function vout = indicators(vin,mode,varargin)
%INDICATORS calculates various technical indicators
%
% Description
% INDICATORS is a technical analysis tool that calculates various
% technical indicators. Technical analysis is the forecasting of
% future financial price movements based on an examination of past
% price movements. Most technical indicators require at least 1
% variable argument. If these arguments are not supplied, default
% values are used.
%
% Syntax
% Momentum
% cci = indicators([hi,lo,cl] ,'cci' ,tp_per,md_per,const)
% roc = indicators(price ,'roc' ,period)
% rsi = indicators(price ,'rsi' ,period)
% [fpctk,fpctd] = indicators([hi,lo,cl] ,'fsto' ,k,d)
% [spctk,spctd] = indicators([hi,lo,cl] ,'ssto' ,k,d)
% [fpctk,fpctd,jline] = indicators([hi,lo,cl] ,'kdj' ,k,d)
% willr = indicators([hi,lo,cl] ,'william',period)
% [dn,up,os] = indicators([hi,lo] ,'aroon' ,period)
% tsi = indicators(cl ,'tsi' ,r,s)
% Trend
% sma = indicators(price ,'sma' ,period)
% ema = indicators(price ,'ema' ,period)
% [macd,signal,macdh] = indicators(cl ,'macd' ,short,long,signal)
% [pdi,mdi,adx] = indicators([hi,lo,cl] ,'adx' ,period)
% t3 = indicators(price ,'t3' ,period,volfact)
% Volume
% obv = indicators([cl,vo] ,'obv')
% cmf = indicators([hi,lo,cl,vo] ,'cmf' ,period)
% force = indicators([cl,vo] ,'force' ,period)
% mfi = indicators([hi,lo,cl,vo] ,'mfi' ,period)
% Volatility
% [middle,upper,lower] = indicators(price ,'boll' ,period,weight,nstd)
% [middle,upper,lower] = indicators([hi,lo,cl] ,'keltner',emaper,atrmul,atrper)
% atr = indicators([hi,lo,cl] ,'atr' ,period)
% Other
% [index,value] = indicators(price ,'zigzag' ,moveper)
% change = indicators(price ,'compare')
% [pivot sprt res] = indicators([dt,op,hi,lo,cl],'pivot' ,type)
% sar = indicators([hi,lo] ,'sar' ,step,maximum)
%
% Arguments
% Outputs
% cci/roc/rsi/willr/tsi/sma/ema/t3/obv/cmf/force/mfi/atr/change/sar
% - single output vector
% macd/signal/macdh
% - moving average convergence divergence vector/
% signal line/ macd histogram
% middle/upper/lower
% - middle/upper/lower band for bollinger bands or keltner
% channels
% fpctk/fpctd/spctk/spctd/jline
% - fast/slow percent k/d and the J Line
% index/value
% - index and value for each point in zigzag
% pivot/sprt/res
% - vectors for pivot point, support lines, resistance
% lines
% dn/up/os
% - aroon down/aroon up/aroon oscillator
% Inputs
% price - any price vector (e.g. open, high, ...)
% dt/op/hi/lo/cl/vo
% - matlab serial date, open/high/low/close price, and volume of data
% Mode
% Momentum
% cci - Commodity Channel Index
% roc - Rate of Change
% rsi - Relative Strength Index
% fsto - Fast Stochastic Oscillator
% ssto - Slow Stochastic Oscillator
% kdj - KDJ Indicator
% william - William's %R
% aroon - Aroon
% tsi - True Strength Index
% Trend
% sma - Simple Moving Average
% ema - Exponential Moving Average
% macd - Moving Average Convergence Divergence
% adx - Wildmer's DMI (ADX)
% t3 - Triple EMA (Not the same as EMA3)
% Volume
% obv - On-Balance Volume
% cmf - Chaikin Money Flow
% force - Force Index
% mfi - Money Flow Index
% Volatility
% boll - Bollinger Bands
% keltner - Keltner Channels
% atr - Average True Range
% Other
% zigzag - ZigZag
% compare - relative price compared to first input
% pivot - Pivot Points
% sar - Parabolic SAR (Stop And Reverse)
%
% Variable Arguments
% period - number of periods over which to make calculations
% short/long/signal
% - number of periods for short/long/signal for macd
% period/weight/nstd
% - number of periods/weight factor/number of standard
% deviations
% k/d/j - number of periods for %K/%D/JLine
% r/s - number of periods for momentum/smoothed momentum
% emaper/atrmul/atrper
% - number of periods for ema/atr multiplier/number of
% periods for atr for keltner
% tp_per/md_per/const
% - number of periods for true price/number of periods for
% mean deviation/constant for cci
% moveper - movement percent for zigzag
% type - string for pivot point method pick one of 's', 'f', 'd'
% which stand for 'standard', 'fibonacci', 'demark'
% step/maximum
% - value to add to acceleration factor at each increment
% maximum value acceleration factor can reach
% volfact - volume factor for t3
%
% Notes
% - there are no argument checks
% - all prices must be column oriented
% - the parabolic sar indicator is not completely correct
% - if there is a tie between price points, the aroon indicator uses
% the one farthest back
% - 2 methods are available to calculate the ema for the tsi
% indicator. Simply uncomment whichever one is desired.
% - the t3 indicator uses a different ema than ta-lib
% - 3 methods are available to calculate the kdj. Simply uncomment
% whichever one is desired.
