一組 bodyRatio、upperWickRatio、lowerWickRatio、candleDir分別為實體、上影線、下影線佔全根範圍(High-Low)之比例與收開差方向強度，範圍(High-Low)為0時四值皆回0 engulfStrength僅於當根與前一根方向相反且前一根有範圍時計算，為方向號誌乘上當根實體與前一根範圍之比值，否則回0 因engulfStrength需前一根資料，第一筆對應輸入第2根(index 1)，資料筆數n小於2時回空陣列

  Unit Test: Github

Source:

• calcCandle.mjs, line 42

Example

let arr = [
    { time: '2020-01-01T00:00:00', Open: 10, High: 12, Low: 9, Close: 11 },
    { time: '2020-01-01T04:00:00', Open: 11, High: 13, Low: 10, Close: 12 },
    { time: '2020-01-01T08:00:00', Open: 12, High: 12.5, Low: 9, Close: 9.5 },
]

calcCandle(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '4hr',
        //   len: 1,
        //   vs: [
        //     { time: '2020-01-01T04:00:00', bodyRatio: 0.3333333333333333, upperWickRatio: 0.3333333333333333, lowerWickRatio: 0.3333333333333333, candleDir: 0.3333333333333333, engulfStrength: 0 },
        //     { time: '2020-01-01T08:00:00', bodyRatio: 0.7142857142857143, upperWickRatio: 0.14285714285714285, lowerWickRatio: 0.14285714285714285, candleDir: -0.7142857142857143, engulfStrength: -0.8333333333333334 }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcCvd(arr, key, optopt) → {Promise}

Description:

• 計算各週期之CVD(Cumulative Volume Delta，累積量能差)

  delta(i)=(2*TakerBuyBaseAssetVolume(i)-Volumn(i))/Volumn(i)，Volumn(i)小於等於0時delta回0 cvd為視窗[i-len+1,i]內delta加總後除以√len之歸一化累積值，第一筆對應輸入索引len-1 cvdSlope=(cvd(i)-cvd(i-3))/3，固定回看3根(對應12hr)以呈現累積動能之變化速度 資料筆數n小於len+3時該期vs回空陣列

  Unit Test: Github

Source:

• calcCvd.mjs, line 46

Example

let arr = [
    { time: '2020-01-01T00:00:00', Volumn: 1000, TakerBuyBaseAssetVolume: 600 },
    { time: '2020-01-01T04:00:00', Volumn: 1200, TakerBuyBaseAssetVolume: 500 },
    { time: '2020-01-01T08:00:00', Volumn: 900, TakerBuyBaseAssetVolume: 400 },
    { time: '2020-01-01T12:00:00', Volumn: 1100, TakerBuyBaseAssetVolume: 700 },
    { time: '2020-01-01T16:00:00', Volumn: 1300, TakerBuyBaseAssetVolume: 800 },
    { time: '2020-01-01T20:00:00', Volumn: 1000, TakerBuyBaseAssetVolume: 450 },
    { time: '2020-01-02T00:00:00', Volumn: 1050, TakerBuyBaseAssetVolume: 600 },
]

calcCvd(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [
        //     { time: '2020-01-01T20:00:00', cvd: 0.23295881491077913, cvdSlope: 0.09262127861591667 },
        //     { time: '2020-01-02T00:00:00', cvd: 0.15797826047056798, cvdSlope: 0.053631390307006886 }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcDonchian(arr, key, optopt) → {Promise}

Description:

• 計算各週期之唐奇安通道DC(Donchian Channel)

  各週期len為該期涵蓋之K線根數(以4hr K線為基準，1day=6根)，資料筆數不足該期len時，該期vs為空陣列 opt.excludeCurrent=false(預設)時，視窗為[i-len+1,i]含當前根，第1筆對應輸入之第len根(index len-1)； 為true時視窗為[i-len,i-1]不含當前根，第1筆對應輸入之第len+1根(index len) 其後皆逐根對齊至最後一根

  Unit Test: Github

Source:

• calcDonchian.mjs, line 52

Example

let arr = [
    { time: '2020-01-01T00:00:00', Open: 101, High: 101.5, Low: 100.5, Close: 101 },
    { time: '2020-01-01T04:00:00', Open: 101, High: 105.17, Low: 99.28, Close: 103.45 },
    { time: '2020-01-01T08:00:00', Open: 103.45, High: 108.74, Low: 101.35, Close: 106.64 },
]

calcDonchian(arr, 'Close')
    .then((rs) => {
        let r = rs.find((v) => v.period === '12hr')
        console.log(r.vs)
        // => [
        //   {
        //     time: '2020-01-01T08:00:00',
        //     dcPctB: 0.7780126849894297,
        //     dcWidth: 0.09095279300067297,
        //     dcWidthChange: 0,
        //     dcWidthMod: 0.09095279300067297,
        //     dcWidthChangeMod: 0
        //   }
        // ]
    })

