Kinito.Finance

Kinito Finance Repository

Web Scrapping

Medium Artcile Link: https://medium.freecodecamp.org/how-to-scrape-websites-with-python-and-beautifulsoup-5946935d93fe

MatPlotLib

https://pbpython.com/effective-matplotlib.html

Matplotlib Tutorial: https://matplotlib.org/tutorials/introductory/lifecycle.html

Technical Indicators

Moving Average Convergence Divergence - MACD (Lagging)

Many false positives on sideways market, to be used with others

Typical made by 3 periods:

12 Slow Moving Averga
26 Fast Moving Average
9 Signal

Average True Rate & Bollinger Bands (Volatility)

Bollinger Bands made of N lines (2) from M periods (20) moving average line. Delta is wide on high volatility & short during reduced volatility

ATR range perspective as swinging moves

Relative Strenght Index - RSI

Momentum oscilation [0,100]

[70,100] overbought (Emerging markets use 80)
[0,30] oversold (Emerging markets use 80)

Average Directional IndeX - ADX

Trend stength [0,100] comparing successive signal, then calculating the smoothed average of the difference of these

[0,25] abscent or weak
[25,50] strong
[50,75] very strong
[75,100] exteemely strong

OBV (Leading)

Volume preceeding price movement. Higher OBV price increase, lower OBV price decrease. False signals Used with MACD

Slope Charts - Supertrend

OLS aproximation on actual moving average trend

https://github.com/ChillarAnand/stocktrends

Renko Charts

Timeseries data on price movement, filtering noise to see real trend. Built on bricks. Time axis but not fix. Closing price based calculation

pip install stocktrends

TA-Lib Technical Analysis (C++) for Python 3.6 x32

TA-Lib website: http://ta-lib.org/

TA-Lib Python Wrapper Github Page: https://mrjbq7.github.io/ta-lib/

TA-Lib documentation of pattern recognition: https://mrjbq7.github.io/ta-lib/func_groups/pattern_recognition.html

Discussion on installation problems: TA-Lib/ta-lib-python#127

Command to install TA-lib for python 3.5 and 3.6:

pip install -i https://pypi.anaconda.org/masdeseiscaracteres/simple ta-lib

Good website on chart patterns: http://thepatternsite.com

Performance Measurements KPIs

Compounded Annual Growth Rate - CAGR

Comparison of different trading strategies for asset or portfolio. Does not reflect investment risk, use with volatility measure.

Volatility Ratio

Standard deviation of the returns to capture variability of returns from the mean return. Annualization by multiplying volatility with aquare root of annualization factor:

annualize daily volatility * sqrt(252) => 252 traiding days
annualize weekly volatility * sqrrt(52) => 52 traiding weeks
annualize monthly volatility * sqrt(12) => 12 traiding months

Used as risk measure assuming normal distribution of returns (false) Does not capture tail risk.

Sharpe Ratio

Average return earned in excess of the risk free rate per unit of volatility Measure of risk adjusted return Fails to dicern between upside and downside fluctuation Used to compare funds

[1,2] good
[2,3] very good
[3,more] excellent

Sortino Ratio

Similar to Sharpe Ratio but uses standard deviation of only negative returns. Used to dicern between upside and downside fluctuation => consider harmful volatility

Maximum Drawdown Ratio

Largest percentage drop in asset price over a specified time period (peak distance difference)

Calmar Ratio

Ratio of CAGR and Maximum Draw Down. Measure of risk adjusted return

Strategies

First test your system/ strategy with a set of rules with trading signal mocking data with trading conditions Factor in slippage (trading/ brokerage costs) when assessing performance. Be conservative on side of caution

Rebalancing - long term

Find a logical subset (L/ M/ S cap, industry, factor) specific and stick to this group. Build fixed individual position sized long only portfolio by picking:

M number of stocks based on monthly returns
any other criteria

Rebalance every month replacing N worse stocks by top N stocks Backtest it and compare the KPIs with that simple buy and hold strategy of corresponding index

Intraday Resistance Breakout

Helps figure out when stock reach resistance (temporary historical top value - upper line) Choose high volume, high activity stocks for this strategy (pre market movers, historically high volumes stocks)