%
% Example
% load disney.mat
% [fpctk,fpctd] = indicators([dis_HIGH,dis_LOW,dis_CLOSE],'fsto',14,3)
% plot(ones(length(fpctk),1),fpctk,'b',ones(length(fpctd),1),fpctd,'g')
% title('Fast Stochastics for Disney')
%
% Further Information
% For an in depth analysis of how many of these technical indicators
% work, refer to one of the following websites or Google it.
% http://stockcharts.com/
% http://www.investopedia.com/
% http://www.ta-lib.org/
%
% Version : 1.1.2 (03/21/2013)
% Author : Nate Jensen
% Created : 10/10/2011
% History :
% - v1.0 10/25/2011 : initial release of 21 indicators
% - v1.1 03/04/2012 : 23 indicators, fixed date conversion issue
% - v1.1.1 03/25/2012 : 24 indicators
% - v1.1.2 03/21/2013 : 25 indicators
% To Do
% - add more indicators
%%% Main Function
% Initialize output vector
vout = [];
% Number of observations
observ = size(vin,1);
% Switch between the various modes
switch lower(mode)
%%% Momentum
%==========================================================================
case 'cci' % Commodity Channel Index
% Input Data
high = vin(:,1);
low = vin(:,2);
close = vin(:,3);
% Variable Argument Input
if isempty(varargin)
tp_per = 20;
md_per = 20;
const = 0.015;
else
tp_per = varargin{1};
md_per = varargin{2};
const = varargin{3};
end
% Typical Price
tp = (high+low+close)/3;
% Simple moving average of typical price
smatp = sma(tp,tp_per,observ);
% Sum of the mean absolute deviation
smad = nan(observ,1);
cci = smad; % preallocate cci
for i1 = md_per:observ
smad(i1) = sum(abs(smatp(i1)-tp(i1-md_per+1:i1)));
end
% Commodity Channel Index
i1 = md_per:observ;
cci(i1) = (tp(i1)-smatp(i1))./(const*smad(i1)/md_per);
% Format Output
vout = cci;
case 'roc' % Rate of Change
% Input Data
close = vin;
% Variable Argument Input
if isempty(varargin)
period = 12;
else
period = varargin{1};
end
% Rate of Change
roc = nan(observ,1);
% calculate rate of change
roc(period+1:observ) = ((close(period+1:observ)- ...
close(1:observ-period))./close(1:observ-period))*100;
% Format Output
vout = roc;
case 'rsi' % Relative Strength Index
% Input Data
close = vin;
% Variable Argument Input
if isempty(varargin)
period = 14;
else
period = varargin{1};
end
% Determine how many nans are in the beginning
nanVals = isnan(close);
firstVal = find(nanVals == 0, 1, 'first');
numLeadNans = firstVal - 1;
% Create vector of non-nan closing prices
nnanvin = close(~isnan(close));
% Take a diff of the non-nan closing prices
diffdata = diff(nnanvin);
priceChange = abs(diffdata);
% Create '+' Delta vectors and '-' Delta vectors
advances = priceChange;
declines = priceChange;
advances(diffdata < 0) = 0;
declines(diffdata >= 0) = 0;
% Calculate the RSI of the non-nan closing prices. Ignore first non-nan
% vin b/c it is a reference point. Take into account any leading nans
% that may exist in vin vector.
trsi = nan(size(diffdata, 1)-numLeadNans, 1);
for i1 = period:size(diffdata, 1)
% Gains/losses
totalGain = sum(advances((i1 - (period-1)):i1));
totalLoss = sum(declines((i1 - (period-1)):i1));
% Calculate RSI
rs = totalGain ./ totalLoss;
trsi(i1) = 100 - (100 / (1+rs));
end
% Pre allocate vector taking into account reference value and leading nans.
% length of vector = length(vin) - # of reference values - # of leading nans
rsi = nan(size(close, 1)-1-numLeadNans, 1);
% Populate RSI
rsi(~isnan(close(2+numLeadNans:end))) = trsi;
% Format Output
vout = [nan(numLeadNans+1, 1); rsi];
case 'fsto' % Fast Stochastic Oscillator
% Input Data
high = vin(:,1);
low = vin(:,2);
close = vin(:,3);
% Variable Argument Input
if isempty(varargin)
kperiods = 14;
dperiods = 3;
else
kperiods = varargin{1};
dperiods = varargin{2};
end
% Fast %K
fpctk = nan(observ,1); % preallocate Fast %K
llv = zeros(observ,1); % preallocate lowest low
llv(1:kperiods) = min(low(1:kperiods)); % lowest low of first kperiods
for i1 = kperiods:observ % cycle through rest of data
llv(i1) = min(low(i1-kperiods+1:i1)); % lowest low of previous kperiods
end
hhv = zeros(observ,1); % preallocate highest high
hhv(1:kperiods) = max(high(1:kperiods)); % highest high of first kperiods
for i1 = kperiods:observ % cycle through rest of data
hhv(i1) = max(high(i1-kperiods+1:i1)); % highest high of previous kperiods
end
nzero = find((hhv-llv) ~= 0);
fpctk(nzero) = ((close(nzero)-llv(nzero))./(hhv(nzero)-llv(nzero)))*100;
% Fast %D
fpctd = nan(size(close));