Parameters:

Returns:

Type

Promise

calcEma(arr, key, optopt) → {Promise}

Description:

• 計算各週期之指數移動平均EMA(Exponential Moving Average)

  各週期len為該期涵蓋之K線根數(以4hr K線為基準，1day=6根)，資料筆數不足該期len時，該期vs為空陣列 第1筆為種子值(取前len筆之算術平均SMA)，對應輸入之第len根，物件僅含{time,param} 其後各筆依EMA公式(alpha=2/(len+1))遞推，逐根對齊至最後一根，物件另含{value,diff,ratio} diff=EMA-現值，ratio=diff/現值(現值為0時ratio給0)

  Unit Test: Github

Source:

• calcEma.mjs, line 39

Example

let arr = [
    { time: '2020-01-01T00:00:00', Open: 101, High: 101.5, Low: 100.5, Close: 101 },
    { time: '2020-01-01T04:00:00', Open: 101, High: 105.17, Low: 99.28, Close: 103.45 },
    { time: '2020-01-01T08:00:00', Open: 103.45, High: 108.74, Low: 101.35, Close: 106.64 },
    { time: '2020-01-01T12:00:00', Open: 106.64, High: 111.54, Low: 104.68, Close: 109.58 },
    { time: '2020-01-01T16:00:00', Open: 109.58, High: 112.76, Low: 108.18, Close: 111.37 },
    { time: '2020-01-01T20:00:00', Open: 111.37, High: 112.3, Low: 110.72, Close: 111.65 },
]

calcEma(arr, 'Close')
    .then((rs) => {
        console.log(rs[0].vs)
        // => [ { time: '2020-01-01T20:00:00', param: 107.28166666666665 } ]
    })

Parameters:

Returns:

Type

Promise

calcIchimoku(arr, key, optopt) → {Promise}

Description:

• 計算各週期之一目均衡表Ichimoku Cloud(Ichimoku Kinko Hyo)

  三層lookback依經典9:26:52比例縮放，各期(period)對應之kp設定為{tenkan,kijun,senkouB}三參數配對(單位皆為以4hr K線為基準之根數)， period標籤本身即對應senkouB(最長lookback)所涵蓋之天數，僅保留4day~30day四期(短於4day時tenkan會<4而失去意義)： '4day': { tenkan: 4, kijun: 12, senkouB: 24 } '7day': { tenkan: 7, kijun: 21, senkouB: 42 } '15day': { tenkan: 16, kijun: 45, senkouB: 90 } '30day': { tenkan: 31, kijun: 90, senkouB: 180 } Tenkan=(視窗內最高High+最低Low)/2(視窗為最近tenkan根)，Kijun同理(視窗為最近kijun根) SenkouA[i]=(Tenkan[i-kijun]+Kijun[i-kijun])/2，SenkouB[i]=(i-kijun期回看senkouB根之最高High+最低Low)/2 cloudTop=max(SenkouA,SenkouB)，cloudBot=min(SenkouA,SenkouB)，cloudMid=(cloudTop+cloudBot)/2 第1筆對應輸入索引startIdx=max(2*kijun-1, kijun+senkouB-1)，資料筆數不足startIdx+1時，該期vs為空陣列 不輸出Chikou(遲行線)，因其原版定義為Close[i+26]之look-ahead資料，回測不可用

  Unit Test: Github

Source:

• calcIchimoku.mjs, line 71

Example

//最短之4day期需24根senkouB且再回看kijun根，故至少需36根K線，此處給40根
let arr = []
for (let i = 0; i < 40; i++) {
    let c = 100 + Math.sin(i * 0.7) * 5 + Math.cos(i * 0.31) * 3
    arr.push({
        time: new Date(Date.UTC(2020, 0, 1) + i * 4 * 3600 * 1000).toISOString().slice(0, 19),
        High: c + 1,
        Low: c - 1,
        Close: c,
    })
}