Define breakout rule:

use price breach 20 * period rolling max/ min price & vlume breaching rolling max volume (go long/ short stocks based on signals)

Define exit/ stop loss signal:

use previous price +/- 20 * period ATR as the roling stop loss price

Calculate strategy cumulative return for each stock

Intraday Renko Ratio - On Balance Volume

Pick high volume, high activity stock (pre market movers, historically high volumes stocks)

Define Buy Signal:

Renko Ratio bar >= 2
5 day OBV Slope >= 30 degrees
exit if Renko Ratio bar < 2

Define Sell Signal:

Renko Ratio bar <= -2
5 day OBV Slope <= -30 degrees
exit if Renko Ratio bar > -2

Intraday Renko Ratio - MACD

Pick high volume, high activity stock (pre market movers, historically high volumes stocks)

Define Buy Signal:

Renko Ratio bar >= 2
MACD line >= signal line
MACD line's slope (over 5 day period) >= signal line's slope (over the last 5 periods)
exit if (MACD line < signal line) & (MACD line's slope < signal line's slope)

Define Sell Signal:

Renko Ratio bar <= -2
MACD line < signal line
MACD line's slope (over 5 day period) < signal line's slope (over the last 5 periods)
exit if (MACD line >= signal line) & (MACD line's slope >= signal line's slope)

Value Investing - Fundamental Analysis

Price is what you pay, value is what you get.

Quantitative approach if you can afford computational power. Quantitative methods:

(Joel Greenblatt's) Magic Formula
(Stanford Accounting Professor Joseph Piotroski) F-score

Excellent artcile on HFT based on actual events:

https://www.nytimes.com/2014/04/06/magazine/flash-boys-michael-lewis.html

Magic Formula

Best seller ''The little book that beats the market'':

identify wonderfull stocks at bargain price

Wonderful stocks : Return On Invested Capital (ROIC) focussing on assets generating the return

Bargain price: Earning Yield is the ratio of EBIT to Enterprise Value (like P/E ratio but capital structure independent)

Invest in the top 20-30 companies (exclude finance & insurance), accumulating 2-3 positions per month over 12 months period and rebalance each year.

F-score (mid & small caps)

Number [0-9] which assess company's financial position strength, 9 bieng best. Pick sctock [8,9] and rebalance once or twice a year.

profiltability criteria:
- positive return on assets in the current year (1 point)
- positive perating cash flow in the current year (1 point)
- return on asset higher in the current year compared to previous year (1 point)
- cash flow from operations divided by total assets greater than ROA in current year - Accruals - (1 point)
leverage, liquidity & source of funds criteria:
- lower ratio of long term debt in the current period, compared to previous - decreased leverage - (1 point)
- higher current ratio this year compared to previous year, more liquidity (1 point)
- no new shares were issued in the last year, lack of dilution (1 point)
operating efficiency criteria:
- higher gross margin compared to previous year (1 point)
- higher asset turnover ratio compared to previous year (1 point)

Trading System - building algorithmic automated

direct access to stock exchange
- high frequency
- colocation
- milliseconds downtime
API trading
- RESTful
- similar to any broker
- cheap
- test performances
- FXCM demo account https://www.fxcm.com/uk/forex-trading-demo/ u=D291103820 p=1504 token=49efa2a744bfcaa80119ebb7fd12ddb777ea2e70
running code continiously/ periodically
- crontab
- taskmgr
- Python 'timed' scripts

Fxcm way

Oanda way

API-Key 8f61d5ed3ce2f93b2cbc3d753abb6f82-e292f680364fc69b424f86fe2a817b3f

OANDA web trading terminal https://trade.oanda.com/
OANDA Python wrapper installation https://pypi.org/project/oandapyV20/

pip install oandapyV20

OANDA API (python wrapper) documentation https://buildmedia.readthedocs.org/media/pdf/oanda-api-v20/latest/oanda-api-v20.pdf
OANDA developer account intro http://developer.oanda.com/