fpctd(~isnan(fpctk)) = ema(fpctk(~isnan(fpctk)),dperiods,observ);
% Format Output
vout = [fpctk,fpctd];
case 'ssto' % Slow Stochastic Oscillator
% Input data
high = vin(:,1);
low = vin(:,2);
close = vin(:,3);
% Variable Argument Input
if isempty(varargin)
kperiods = 14;
dperiods = 3;
else
kperiods = varargin{1};
dperiods = varargin{2};
end
% Fast %K
fpctk = nan(observ,1); % preallocate Fast %K
llv = zeros(observ,1); % preallocate lowest low
llv(1:kperiods) = min(low(1:kperiods)); % lowest low of first kperiods
for i1 = kperiods:observ % cycle through rest of data
llv(i1) = min(low(i1-kperiods+1:i1)); % lowest low of previous kperiods
end
hhv = zeros(observ,1); % preallocate highest high
hhv(1:kperiods) = max(high(1:kperiods)); % highest high of first kperiods
for i1 = kperiods:observ % cycle through rest of data
hhv(i1) = max(high(i1-kperiods+1:i1)); % highest high of previous kperiods
end
nzero = find((hhv-llv) ~= 0);
fpctk(nzero) = ((close(nzero)-llv(nzero))./(hhv(nzero)-llv(nzero)))*100;
% Fast %D
fpctd = nan(size(close));
fpctd(~isnan(fpctk)) = ema(fpctk(~isnan(fpctk)),dperiods,observ);
% Slow %K
spctk = fpctd;
% Slow %D
spctd = nan(size(close));
spctd(~isnan(spctk)) = ema(spctk(~isnan(spctk)),dperiods,observ-dperiods+1);
% Format Output
vout = [spctk,spctd];
case 'kdj' % KDJ Indicator
% Input Data
hi = vin(:,1);
lo = vin(:,2);
cl = vin(:,3);
% Variable Argument Input
if isempty(varargin)
kperiods = 14;
dperiods = 3;
% jperiods = 5;
else
kperiods = varargin{1};
dperiods = varargin{2};
% jperiods = varargin{3};
end
% Fast %K
fpctk = nan(observ,1); % preallocate Fast %K
llv = zeros(observ,1); % preallocate lowest low
llv(1:kperiods) = min(lo(1:kperiods)); % lowest low of first kperiods
for i1 = kperiods:observ % cycle through rest of data
llv(i1) = min(lo(i1-kperiods+1:i1)); % lowest low of previous kperiods
end
hhv = zeros(observ,1); % preallocate highest high
hhv(1:kperiods) = max(hi(1:kperiods)); % highest high of first kperiods
for i1 = kperiods:observ % cycle through rest of data
hhv(i1) = max(hi(i1-kperiods+1:i1)); % highest high of previous kperiods
end
nzero = find((hhv-llv) ~= 0);
fpctk(nzero) = ((cl(nzero)-llv(nzero))./(hhv(nzero)-llv(nzero)))*100;
% Fast %D
fpctd = nan(size(cl));
fpctd(~isnan(fpctk)) = ema(fpctk(~isnan(fpctk)),dperiods,observ);
% Method # 1:
jline = 3*fpctk-2*fpctd;
% Method # 2:
% jline = nan(size(cl));
% jline(~isnan(fpctk)) = ema(fpctk(~isnan(fpctk)),jperiods,observ);
% Method # 3:
% Slow %K
% spctk = fpctd;
% Slow %D
% spctd = nan(size(cl));
% spctd(~isnan(spctk)) = ema(spctk(~isnan(spctk)),dperiods,observ-dperiods+1);
% J Line
% jline = 3*spctk-2*spctd;
% Format Output
% vout = [spctk,spctd,jline];
% Format Output
vout = [fpctk,fpctd,jline];
case 'william' % William's %R
% Input Data
high = vin(:,1);
low = vin(:,2);
close = vin(:,3);
% Variable Argument Input
if isempty(varargin)
period = 14;
else
period = varargin{1};
end
% Highest High and Lowest Low
llv = zeros(observ,1); % preallocate lowest low
llv(1:period) = min(low(1:period)); % lowest low of first kperiods
for i1 = period:observ % cycle through rest of data
llv(i1) = min(low(i1-period+1:i1)); % lowest low of previous kperiods
end
hhv = zeros(observ,1); % preallocate highest high
hhv(1:period) = max(high(1:period)); % highest high of first kperiods
for i1 = period:observ % cycle through rest of data
hhv(i1) = max(high(i1-period+1:i1)); % highest high of previous kperiods
end
% Williams %R
wpctr = nan(observ,1);
nzero = find((hhv-llv) ~= 0);
wpctr(nzero) = ((hhv(nzero)-close(nzero))./(hhv(nzero)-llv(nzero))) * -100;
% Format output
vout = wpctr;
case 'aroon' % Aroon
% Input Data
high = vin(:,1);
low = vin(:,2);
% Variable Argument Input
if isempty(varargin)
period = 25;
else
period = varargin{1};
end
% Cumulative sum of end indices
% Output looks like:
% [1 16 31 46 61 76 91 ... ]
temp_var1 = cumsum([1;(period+1:observ)'-(1:observ-period)'+1]);
% Vector of moving indices
% Output looks like:
% [1 2 3 4 5 2 3 4 5 6 3 4 5 6 7 4 5 6 7 8 ... ]
temp_var2 = ones(temp_var1(observ-period+1)-1,1);
temp_var2(temp_var1(1:observ-period)) = 1-period;
temp_var2(1) = 1;
temp_var2 = cumsum(temp_var2);
% Days since last n periods high/low
[~,min_idx] = min(low(reshape(temp_var2,period+1,observ-period)),[],1);
[~,max_idx] = max(high(reshape(temp_var2,period+1,observ-period)),[],1);
% Aroon Down/Up/Oscillator
aroon_dn = [nan(period,1); ((period-(period+1-min_idx'))/period)*100];