calcIchimoku(arr, 'Close')
    .then((rs) => {
        console.log(rs.map((v) => `${v.period}(len=${v.len}): ${v.vs.length} 筆`))
        // => [
        //   '4day(len=24): 5 筆',
        //   '7day(len=42): 0 筆',
        //   '15day(len=90): 0 筆',
        //   '30day(len=180): 0 筆'
        // ]

        let p4day = rs.find((v) => v.period === '4day')
        console.log(p4day.vs[p4day.vs.length - 1])
        // => {
        //   time: '2020-01-07T12:00:00',
        //   cloudDist: 0.06339165670785038,
        //   cloudThick: 0.011928239637035617,
        //   tkDiff: 0.034558548904317685,
        //   cloudDistMod: 0.06339165670785038,
        //   cloudThickMod: 0.011928239637035617,
        //   tkDiffMod: 0.034558548904317685
        // }
    })

Parameters:

Returns:

Type

Promise

calcKdj(arr, key, optopt) → {Promise}

Description:

• 計算各週期之KDJ隨機指標(Stochastic Oscillator KDJ)

  各週期len即RSV回看期數(以4hr K線為基準，1day=6根)，資料筆數不足該期len時，該期vs為空陣列 K/D以EMA平滑遞推(kAlpha=dAlpha=1/3)，K與D初始值(seed)皆為50 RSV=(Close-Ln)/(Hn-Ln)*100，Hn/Ln為視窗內(最近len根)之最高High/最低Low，Hn===Ln時RSV取中性值50(避免除以0) 輸出自輸入索引len-1起(視窗未滿之根不輸出)，其後逐根對齊至最後一根 opt目前未被讀取(本函式尚未提供可調整之設定項)

  Unit Test: Github

Source:

• calcKdj.mjs, line 48

Example

let arr = [
    { time: '2020-01-01T00:00:00', Open: 101, High: 101.5, Low: 100.5, Close: 101 },
    { time: '2020-01-01T04:00:00', Open: 101, High: 105.17, Low: 99.28, Close: 103.45 },
    { time: '2020-01-01T08:00:00', Open: 103.45, High: 108.74, Low: 101.35, Close: 106.64 },
    { time: '2020-01-01T12:00:00', Open: 106.64, High: 111.54, Low: 104.68, Close: 109.58 },
    { time: '2020-01-01T16:00:00', Open: 109.58, High: 112.76, Low: 108.18, Close: 111.37 },
    { time: '2020-01-01T20:00:00', Open: 111.37, High: 112.3, Low: 110.72, Close: 111.65 },
]

calcKdj(arr, 'Close')
    .then((rs) => {
        console.log(rs[0].vs)
        // => [
        //   {
        //     time: '2020-01-01T20:00:00',
        //     K: 63.92185954500495,
        //     D: 54.640619848334985,
        //     J: 82.48433893834486,
        //     KminusD: 9.281239696669964
        //   }
        // ]
    })

Parameters:

Returns:

Type

Promise

calcKeltner(arr, key, optopt) → {Promise}

Description:

• 計算各週期之肯特納通道Keltner Channel

  各期(period)對應之kp設定為{len,mult}參數配對(len單位為以4hr K線為基準之根數)， len為EMA(中線)與ATR(通道厚度基準)共用之平滑期數，mult為ATR通道厚度倍數，錨點1day=(len:6,mult:2)對應Linda Raschke經典(20,10,2)標準： '12hr': { len: 3, mult: 1.5 } '16hr': { len: 4, mult: 1.5 } '20hr': { len: 5, mult: 1.5 } '1day': { len: 6, mult: 2 } '2day': { len: 12, mult: 2 } '4day': { len: 24, mult: 2 } '7day': { len: 42, mult: 2.5 } '15day': { len: 90, mult: 2.5 } '30day': { len: 180, mult: 3 } middle=EMA(Close,len)，ATR以Wilder平滑遞推(種子為前len筆TR之算術平均SMA)，upper=middle+multATR，lower=middle-multATR 第1筆對應輸入索引len(此時EMA與ATR皆已就緒)，資料筆數不足len+1時，該期vs為空陣列，其後逐根對齊至最後一根

  Unit Test: Github

Source:

• calcKeltner.mjs, line 56

Example

let arr = [
    { time: '2020-01-01T00:00:00', Open: 101, High: 101.5, Low: 100.5, Close: 101 },
    { time: '2020-01-01T04:00:00', Open: 101, High: 105.17, Low: 99.28, Close: 103.45 },
    { time: '2020-01-01T08:00:00', Open: 103.45, High: 108.74, Low: 101.35, Close: 106.64 },
    { time: '2020-01-01T12:00:00', Open: 106.64, High: 111.54, Low: 104.68, Close: 109.58 },
]

calcKeltner(arr, 'Close')
    .then((rs) => {
        let r = rs.find((v) => v.period === '12hr')
        console.log(r.vs)
        // => [
        //   {
        //     time: '2020-01-01T12:00:00',
        //     kcPctB: 0.6460609069844429,
        //     kcWidth: 0.18886266664582768,
        //     kcWidthMod: 0.18886266664582768
        //   }
        // ]
    })

Parameters:

Returns:

Type

Promise

calcKlr(arr, key, optopt) → {Promise}

Description:

• 計算各週期之K線OHLC比值(K-Line OHLC Ratio)

  僅一期'4hr'(len固定為1)，逐根獨立計算同根Open/High/Low/Close間之六種比值，不需回看歷史根數， 輸出自輸入索引0起，逐根對齊至最後一根，資料筆數不足len(即輸入為空陣列)時，該期vs為空陣列 ho=H/O，lo=L/O，co=C/O，hc=H/C，lc=L/C，hl=H/L，六比值皆先將O/H/L/C同步加上opt.plusClose後才經diviProt計算(分母保護：|分母|<0.00001時clamp至±0.00001，保留符號)

  Unit Test: Github

Source:

• calcKlr.mjs, line 45

Example

let arr = [
    { time: '2020-01-01T00:00:00', Open: 101, High: 101.5, Low: 100.5, Close: 101 },
]

calcKlr(arr, 'Close')
    .then((rs) => {
        console.log(rs[0].vs)
        // => [
        //   {
        //     time: '2020-01-01T00:00:00',
        //     ho: 1.004950495049505,
        //     lo: 0.995049504950495,
        //     co: 1,
        //     hc: 1.004950495049505,
        //     lc: 0.995049504950495,
        //     hl: 1.0099502487562189
        //   }
        // ]
    })

Parameters:

Returns:

Type

Promise

calcMa(arr, key, optopt) → {Promise}

Description:

• 計算各週期之移動平均MA(Moving Average)

  各週期len為該期涵蓋之K線根數(以4hr K線為基準，1day=6根)，資料筆數不足該期len時，該期vs為空陣列 各期第1筆對應輸入之第len根，其後逐根對齊至最後一根

  Unit Test: Github

Source:

• calcMa.mjs, line 45

Example

let arr = [
    { time: '2020-01-01T00:00:00', Close: 7173.32 },
    { time: '2020-01-01T04:00:00', Close: 7195.23 },
    { time: '2020-01-01T08:00:00', Close: 7225.01 },
    { time: '2020-01-01T12:00:00', Close: 7209.83 },
    { time: '2020-01-01T16:00:00', Close: 7188.77 },
    { time: '2020-01-01T20:00:00', Close: 7200.85 },
]

calcMa(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '1day',
        //   len: 6,
        //   vs: [ { time: '2020-01-01T20:00:00', param: 7198.835 } ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcMacd(arr, key, optopt) → {Promise}

Description:

• 計算各週期之指數平滑異同移動平均線MACD(Moving Average Convergence Divergence)

  各週期len為基準倍率(以4hr K線為基準)，實際採用快線期數fastPeriod=len12、慢線期數slowPeriod=len26、訊號線期數signalPeriod=len9(即標準MACD(12,26,9)之等比放大) DIF(快線)為收盤價快線EMA減慢線EMA，DEM(慢線，即訊號線)為DIF之EMA，OSC(柱狀動能)為DIF減DEM 資料筆數n小於len或小於slowPeriod+signalPeriod-1時，該期vs為空陣列 各期第1筆對應輸入之第35len-2根(即索引(slowPeriod-1)+(signalPeriod-1))，其後逐根對齊至最後一根

  Unit Test: Github

Source:

• calcMacd.mjs, line 45

Example

//收盤價1至34之等差序列，time自2020-01-01T00:00:00起每根間隔4小時
let arr = []
for (let i = 0; i < 34; i++) {
    let d = new Date(Date.UTC(2020, 0, 1) + i * 4 * 3600 * 1000)
    arr.push({ time: d.toISOString().slice(0, 19), Close: i + 1 })
}

calcMacd(arr, 'Close')
    .then((rs) => {
        let r = rs.find((v) => v.period === '4hr')
        console.log(r)
        // => {
        //   period: '4hr',
        //   len: 1,
        //   vs: [ { time: '2020-01-06T12:00:00', DIF: 7, DEM: 7, OSC: 0 } ]
        // }
        //其餘各期(8hr,12hr,16hr,20hr,1day)因資料筆數未達門檻(35*len-2)，vs皆為[]
    })

