Why?

• Because a person new to the project has no way to know what sources of financial data she can use.
• Because browsing the different quotes-from-xxx.service.ts classes gives no hint about what data they actually process.

Condition of satisfaction

• Redact the concerned section of the README.md so that the potential user understands how to list the financial sources he can use.
• Avoid listing the financial sources in README.md, because they'll change, and we'll never remember to update the documentation.
• Improve the code in src/app/services/quotes to match the README.md in a satisfactory way.

Additional notes

• This is not about adding new features.
• This is about making the code more readable by someone who has very superficial programming skills.

Why?

• Because it is mentioned in the book I am reading.
• Because I can compare the results in my simulation with those of the book.

Condition of satisfaction

• Obtain same results as in the book.
• Or invalidate the book.

Why?

• So anyone can start using the project.
• To acquire a critical point of view about the project.

Condition of satisfaction

• Anyone, with very superficial programming knowledge, can download financial data, add them to the project, and launch a simulation based on an existing example.
• Write a reusable example of a simulation.

Why?

Fetching online data is nice because you always have it fresh, but there are some shortcomings:

• Online data are usually not meant to be consumed externally, so the format can change without notice, and your application will stop working.
• Most of online data requires that you login to a banking system, and those are heavily defended against data retrieval from third party applications.
• Typescript, Javascript and Internet security are very worried when they see a script from one website trying to access data from another website.

The bottom line is, fetching online data has too many disadvantages.

Condition of satisfaction

• Keep data in separate folders, using the name of the provider and the instrument ISIN, so they can be found easily.
• Keep the data untouched, just saved from the provider.
• Implement the mapping so the data can be used as native Stock and StockData.

Why?

• Because trading costs have an impact of the return of any strategy.
• An important trading cost is the bid-ask spread.
• The bid-ask spread is not generally available from quote providers.
• Therefore they need to be estimated.

Condition of satisfaction

• quote services are able to include an estimation of the bid-ask spread while loading the data.
• Big-ask spread is used in all order simulations.
• Use the Farshid Abdi, Angelo Ranaldo 2017 paper:
  • A Simple Estimation of Bid-Ask Spreads from Daily Close, High, and Low Prices
  • https://academic.oup.com/rfs/article/30/12/4437/4047344

Why?

• Because they offer more options.
• In particular, Apex makes easy to have two parallel charts.
• And also Apex is stronger on markers.

Condition of satisfaction

• Use Apex Charts.
• Implement an options for two charts.
• Use a marker to signal the Bull / Bear changes

Why?

• Because it is a simple way of including currency risk to any instrument.

Condition of satisfaction

• In asset configuration, add a exchangeRate, pointing to exchange rate data.

Why?

Where to invest when market goes bearish?

We want to compare these two options:

• AGG with dividends and the risk of adverse USD / CHF exchange rate
• CHF without any dividends but also without any risk.

Condition of satisfaction

Make a simulation comparing the USD/CHF exchange rate and the following two strategies:

• Superthon / Candle Filter with SPY for bull market and and AGG for bear market.
• Superthon / Candle Filter with SPY for bull market and and CHF for bear market.

Why?

• Because ATR is the measure for detecting that something is not normal in the price evolution.
• Because, right now, Stop Loss Market Timing is useless.

Condition of satisfaction

• Specify the start of BEAR by a multiple of ATR.
• Specify the end of BEAR by a moving linear regression.

Why?

Because reinvesting dividends happens inevitably if you use market timing. Consider a strategy set with reinvestDividends=false in the following scenario:

• In BULL period, as dividends are not reinvested, they're accumulated in cash.
• At the beginning of the BEAR period, all assets are sold and more cash is added to the account.
• At the beginning of the next BULL period, all cash is invested in the targeted asset.

So the dividends are reinvested, independently of the value of reinvestDividends, and there is no fix, as the cash from dividends is mixed with the cash from selling the assets.

Condition of satisfaction

• Expand the RegularTransfer class so we can have a Transfer that considers the regular withdrawal (as of now) plus transferring the dividends when there are any.
• Also add the possibility of specifying a % of dividends that would be lost in charges or taxes.

Why?

• Because the normalized value is visible anyway against the vertical axis.
• Because the normalized value is only useful to visually compare performances, but it is not a real value.

Condition of satisfaction

• The tooltip displays the real reported value, not the normalized one.

Why?

• Until now, you can create a scenario, change the simulation component, and see it. It is simple and satisfactory of a single individual, just wanting to test something.
• But what if I want to keep track of my scenarios to show to someone else? Or just not forget about them?
• What if there are several people committing to the same repository?

Condition of satisfaction

• Add some kind of navigation, so it is possible to group scenarios in different pages.
• Run only scenarios when the become visible, so as to avoid a long start up time.

Why?

Because sometimes it is impossible to find the historical data for dividends (the SP500 is a special case). Just using the 'total return' index would suit a simulation where we reinvest all dividends, however this is hardly convenient because of several reasons:

• Often dividends have tax impositions that depend on your tax residence, so you can't reinvest all dividends.
• If you want to obtain a revenue from your portfolio, you may want to use the dividends when possible, to reduce operational costs.