aroon_up = [nan(period,1); ((period-(period+1-max_idx'))/period)*100];
aroon_os = aroon_up-aroon_dn;
% Format Output
vout = [aroon_dn,aroon_up,aroon_os];
case 'tsi' % True Strength Index
% Input Data
close = vin(:,1);
% Variable Argument Input
if isempty(varargin)
slow = 25;
fast = 13;
else
slow = varargin{1};
fast = varargin{2};
end
% If the lag is greater than or equal to the number of observations
if slow >= observ || fast >= observ
return
end
% Momentum
mtm = [0; (close(2:end,1)) - close(1:end-1,1)];
absmtm = abs(mtm);
% Calculate the exponential percentage
k1 = 2/(slow+1);
k2 = 2/(fast+1);
% Wikipedia method for calculating ema
% Preallocate
ema1 = zeros(observ,1);
ema2 = ema1;
ema3 = ema1;
ema4 = ema1;
% EMA's
for i1 = 2:observ
ema1(i1) = k1 * (mtm(i1)-ema1(i1-1)) + ema1(i1-1);
ema2(i1) = k2 * (ema1(i1)-ema2(i1-1)) + ema2(i1-1);
ema3(i1) = k1 * (absmtm(i1)-ema3(i1-1)) + ema3(i1-1);
ema4(i1) = k2 * (ema3(i1)-ema4(i1-1)) + ema4(i1-1);
end
% True Strength Index
tsi = 100*ema2./ema4;
% % Matlab method for calculating ema
% % Preallocate EMA's
% ema1 = nan(observ,1);
% ema2 = ema1;
% ema3 = ema1;
% ema4 = ema1;
%
% % Calculate the simple moving average for the first 'exp mov avg' value.
% ema1(slow) = sum(mtm(1:slow))/slow;
% ema3(slow) = sum(absmtm(1:slow))/slow;
%
% % K*vin; 1-k
% kvin1 = mtm(slow:observ) * k1;
% oneK1 = 1-k1;
% kvin3 = absmtm(slow:observ) * k1;
% oneK2 = 1-k2;
%
% % First period calculation
% ema1(slow) = kvin1(1) + (ema1(slow) * oneK1);
% ema3(slow) = kvin3(1) + (ema3(slow) * oneK1);
%
% % Remaining periods calculation
% for i1 = slow+1:observ
% ema1(i1) = kvin1(i1-slow+1) + (ema1(i1-1) * oneK1);
% ema3(i1) = kvin3(i1-slow+1) + (ema3(i1-1) * oneK1);
% end
%
% % Calculate the simple moving average for the first 'exp mov avg' value.
% ema2(slow+fast-1) = sum(ema1(slow:slow+fast-1))/fast;
% ema4(slow+fast-1) = sum(ema3(slow:slow+fast-1))/fast;
%
% % K*vin; 1-k
% kvin2 = ema1(slow+fast-1:observ) * k2;
% kvin4 = ema3(slow+fast-1:observ) * k2;
%
% % First period calculation
% ema2(slow+fast-1) = kvin2(1) + (ema2(slow+fast-1) * oneK2);
% ema4(slow+fast-1) = kvin4(1) + (ema4(slow+fast-1) * oneK2);
%
% % Remaining periods calculation
% for i1 = slow+fast:observ
% ema2(i1) = kvin2(i1-fast-slow+2) + (ema2(i1-1) * oneK2);
% ema4(i1) = kvin4(i1-fast-slow+2) + (ema4(i1-1) * oneK2);
% end
%
% % True Strength Index
% tsi = 100*ema2./ema4;
% Format Output
vout = tsi;
%--------------------------------------------------------------------------
%%% Trend
%==========================================================================
case 'sma' % Simple Moving Average
% Input Data
price = vin;
% Variable Argument Input
if isempty(varargin)
period = 20;
else
period = varargin{1};
end
% Simple Moving Average
simmovavg = sma(price,period,observ);
% Format Output
vout = simmovavg;
case 'ema' % Exponential Moving Average
% Input Data
price = vin;
% Variable Argument Input
if isempty(varargin)
period = 20;
else
period = varargin{1};
end
% Exponential Moving Average
expmovavg = ema(price,period,observ);
% Format Output
vout = expmovavg;
case 'macd' % Moving Average Convergence Divergence
% Input Data
close = vin;
% Variable Argument Input
if isempty(varargin)
short = 12;
long = 26;
signal = 9;
else
short = varargin{1};
long = varargin{2};
signal = varargin{3};
end
% EMA of Long Period
[ema_lp status] = ema(close,long,observ);
if ~status
return
end
% EMA of Short Period
ema_sp = ema(close,short,observ);
% MACD
MACD = ema_sp-ema_lp;
% Signal
[signal status] = ema(MACD(~isnan(MACD)),signal,observ-long+1);
if ~status
return
end
signal = [nan(long-1,1);signal];
% MACD Histogram
MACD_h = MACD-signal;
% Format Output
vout = [MACD,signal,MACD_h];
case 'adx' % Wilder's DMI (ADX)
% Input Data
high = vin(:,1);
low = vin(:,2);
close = vin(:,3);
% Variable argument input
if isempty(varargin)
period = 14;
else
period = varargin{1};
end
% True range
h_m_l = high-low; % high - low
h_m_c = [0;abs(high(2:observ)-close(1:observ-1))]; % abs(high - close)
l_m_c = [0;abs(low(2:observ)-close(1:observ-1))]; % abs(low - close)
tr = max([h_m_l,h_m_c,l_m_c],[],2); % true range
% Directional Movement
h_m_h = high(2:observ)-high(1:observ-1); % high - high
l_m_l = low(1:observ-1)-low(2:observ); % low - low
pdm1 = zeros(observ-1,1); % preallocate pdm1
max_h = max(h_m_h,0);
pdm1(h_m_h > l_m_l) = max_h(h_m_h > l_m_l); % plus