Parameters:

Returns:

Type

Promise

calcMfi(arr, key, optopt) → {Promise}

Description:

• 計算各週期之資金流量指標MFI(Money Flow Index)

  各週期len為該期涵蓋之K線根數(以4hr K線為基準，1day=6根)，資料筆數n小於len+1時，該期vs為空陣列 典型價tp=(High+Low+Close)/3，原始資金流rawMF=tp*Volumn；tp較前一根上升時計入正向資金流posMF，下降時計入負向資金流negMF，持平則兩者皆不計入 MFI=100-100/(1+視窗內sumPosMF/sumNegMF之比值)；sumPosMF與sumNegMF皆小於等於opt.eps時給50(完全平盤)，僅sumNegMF小於等於opt.eps時給100(全部正向)，僅sumPosMF小於等於opt.eps時給0(全部負向) MFIratio=sumPosMF/(sumPosMF+sumNegMF)，兩者總和小於等於opt.eps時給0.5 MFIdiv為MFI與len根前之MFI取正負號比較，與Close之len根漲跌方向相同給1，相反給-1，任一方向為0或無歷史值則給0 各期第1筆對應輸入之第len根，其後逐根對齊至最後一根

  Unit Test: Github

Source:

• calcMfi.mjs, line 44

Example

let arr = [
    { time: '2020-01-01T00:00:00', High: 10, Low: 10, Close: 10, Volumn: 100 },
    { time: '2020-01-01T04:00:00', High: 12, Low: 12, Close: 12, Volumn: 100 },
    { time: '2020-01-01T08:00:00', High: 12, Low: 12, Close: 12, Volumn: 100 },
    { time: '2020-01-01T12:00:00', High: 9, Low: 9, Close: 9, Volumn: 100 },
]

calcMfi(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [ { time: '2020-01-01T12:00:00', MFI: 57.14285714285714, MFIratio: 0.5714285714285714, MFIdiv: 0 } ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcObv(arr, key, optopt) → {Promise}

Description:

• 計算各週期之能量潮OBV(On Balance Volume)

  各週期len為該期涵蓋之K線根數(以4hr K線為基準，1day=6根)，資料筆數n小於len+1時，該期vs為空陣列 累積OBV：Close較前一根上升時obv累加該根Volumn，下降時obv累減該根Volumn，持平則obv不變，obv[0]=0 OBVnorm=(obv-視窗len根obv之算術平均OBV_MA)/|OBV_MA|，OBV_MA=0時給0 OBVslope=(obv-len根前之obv)/(len*|len根前之obv|)，len根前之obv為0時分母改為len OBVdiv為obv與len根前之obv取正負號比較，與Close之len根漲跌方向相同給1，相反給-1，任一方向為0則給0 各期第1筆對應輸入之第len根，其後逐根對齊至最後一根

  Unit Test: Github

Source:

• calcObv.mjs, line 42

Example

let arr = [
    { time: '2020-01-01T00:00:00', Close: 10, Volumn: 1 },
    { time: '2020-01-01T04:00:00', Close: 12, Volumn: 2 },
    { time: '2020-01-01T08:00:00', Close: 11, Volumn: 3 },
    { time: '2020-01-01T12:00:00', Close: 13, Volumn: 4 },
]

calcObv(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [ { time: '2020-01-01T12:00:00', OBVnorm: 1.2500000000000002, OBVslope: 1, OBVdiv: 1 } ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcOrig(arr, key) → {Promise}

Description:

• 計算各週期之原始數值Orig(Original Value，逐筆直接複製輸入數值，不做任何運算)

  僅逐筆對應輸入，不做視窗/累積運算，亦不篩選或檢查數值合法性；資料筆數與輸入相同，缺key欄位或非數值時param為undefined或原樣值，該筆仍保留 固定回傳單一區塊，period為'none'，len為0

  Unit Test: Github

Source:

• calcOrig.mjs, line 38

Example

let arr = [
    { time: '2020-01-01T00:00:00', Close: 7173.32 },
    { time: '2020-01-01T04:00:00', Close: 7195.23 },
    { time: '2020-01-01T08:00:00', Close: 7225.01 },
]

calcOrig(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: 'none',
        //   len: 0,
        //   vs: [
        //     { time: '2020-01-01T00:00:00', param: 7173.32 },
        //     { time: '2020-01-01T04:00:00', param: 7195.23 },
        //     { time: '2020-01-01T08:00:00', param: 7225.01 }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcRsi(arr, key, optopt) → {Promise}