SMA crosover

SMA Crossover Strategy Link https://www.babypips.com/trading/forex-system-20150605
Stochastic https://excellenceassured.com/trading/stochastic-oscillator-trading-indicator

Cloud

Faster, reliable, try EC2 free on AWS https://ca-central-1.console.aws.amazon.com/console/home?region=ca-central-1

i-05c31a451783d16d4 ec2-15-222-12-63.ca-central-1.compute.amazonaws.com

cd /path/to/PEMfile
chmod 400 udemy_demo.pem
chmod go-rwx filename.pem
chmod u-wx filename.pem
ssh -i "udemy_demo.pem" [email protected]

link to documentation on installing python and pip in your ec2 instance https://docs.aws.amazon.com/elasticbeanstalk/latest/dg/eb-cli3-install-linux.html

sudo yum install python37
curl -O https://bootstrap.pypa.io/get-pip.py
python3 get-pip.py --user

Basic introduction of crontab - https://www.thegeekstuff.com/2009/06/15-practical-crontab-examples/

sample crontab command to run you file every day at 9 am (monday to friday)

crontab -e

00 09 * * 1-5 /path/to/file

Sentiment Analysis

NLTK - Natural Language Processing

pip install -U nltk wordnet

Lexicon based approach

Relies on underlyng sentiment, some manually constructed

LIWC
ANEW
SentiWordNet
SenticNet
VADER

VADER - Valence Aware Dictionary and sEntiment Reasoner

Inludes abbreviations and emoticons, Successfule in social media data. VADER [-4,4] Paper by Georgia Tech Computer Department: http://comp.social.gatech.edu/papers/icwsm14.vader.hutto.pdf

VADER Python library gitgub page: https://github.com/cjhutto/vaderSentiment

pip install --upgrade vaderSentiment textblob

Machine Learning approach on Natural Language

Probalility

Naive Bayes

Probability Prediction

Training Data (labeled data)

Constructing feature matrix

TF-IDF (Term Frequency - Inverse Document Frequency)

Algorithm assigning weighting factor to get important features of a document

TF: number of times the term t appear in the doc / total number of words in the doc

IDF: log(total number of documents/ number of documents with the term t in it)

TF-IDF: TF * IDF

Application

Evolving area Commercial product lexicon based therefore un accurate Custom ML can be a product to sell Challenges:

label large volumes of news data accurately
equities need industry specific classifiers
streaming news expensive

export PATH=` echo $PATH | sed 's/:/\n/g' | grep -v mnt | xargs | sed 's/\ /\:/g' `
jupyter nbconvert --to script Section-17_98-WhatIsaNormalDistribution-Note.ipynb

Calculating & Comparing risk

Calculating Return

Annual

Annual Return = ((Daily Return + 1)^365)*100-1

Single Equity

Logarithmic Rate of Return

$$ ln( \frac{P_t}{P_{t-1}} ) $$

Multiple Equities

Simple Rate of Return

$$ \frac{P_1 - P_0}{P_0} = \frac{P_1}{P_0} - 1 $$

Normalization to 100 to visualize

$$ \frac{P_1} {P_0} * 100 $$

Dot Product from Average Retrurn by Weights

Run permutation calculations

Popular Market Indices

S&P500: '^GSPC' 500 largest US companies reflects its diversity

DowJones: '^DJI' Industrial Average 30 largest

NASDAQ: '^IXIC' Composite Grouped Securities & IT Companies

FTSE100: '^FTSE' UK

DAX30: '^GDAXI' GER

NIKKEI225: JPN

MSCI: Morgan Stanley Composite International

Calculating Rate of Return RoR

Measuring Investment Risk

Variance & Standard Deviation

Coviariance between 2 values

cov > 0 : both move in the same direction
cov = 0 : both are independent
cov < 0 : both move in opposite direction

Correlation between 2 values

Corr(returns) : Dependence between prices at different times, focus on returns of your portfolio.
Corr(prices) : focus on price levels
variance = Dot Product (weights as Vector, Dot Product (252*cov(returns), weights) )
volatility = sqrt(Dot Product (weights as Vector, Dot Product (252*cov(returns), weights) ))