mdm1 = zeros(observ-1,1); % preallocate mdm1
max_l = max(l_m_l,0);
mdm1(l_m_l > h_m_h) = max_l(l_m_l > h_m_h); % minus
pdm1 = [nan;pdm1];
mdm1 = [nan;mdm1];
% Preallocate 14 period tr, pdm, mdm, adx
tr14 = nan(observ,1); % 14 period true range
pdm14 = tr14; % 14 period plus directional movement
mdm14 = tr14; % 14 period minus directional movement
adx = tr14; % average directional index
% Calculate tr14, pdm14, mdm14, pdi14, mdi14, dmx
tr14(period+1) = sum(tr(period+1-period+1:period+1));
pdm14(period+1) = sum(pdm1(period+1-period+1:period+1));
mdm14(period+1) = sum(mdm1(period+1-period+1:period+1));
for i1 = period+2:observ
tr14(i1) = tr14(i1-1)-tr14(i1-1)/period+tr(i1);
pdm14(i1) = pdm14(i1-1)-pdm14(i1-1)/period+pdm1(i1);
mdm14(i1) = mdm14(i1-1)-mdm14(i1-1)/period+mdm1(i1);
end
pdi14 = 100*pdm14./tr14; % 14 period plus directional indicator
mdi14 = 100*mdm14./tr14; % 14 period minus directional indicator
dmx = 100*abs(pdi14-mdi14)./(pdi14+mdi14);% directional movement index
% Average Directional Index
adx(2*period) = sum(dmx(period+1:2*period))/(2*period-period-1);
for i1 = 2*period+1:observ
adx(i1) = (adx(i1-1)*(period-1)+dmx(i1))/period;
end
% Format Output
vout = [pdi14,mdi14,adx];
case 't3' % T3
% Input Data
price = vin;
% Variable Argument Input
if isempty(varargin)
period = 5;
volfact = 0.7;
else
period = varargin{1};
volfact = varargin{2};
end
% EMA
ema1 = ema(price,period,observ);
ema2 = [nan(period,1); ema(ema1(~isnan(ema1)),period,observ-period+1)];
ema3 = [nan(2*period-1,1); ema(ema2(~isnan(ema2)),period,observ-2*period+1)];
ema4 = [nan(3*period-1,1); ema(ema3(~isnan(ema3)),period,observ-3*period+1)];
ema5 = [nan(4*period-1,1); ema(ema4(~isnan(ema4)),period,observ-4*period+1)];
ema6 = [nan(5*period-1,1); ema(ema5(~isnan(ema5)),period,observ-5*period+1)];
% Constants
c1 = -(volfact*volfact*volfact);
c2 = 3*(volfact*volfact-c1);
c3 = -6*volfact*volfact-3*(volfact-c1);
c4 = 1+3*volfact-c1+3*volfact*volfact;
% T3
t3 = c1*ema6+c2*ema5+c3*ema4+c4*ema3;
% Format Output
vout = t3;
%--------------------------------------------------------------------------
%%% Volume
%==========================================================================
case 'obv' % On-Balance Volume
% Input data
close = vin(:,1);
volume = vin(:,2);
% On-Balance Volume
obv = volume;
for i1 = 2:observ
if close(i1) > close(i1-1)
obv(i1) = obv(i1-1)+volume(i1);
elseif close(i1) < close(i1-1)
obv(i1) = obv(i1-1)-volume(i1);
elseif close(i1) == close(i1-1)
obv(i1) = obv(i1-1);
end
end
% Format Output
vout = obv;
case 'cmf' % Chaikin Money Flow
% Input Data
high = vin(:,1);
low = vin(:,2);
close = vin(:,3);
volume = vin(:,4);
% Variable Argument Input
if isempty(varargin)
period = 20;
else
period = varargin{1};
end
% Money Flow Multiplier
mfm = ((close-low)-(high-close))/(high-low);
% Money Flow Volume
mfv = mfm*volume;
% Chaikin Money Flow
cmf = nan(observ,1);
for i1 = period:observ
cmf(i1) = sum(mfv(i1-period+1:i1))/sum(volume(i1-period+1:i1));
end
% Format Output
vout = cmf;
case 'force' % Force Index
% Input Data
close = vin(:,1);
volume = vin(:,2);
% Variable Argument Input
if isempty(varargin)
period = 13;
else
period = varargin{1};
end
% Force Index
force = [nan; (close(2:observ)-close(1:observ-1)).*volume(2:observ)];
force = [nan; ema(force(2:observ),period,observ-1)];
% Format Output
vout = force;
case 'mfi' % Money Flow Index
% Input Data
% Input Data
high = vin(:,1);
low = vin(:,2);
close = vin(:,3);
volume = vin(:,4);
if isempty(varargin)
period = 14;
else
period = varargin{1};
end
% Typical Price
tp = (high+low+close)/3;
% Up or Down
upordn = ones(observ-1,1);
upordn(tp(2:observ) <= tp(1:observ-1)) = -1;
% Raw Money Flow
rmf = tp(2:observ).*volume(2:observ);
% Positive Money Flow
pmf = zeros(observ-1,1);
pmf(upordn == 1) = rmf(upordn == 1);
% Negative Money Flow
nmf = zeros(observ-1,1);
nmf(upordn == -1) = rmf(upordn == -1);
% Cumulative sum of end indices
% Output looks like:
% [1 16 31 46 61 76 91 ... ]
temp_var1 = cumsum([1;(period:observ-1)'-(1:observ-period)'+1]);
% Vector of moving indices
% Output looks like:
% [1 2 3 4 5 2 3 4 5 6 3 4 5 6 7 4 5 6 7 8 ... ]
temp_var2 = ones(temp_var1(observ-period+1)-1,1);
temp_var2(temp_var1(1:observ-period)) = 2-period;
temp_var2(1) = 1;
temp_var2 = cumsum(temp_var2);
% Money Flow Ratio
mfr = sum(pmf(reshape(temp_var2,period,observ-period)),1)'./ ...