Description:

• 計算各週期之相對強弱指標RSI(Relative Strength Index，採Wilder平滑法)

  各週期len為該期涵蓋之K線根數(以4hr K線為基準，1day=6根)，資料筆數n小於len+1時，該期vs為空陣列 每根漲跌diff=Close-前一根Close，漲幅up=max(diff,0)，跌幅down=max(-diff,0) 種子(opt.seed='sma'，預設)：avgGain為第1至len根up之算術平均，avgLoss為第1至len根down之算術平均 種子(opt.seed='first')：avgGain、avgLoss分別取第1根之up、down，再以Wilder遞推方式(avg=(avg*(len-1)+x)/len)推進至第len根 其後每根皆以avg=(avg*(len-1)+x)/len遞推；RS=avgGain/avgLoss，RSI=100-100/(1+RS) avgGain與avgLoss皆小於等於opt.eps時RSI給50(完全平盤)，僅avgLoss小於等於opt.eps時給100(一路上漲)，僅avgGain小於等於opt.eps時給0(一路下跌) 各期第1筆對應輸入之第len根，其後逐根對齊至最後一根

  Unit Test: Github

Source:

• calcRsi.mjs, line 46

Example

let arr = [
    { time: '2020-01-01T00:00:00', Close: 10 },
    { time: '2020-01-01T04:00:00', Close: 11 },
    { time: '2020-01-01T08:00:00', Close: 10.5 },
    { time: '2020-01-01T12:00:00', Close: 12 },
]

calcRsi(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [ { time: '2020-01-01T12:00:00', RSI: 83.33333333333334, rsiAvgGain: 0.8333333333333334, rsiAvgLoss: 0.16666666666666666, rsiUp: 1.5, rsiDown: 0 } ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcSupertrend(arr, key, optopt) → {Promise}

Description:

• 計算各週期之Supertrend(Olivier Seban ATR通道趨勢指標)

  各週期kp值為{len,mult}物件，len為ATR平滑期數(以4hr K線為基準之根數)，mult為ATR通道厚度倍數；資料筆數n小於len+1時，該期vs為空陣列 TR(真實區間)=max(|High-Low|,|High-前一根Close|,|Low-前一根Close|)，ATR種子為第1至len根TR之算術平均，其後以Wilder平滑遞推(atr=(atr*(len-1)+TR)/len) basicUpper=(High+Low)/2+multATR，basicLower=(High+Low)/2-multATR finalUpper僅在basicUpper較前值小或前一根Close高於前值時更新為basicUpper，否則鎖死沿用前值；finalLower僅在basicLower較前值大或前一根Close低於前值時更新為basicLower，否則鎖死沿用前值 方向dir沿續前一根，多頭且Close跌破finalLower時翻空，空頭且Close突破finalUpper時翻多；首根依Close是否小於等於finalUpper判定初始方向 ST線line：多頭取finalLower，空頭取finalUpper stDistance=(Close-line)/Close，Close=0時給0 stDistanceMod=diviProt(Close+opt.plusClose-line, Close+opt.plusClose)，opt.plusClose=0時退化為stDistance；diviProt對分母絕對值小於0.00001時會鎖定為±0.00001(保留正負號)以避免除以0爆炸 各期第1筆對應輸入之第len根，其後逐根對齊至最後一根

  Unit Test: Github

Source:

• calcSupertrend.mjs, line 48

Example

let arr = [
    { time: '2020-01-01T00:00:00', Open: 10, High: 10, Low: 10, Close: 10 },
    { time: '2020-01-01T04:00:00', Open: 10, High: 12, Low: 10, Close: 12 },
    { time: '2020-01-01T08:00:00', Open: 12, High: 12, Low: 11, Close: 11 },
    { time: '2020-01-01T12:00:00', Open: 11, High: 13, Low: 11, Close: 13 },
]

calcSupertrend(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [ { time: '2020-01-01T12:00:00', stDistance: -0.17948717948717954, stDistanceMod: -0.17948717948717954 } ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcVec(arr, key, optopt) → {Promise}

Description:

• 計算各週期之差值向量Vec(後一值減前一值，非真實速度)

  各週期len為該期涵蓋之K線根數(以4hr K線為基準，1day=6根)，資料筆數不足該期len時，該期vs為空陣列 各期第1筆對應輸入之第len根(非len-1，因需與len根前之值相減)，其後逐根對齊至最後一根