Regressions for Financial Analysis

univariate regression: one variable (line = intercept + slope * x + residuals)
multivariate regression: multiple variables

Good regression

Higher percent implies more predictive is: R square = 1 - (SSR / TSS)

Markowitz Portfolio Optimization

Distribute to have low correlation

Perform many iterations for performance graph

Capital Asset Pricing Model

Capital Market Line

Beta $$ \beta_{stock,market} = \frac{\sigma_{stock,market}}{\sigma_{market}^2} $$

Beta = cov(stock, market) / variance of the market Risk that cannot be avoided using diversification

beta > 1 aggressive - performance follows the market
beta = 0 no relationship
beta < 1 defensive - performance follows slowly

CAPM $$ \overline{r_{stock}} = r_{free} + \beta_{stock,market}(\overline{r_{market}} - r_{free}) $$

Equity Risk Premium $$(\overline{r_{market}} - r_{free})$$

Example for a US stock

risk free: 2.5% (10 year US government bond yield)
beta: 0.62 (S&P500 approximation) aka average premium
equity risk premium: [4.5%, 5.5%] (historical in US)
r(stock) = 2.5% + 0.62 * 5% = 5.6%

Sharpe Ratio allows to compare 2 stocks or 2 portfolios

Sharpe ratio: $$ Sharpe = \frac{\overline{r_{stock}} - r_{free}}{\sigma_{stock}} $$

CAPM alpha

Multivariate Regression Analysis

Univariate regression has a unique beta coefficient and unique explanatory variable

Multivariate regression has a multiple beta coefficients and multiple explanatory variables

Monte Carlo Simulations for Decision Making Tool

Using historical data

On revenues

cogs: Gross Profits - cost of goods sold
opex: Operating Profits

Price Evolution

Brownian Motion

Drift aka Expected Daily Return

$$ Drift = \mu - \frac{1}{2} \cdot var = \mu - \frac{1}{2} \cdot \sigma^2 $$

Direction headed in the past (calculate average, standard deviation, and variance f daily returns) in the historical period.

Volatility aka Random Variable

$$Random Variable = \sigma \cdot Z(Random(0,1))$$

sigma is the historical variability
Zeta is the number of standard deviation from mean

$$ \frac{P_t - P_{t-2}}{P_{t-2}} $$

$$ ln(\frac{P_t}{P_{t-1}} ) = ln( \frac{P_t - P_{t-1}}{P_{t-1}} + \frac{P_{t-1}}{P_{t-1}}) = ln(\ simple.returns + 1) $$

$$ daily_returns = exp({drift} + {stdev} * z) $$

$$ where\ z = norm.ppf(np.random.rand(t_intervals, iterations)) $$

Historical Data

Log Returns: LogReturns = log(1 + returns_pct_change)

Mean Calculation : Mean = mean(LogReturns)
Variance Calculation : Variance = variance(LogReturns)
Drift Calculation : Drift = Mean - (Variance / 2)
Standard Deviation Calculation : Std = std(LogReturns)
Drift into Array Conversion
Set normalization at 95% norm.ppf(0.95)
Set array 10x2 of random values
Z Calculation : Z = norm.ppf(randomArray)
Set intervals & interations (1000,10)
r calculation : r = Drift + Std * Z
Daily Returns Calculation : DailyReturns = e^r
Price List Calculation : iterate to create matrix (first row has latest price value) m[t] = m[t-1] * DailyReturns[t]

r = drift + stdev * Z Z = norm.ppf(np.random.rand(t_intervals, iterations) dailyReturns = exp(r) = exp(drift + stdev * Z) = exp(drift + stdev * norm.ppf(np.random.rand(tIntervals, iterations))

Derivative Contracts

Forwards: price preset, one sells in N time to another
Futures: highly standardized foward contract on market place
Swaps: echange cashflows in N times of an asset (interest rate, stock price, bond price, commodity)
Options: owner buys (call options) or sells (put options) asset at a given price (strike price)

Derivative Pricing aka Black-Sholes-Merton

d1 : how much can we expect
d2 : how much we must pay

webclinic017 / kinito.finance Goto Github PK

kinito.finance's Introduction