sum(nmf(reshape(temp_var2,period,observ-period)),1)';
mfr = [nan(period,1); mfr];
% Money Flow Index
mfi = 100-100./(1+mfr);
% Format Output
vout = mfi;
%--------------------------------------------------------------------------
%%% Volatility
%==========================================================================
case 'boll' % Bollinger Bands
% Input data
close = vin;
% Variable argument input
if isempty(varargin)
period = 20;
weight = 0;
nstd = 2;
else
period = varargin{1};
weight = varargin{2};
nstd = varargin{3};
end
% Create output vectors.
mid = nan(size(close, 1), 1);
uppr = mid;
lowr = mid;
% Create weight vector.
wtsvec = ((1:period).^weight) ./ (sum((1:period).^weight));
% Save the original data and remove NaN's from the data to be processed.
nnandata = close(~isnan(close));
% Calculate middle band moving average using convolution.
cmid = conv(nnandata, wtsvec);
nnanmid = cmid(period:length(nnandata));
% Calculate shift for the upper and lower bands. The shift is a
% moving standard deviation of the data.
mstd = nnandata(period:end); % Pre-allocate
for i1 = period:length(nnandata)
mstd(i1-period+1, :) = std(nnandata(i1-period+1:i1));
end
% Calculate the upper and lower bands.
nnanuppr = nnanmid + nstd.*mstd;
nnanlowr = nnanmid - nstd.*mstd;
% Return the values.
nanVec = nan(period-1,1);
mid(~isnan(close)) = [nanVec; nnanmid];
uppr(~isnan(close)) = [nanVec; nnanuppr];
lowr(~isnan(close)) = [nanVec; nnanlowr];
% Format output
vout = [mid,uppr,lowr];
case 'keltner' % Keltner Channels
% Input data
high = vin(:,1);
low = vin(:,2);
close = vin(:,3);
% Variable argument input
if isempty(varargin)
emaper = 20;
atrmul = 2;
atrper = 10;
else
emaper = varargin{1};
atrmul = varargin{2};
atrper = varargin{3};
end
% True range
h_m_l = high-low; % high - low
h_m_c = [0;abs(high(2:observ)-close(1:observ-1))]; % abs(high - close)
l_m_c = [0;abs(low(2:observ)-close(1:observ-1))]; % abs(low - close)
tr = max([h_m_l,h_m_c,l_m_c],[],2); % true range
% Average true range
atr = ema(tr,atrper,observ);
% Middle/Upper/Lower bands of keltner channels
midd = ema(close,emaper,observ);
uppr = midd+atrmul*atr;
lowr = midd-atrmul*atr;
% Format output
vout = [midd,uppr,lowr];
case 'atr' % Average True Range
% Input data
high = vin(:,1);
low = vin(:,2);
close = vin(:,3);
% Variable argument input
if isempty(varargin)
period = 20;
else
period = varargin{1};
end
% True range
h_m_l = high-low; % high - low
h_m_c = [0;abs(high(2:observ)-close(1:observ-1))]; % abs(high - close)
l_m_c = [0;abs(low(2:observ)-close(1:observ-1))]; % abs(low - close)
tr = max([h_m_l,h_m_c,l_m_c],[],2); % true range
% Average true range
atr = ema(tr,period,observ);
% Format Output
vout = atr;
%--------------------------------------------------------------------------
%%% Other
%==========================================================================
case 'zigzag' % ZigZag
% Input data
close = vin;
% Variable argument input
if isempty(varargin)
moveper = 7;
else
moveper = varargin{1};
end
% Preallocate zigzag
zigzag = nan(observ,1);
% First zigzag is first data point of input vector
zigzag(1) = 1;
i1 = 1; % index of input data
i2 = 1; % index of output vector
% The number of outputs is unknown
while 1
% Find the first value in the input, from the current index to
% the number of observations, that has a price movement of
% moveper greater or less than the value of current index of
% the input and return the index of that value
% If all of the following conditions are met, temp_var1 is the
% index a zigzag
temp_var1 = find(close(i1:observ) > close(i1)+close(i1)*moveper/100 | ...