  Unit Test: Github

Source:

• calcVec.mjs, line 49

Example

let arr = [
    { time: '2020-01-01T00:00:00', Close: 1 },
    { time: '2020-01-01T04:00:00', Close: 2 },
    { time: '2020-01-01T08:00:00', Close: 3 },
    { time: '2020-01-01T12:00:00', Close: 4 },
    { time: '2020-01-01T16:00:00', Close: 5 },
    { time: '2020-01-01T20:00:00', Close: 6 },
]

calcVec(arr, 'Close')
    .then((rs) => {
        console.log(rs[0])
        // => {
        //   period: '4hr',
        //   len: 1,
        //   vs: [
        //     { time: '2020-01-01T04:00:00', param: 1 },
        //     { time: '2020-01-01T08:00:00', param: 1 },
        //     { time: '2020-01-01T12:00:00', param: 1 },
        //     { time: '2020-01-01T16:00:00', param: 1 },
        //     { time: '2020-01-01T20:00:00', param: 1 }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcVortex(arr, key, optopt) → {Promise}

Description:

• 計算各週期之Vortex Indicator(VI+/VI-，Etienne Botes與Douglas Siepman原始定義)

  各週期len為該期涵蓋之K線根數(以4hr K線為基準，1day=6根)，直接讀取輸入arr內High、Low、Close欄位計算，資料筆數不足len+1根時，該期vs為空陣列 各期第1筆對應輸入之第len根(因每根計算VM+/VM-/TR皆需前一根資料，故自輸入第1根起算，累積滿len根視窗後才有第1筆輸出)，其後逐根對齊至最後一根 sum(TR)為0(視窗內無波動)時，VI+與VI-皆直接給0以避免除以0

  Unit Test: Github

Source:

• calcVortex.mjs, line 48

Example

let arr = [
    { time: '2020-01-01T00:00:00', High: 105, Low: 95, Close: 100 },
    { time: '2020-01-01T04:00:00', High: 110, Low: 100, Close: 108 },
    { time: '2020-01-01T08:00:00', High: 112, Low: 104, Close: 106 },
    { time: '2020-01-01T12:00:00', High: 109, Low: 101, Close: 103 },
]

calcVortex(arr, 'Close')
    .then((rs) => {
        let r = rs.find((v) => v.period === '12hr')
        console.log(r)
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [
        //     {
        //       time: '2020-01-01T12:00:00',
        //       VIplus: 1.2307692307692308,
        //       VIminus: 0.8461538461538461,
        //       VIdiff: 0.3846153846153847,
        //       VIsum: 2.076923076923077
        //     }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

calcVwap(arr, key, optopt) → {Promise}

Description:

• 計算各週期之Rolling VWAP偏離度(vwapDist)與其以ATR正規化後之偏離度(vwapDistATR)

  各週期len為該期涵蓋之K線根數(以4hr K線為基準，1day=6根)，直接讀取輸入arr內High、Low、Close、Volumn欄位計算，資料筆數不足len+1根時，該期vs為空陣列 ATR以Wilder平滑計算，種子為前len根TR之簡單平均；VWAP為典型價((High+Low+Close)/3)乘Volumn後之rolling窗口總和除以Volumn總和 各期第1筆對應輸入之第len根(與ATR對齊)，其後逐根對齊至最後一根 vwapDist於Close為0時給0；vwapDistATR於ATR為0時給0

  Unit Test: Github

Source:

• calcVwap.mjs, line 47

Example

let arr = [
    { time: '2020-01-01T00:00:00', High: 105, Low: 95, Close: 100, Volumn: 10 },
    { time: '2020-01-01T04:00:00', High: 110, Low: 100, Close: 108, Volumn: 12 },
    { time: '2020-01-01T08:00:00', High: 112, Low: 104, Close: 106, Volumn: 8 },
    { time: '2020-01-01T12:00:00', High: 109, Low: 101, Close: 103, Volumn: 15 },
]

calcVwap(arr, 'Close')
    .then((rs) => {
        let r = rs.find((v) => v.period === '12hr')
        console.log(r)
        // => {
        //   period: '12hr',
        //   len: 3,
        //   vs: [
        //     {
        //       time: '2020-01-01T12:00:00',
        //       vwapDist: -0.02515025427646778,
        //       vwapDistATR: -0.29890109890109784
        //     }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