close(i1:observ) < close(i1)-close(i1)*moveper/100,1,'first');
% If no value is found
if isempty(temp_var1)
% If the current index is less than the number of
% observations
if i1 < observ
% If the current index of the output vector is greater
% than 1
if i2 > 1
% If the value of the input of the last recorded
% index is less than the value of the input of the
% index 2 recordings ago and there is a value
% between the value of the last recorded index and
% the number of observations that is less than the
% value of the last recorded index
if close(zigzag(i2)) < close(zigzag(i2-1)) && ...
min(close(i1:observ)) < close(zigzag(i2))
% Find the index of the minimum value that is
% between the index of the last recorded value
% and the number of observations
[~,temp_var1] = min(close(zigzag(i2):observ));
% Set the output of the current index equal to
% the previously calculated index
zigzag(i2) = i1+temp_var1-1;
% The opposite of the previous if statement
elseif close(zigzag(i2)) > close(zigzag(i2-1)) && ...
max(close(i1:observ)) > close(zigzag(i2))
[~,temp_var1] = max(close(zigzag(i2):observ));
zigzag(i2) = i1+temp_var1-1;
% The previous 2 statements are not true
% The output vector is complete
else
break
end
% The previous statement is not true
% The output vector is complete
else
break
end
% The previous statement is not true
% The output vector is complete
else
break
end
% If the current index of the output vector is greater than 1
elseif i2 > 1
% If the value of the index of temp_var1 is greater than
% the value of the last recorded index and the the value of
% the last recorded index is greater than the value of the
% index 2 recordings ago
if close(temp_var1+i1-1) > close(zigzag(i2)) && ...
close(zigzag(i2)) > close(zigzag(i2-1))
% Set the output of the current index equal to the
% index temp_var1
zigzag(i2) = temp_var1+i1-1;
% The opposit of the previous if statement
elseif close(temp_var1+i1-1) < close(zigzag(i2)) && ...
close(zigzag(i2)) < close(zigzag(i2-1))
zigzag(i2) = temp_var1+i1-1;
% If the value of the input of the last recorded index is
% less than the value of the input of the index 2
% recordings ago and there is a value between the value of
% the last recorded index and temp_var1 that is less than
% the value of the last recorded index
elseif close(zigzag(i2)) < close(zigzag(i2-1)) && ...
min(close(zigzag(i2):temp_var1+i1-1)) < close(zigzag(i2))
% Find the index of the minimum value that is between
% the index of the last recorded value and temp_var1
[~,temp_var1] = min(close(zigzag(i2):temp_var1+i1-1));
% Set the output of the current index equal to the
% previously calculated index
zigzag(i2) = temp_var1+i1-1;
% The opposite of the previous statement
elseif close(zigzag(i2)) > close(zigzag(i2-1)) && ...
max(close(zigzag(i2):temp_var1+i1-1)) > close(zigzag(i2))
[~,temp_var1] = max(close(zigzag(i2):temp_var1+i1-1));
zigzag(i2) = temp_var1+i1-1;
% The previous 4 statements are not true
else
% Increment the index of the output vector
% set the output of the incremented index equal to
% temp_var1
i2 = i2+1;
zigzag(i2) = temp_var1+i1-1;
end
% The current index of the output is equal to 1
else
% increment the index of the output vector
% set the output of the incremented index equal to
% temp_var1
i2 = i2+1;
zigzag(i2) = temp_var1+i1-1;
end
% Increment the index of the input data
i1 = temp_var1+i1-1;
end
% Redefine the output data equal to the index of each zigzag, and
% the value of the index of each zigzag
zigzag = [zigzag(~isnan(zigzag)),close(zigzag(~isnan(zigzag)))];
% Format output
vout = zigzag;
case 'compare' % Price Comparison
% Input data
numvars = size(vin,2);
price = vin;
% Percent change relative to first price
delta_percent = nan(observ,numvars);
delta_percent(2:observ,:) = 100*(price(2:observ,:)-price(1,:))./price(1,:);
% Format output
vout = delta_percent;
case 'pivot' % Pivot Points
% Input Data
date = vin(:,1);
open = vin(:,2);
high = vin(:,3);
low = vin(:,4);
close = vin(:,5);
% Variable Argument Input
if isempty(varargin)
type = 's';
else
type = varargin{1};
end
% Convert Matlab time to years, months, and days
[year,month,day,~,~,~] = datevecmx(date);
% Frequency
freq = diff(date);
if sum(freq)/observ < 1 % Intraday
freq = day;
elseif sum(freq)/observ < 7 % Daily
freq = month;
else % Weekly/Monthly
freq = year;
end
% Reassign open, high, low, and close based on frequency
temp_var1 = unique(freq);
num_dates = length(temp_var1);
new_open = nan(observ,1);
new_high = nan(observ,1);
new_low = nan(observ,1);
new_close = nan(observ,1);
for i1 = 2:num_dates
last_per = freq == temp_var1(i1-1);
this_per = freq == temp_var1(i1);
temp_var2 = open(last_per);
new_open (this_per) = temp_var2(1);
new_high (this_per) = max(high(last_per));
new_low (this_per) = min(low(last_per));
temp_var2 = close(last_per);
new_close(this_per) = temp_var2(end);
end
% Pivot Point
switch type
case 's' % Standard
pivot = (new_high+new_low+new_close)/3;
sprt(:,1) = pivot*2-new_high;
sprt(:,2) = pivot-(new_high-new_low);
res(:,1) = pivot*2-new_low;
res(:,2) = pivot+new_high-new_low;
case 'f' % Fibonacci
pivot = (new_high+new_low+new_close)/3;
sprt(:,1) = pivot-0.382*(new_high-new_low);
sprt(:,2) = pivot-0.612*(new_high-new_low);
sprt(:,3) = pivot-(new_high-new_low);
res(:,1) = pivot+0.382*(new_high-new_low);
res(:,2) = pivot+0.612*(new_high-new_low);
res(:,3) = pivot+(new_high-new_low);
case 'd' % Demark
X = nan(observ,1);
temp_var1 = new_high+2*new_low+new_close;
temp_var2 = 2*new_high+new_low+new_close;
temp_var3 = new_high+new_low+2*new_close;
X(new_close < new_open) = temp_var1(new_close < new_open);
X(new_close > new_open) = temp_var2(new_close > new_open);
X(new_close == new_open) = temp_var3(new_close == new_open);
pivot = X/4;
sprt = X/2-new_high;
res = X/2-new_low;
end
% Format Ouput
vout = [pivot sprt res];
case 'sar' % Parabolic SAR (Stop And Reverse)
% Input Data
high = vin(:,1);
low = vin(:,2);
% Variable Argument Input
if isempty(varargin)
step = 0.02;
maximum = 0.2;
else
step = varargin{1};
maximum = varargin{2};
end
af = step;
% Directional Movement
h_m_h = high(2)-high(1); % high - high
l_m_l = low(1)-low(2); % low - low
pdm = 0; % preallocate pdm1
mdm = 0; % preallocate mdm1
max_h = max(h_m_h,0); % max high
max_l = max(l_m_l,0); % max low
pdm(h_m_h > l_m_l) = max_h(h_m_h > l_m_l); % +DM
mdm(l_m_l > h_m_h) = max_l(l_m_l > h_m_h); % -DM
% false is long true is short
new_dir = false;
new_dir(mdm < pdm) = true;
% Defaults
out_sar = nan(observ,1);
ep (new_dir) = high(2);
sar(new_dir) = low(1);
ep (~new_dir) = low(2);
sar(~new_dir) = high(1);
for i1 = 1:observ-1
if new_dir
% Switch to short if the low penetrates the SAR value
if low(i1+1) <= sar
new_dir = false;
sar = ep;
% sar(sar < high(i1)) = high(i1);
% sar(sar < high(i1+1)) = high(i1+1);
out_sar(i1+1) = sar;
af = step;
ep = low(i1+1);
sar = sar+af*(ep-sar);
% sar(sar < high(i1)) = high(i1);
% sar(sar < high(i1+1)) = high(i1+1);
else
out_sar(i1+1) = sar;
af(high(i1+1) > ep) = af+step;
ep(high(i1+1) > ep) = high(i1+1);
af(af > maximum) = maximum;
sar = sar+af*(ep-sar);
% sar(sar > low(i1)) = low(i1);
% sar(sar > low(i1+1)) = low(i1+1);
end
else
% Switch to long if the high penetrates the SAR value
if high(i1+1) >= sar
new_dir = true;
sar = ep;
% sar(sar > low(i1)) = low(i1);
% sar(sar > low(i1+1)) = low(i1+1);
out_sar(i1+1) = sar;
af = step;
ep = high(i1+1);
sar = sar+af*(ep-sar);
% sar(sar > low(i1)) = low(i1);
% sar(sar > low(i1+1)) = low(i1+1);
else
out_sar(i1+1) = sar;
af(low(i1+1) < ep) = af+step;
ep(low(i1+1) < ep) = low(i1+1);
af(af > maximum) = maximum;
sar = sar+af*(ep-sar);
% sar(sar < high(i1)) = high(i1);
% sar(sar < high(i1+1)) = high(i1+1);
end
end
end
% Format Output
vout = out_sar;
%--------------------------------------------------------------------------
end
end
%%% Simple Moving Average
%==========================================================================
function [vout status] = sma(vin,lag,observ)
% Set status
status = 1;
% If the lag is greater than or equal to the number of observations
if lag >= observ
% End function, set status
status = 0;
return
end
% Preallocate a vector of nan's
vout = nan(observ,1);
% Simple moving average
ma = filter(ones(1,lag)/lag,1,vin);
% Fill in the nan's
vout(lag:end) = ma(lag:end);
end
%--------------------------------------------------------------------------
%%% Exponential Moving Average
%==========================================================================
function [vout status] = ema(vin,lag,observ)
% Preallocate output
vout = nan(observ,1);
% Set status
status = 1;
% If the lag is greater than or equal to the number of observations
if lag >= observ
status = 0;
return
end
% Calculate the exponential percentage
k = 2/(lag+1);
% Calculate the simple moving average for the first 'exp mov avg' value.
vout(lag) = sum(vin(1:lag))/lag;
% K*vin; 1-k
kvin = vin(lag:observ) * k;
oneK = 1-k;
% First period calculation
vout(lag) = kvin(1) + (vout(lag) * oneK);
% Remaining periods calculation
for i1 = lag+1:observ
vout(i1) = kvin(i1-lag+1) + (vout(i1-1) * oneK);
end
end
%--------------------------------------------------------------------------