(async) dataCalibrate(calcFn, arrSrc, arrRef, pairs, optMethodInit, optopt) → {Promise}

Description:

• 校正calc函式之optMethod: 搜尋參數，使來源側(src)各keyOut之z-score分布形狀對齊參考側(ref)對應keyOut

  本專案用例: src=溢價指數(index) OHLC，ref=價格(price) OHLC，校正index系變換偏移量 搜尋為逐維座標下降: 每維先粗掃(現值+0+正負mags量級)，再以粗掃最佳為中心做倍率細化與微調 防churn: 需(score0-scoreBest)/score0 > threshold，且(score0-scoreBest) > thresholdAbs，兩者皆達門檻才採納搜尋結果，否則optMethod維持optMethodInit

  Unit Test: Github

Source:

• dataCalibrate.mjs, line 81

Example

let mkArr = (n) => {
    let arr = []
    for (let i = 0; i < n; i++) {
        arr.push({ time: `t${i}`, u: Math.sin(i * 1.7) })
    }
    return arr
}

let calcFn = async (arr, keyIn, optMethod) => {
    let a = optMethod.a !== undefined ? optMethod.a : 0
    let vs = arr.map((v) => {
        let x = v.u + a * v.u * v.u
        return { time: v.time, X: x }
    })
    return [{ period: 'p1', len: 1, vs }]
}

let arr = mkArr(400)

dataCalibrate(calcFn, arr, arr, ['X'], { a: 0 }, { optMethodRef: { a: 0.1 } })
    .then((r) => {
        console.log(r)
        // => {
        //   adopt: true,
        //   optMethod: { a: 0.1 },
        //   optMethodBest: { a: 0.1 },
        //   score0: 0.0478199962161957,
        //   scoreBest: 0,
        //   nEval: 26,
        //   detail: [
        //     { keyOutSrc: 'X', keyOutRef: 'X', dist: 0, pNearZeroSrc: 0.0425, pNearZeroRef: 0.0425 }
        //   ]
        // }
    })

Parameters:

Returns:

Type

Promise

(async) dataConvert(name, symbol, interval, cointpe, keyKind, methods, arrOhlc, kp, fdSelfData, optopt) → {Promise}

Description:

• 依methods內各method呼叫kp對應之calc函式產生指標序列，再依各deal之nm(avg,std)做z-score正規化，並輸出各method/keyIn/keyOut組合之統計量

  若opt.funAddKeysIn為函式，會先逐筆對arrOhlc呼叫該函式(可回傳Promise)以轉換/擴增輸入資料 對keysIn(輸入計算所用數值欄位名稱)與methods(指標方法設定)做雙層迴圈，各methods[i]需含method(對應kp之鍵)、deals(輸出設定陣列)、optMethod(傳入calc函式之opt，可選，預設{}) 呼叫kpmethod取得rrs(各期{period,len,vs}結果)後，對每個deal依其keyOut取值並以(值-avg)/std正規化，若deal.keyIn有指定則僅處理keyIn相符者，其餘略過 除非opt.denyOutput為true，否則會將各期正規化後之序列以檔名${cointpe}_${keyKind}_${keyIn}_${keyOut}_${method}_${period}.json寫入fdSelfData資料夾 最終以w-statistic計算全期攤平後序列之avg、std、min、max，彙整為單筆結果

Source:

• dataConvert.mjs, line 85

Example

import calcMa from './calcMa.mjs'

let arr = []
for (let i = 0; i < 200; i++) {
    arr.push({
        time: new Date(Date.UTC(2020, 0, 1) + i * 4 * 3600 * 1000).toISOString().slice(0, 19),
        Close: 100 + Math.sin(i * 0.7) * 5 + Math.cos(i * 0.31) * 3,
    })
}

let kp = { ma: calcMa }

let methods = [
    {
        method: 'ma',
        deals: [
            { keyOut: 'param', nm: { avg: 0, std: 1 } },
        ],
        optMethod: {},
    },
]

dataConvert('demo', 'BTCUSDT', '4hr', 'price', 'ohlc', methods, arr, kp, './tmp/unused', { denyOutput: true, keysIn: ['Close'] })
    .then((rs) => {
        console.log(rs)
        // => [
        //   {
        //     method: 'ma',
        //     keyIn: 'Close',
        //     keyOut: 'param',
        //     avg: 99.968791151179,
        //     std: 1.3207330769565093,
        //     min: 95.38244060380985,
        //     max: 104.64005705802747
        //   }
        // ]
    })

Parameters:

Returns:

Type

Promise