Condition of satisfaction

• Implement a source of dividends that would accept the total return index.
• For testing, use the SP500, as all required data is available: normal index quotation, total return quotation and dividends.

Technical details

• https://www.indexologyblog.com/2017/02/07/index-basics-calculating-an-indexs-total-return/

Why?

• To verify if it works.
• As a prior step, before implementing the investment strategy by Andreas Clenow in his book "Stock on the move".
• In particular, to implement the exponential regression and getting familiar with its behavior.

Condition of satisfaction

Implement a market timing based on the following strategy:

• Buy when the exponential regression x correlation is above a value.
• Sell when asset value moves beyond a number of times its ATR? Or maybe when exponential regression is below a value? Explore possibilities.

Why?

• Because this is how lots of investors do: invest in stock indexes during BULL periods, and switch to bonds indexes during BEAR period.

Condition of satisfaction

• B&H with market timing will have not one, but two securities: A, and B
• During Bull periods, the strategy invests everything in security A.
• During Bear periods, the strategy sells position on A, and invests everything in B.

Additional indications

• Write unit tests for B&H with market timing.
• Remove the older B&H, as all functionality still works with the new one.

Why?

• So it is possible to compare different strategies.

Condition of satisfaction

• It is possible to simulate two different setups on the same graph.

Why?

• Current implementation of stop loss protects against heavy losses.
• However, it also prevents to make nice profits.
• And this is because, on average, re-entering the market is not that easy.

Condition of satisfaction

• Explore the advantages of using RSI for reentering.

Why?

• Because it has an impact on investment yields.
• Because I happen to use IBKR.

Condition of satisfaction

• Implement both FIXED and TIERED price structures.
• https://merelycurious.me/post/interactive-brokers-fixed-vs-tiered-commissions
• https://interactivebrokers.com/en/index.php?f=1590

Why?

• It is a nice scenario to verify
• We've read that book, "Stocks On The Move", by Andreas Clenow

Condition of satisfaction

• Follow the description of the scenario, to back test 1 year.
• Define an universe
  • That will be the S&P500 Components.
  • Take in count also companies entering and leaving the index.
  • https://en.wikipedia.org/wiki/List_of_S%26P_500_companies
• Download data:
  • Use data from https://marketstack.com/
  • Data up to 1 year is free, this is why we limit to 1 year for the moment.
• Implement all the selection rules described in the book:
  • Momentum (as CAGR corrected with R2)
  • 15% gap.
  • Market timing to the index.
  • ATR to distribute risk.

Additional details:

• https://github.com/Auquan/Tutorials/blob/master/momentum_backtest_making_money.ipynb
• https://github.com/Auquan/Tutorials/blob/master/momentum_backtest_losing_money.ipynb

Why?

• Because this is what the information is meant for.
• Because this information is provided by Yahoo and Market stack.

Condition of satisfaction

• Use the adjusted price to calculate dividends and splits.
• When downloading newer data from yahoo or market stack, correct the previous adjusted price, so the whole series is correctly held.

Why?

• Because 12 periods of 1 month seems a bit arbitrary.

Condition of satisfaction

• Add a periodLength for MONTHLY, SEMIMONTHLY and WEEKLY.
• Compare with SP500 and IBEX35 if it is better to use mean over a shorter period, or use a different period length.

Why?

• Because it is a very common strategy.

Condition of satisfaction

The strategy has the following properties:

• X% cash.
• Y1% asset 1
• Y2% asset 2
• etc.
• Where X% + Y1% +Y2% + ... = 100%

Why?

• Because SPY concerns only US market.
• ACWX stands for ""

Condition of satisfaction

Run simulations on ACWX taking into account:

• Dividends
• USD / CHF exchange rate.

Additional details

• https://www.ishares.com/us/products/239594/ishares-msci-acwi-ex-us-etf#/

Why?

• Being consistent in wording is a never ending task.
• https://english.stackexchange.com/questions/107902/normalise-or-normalize-british-english

Condition of satisfaction

• Use 'normalise' everywhere.

Why?

• After implementing the fixed allocation strategy in #46, we need to check how it works with some approved model.
• Also, to confirm the claims of the Ray Dalio.

Condition of satisfaction

Replicate the results of:

• http://www.lazyportfolioetf.com/allocation/ray-dalio-all-weather/

Why?

Because it looks reasonable to buy when the P/E ratio is low, and sell when it is high.
Or is it?

Condition of satisfaction

• Add the SP500 P/E Ratio to the assets.
• Create a market timing indicator based on it.
• Verify the hypothesis.

Why?

• #14 implemented a navigation
• But... it is ugly

Condition of satisfaction

• Use bootstrap
• Make it usable and functional, but don't waste time on design

Why?

• Because it is possible.

Condition of satisfaction

• Each commit runs the tests

Why?

• Because this reflects reality.
• Otherwise it is impossible to compare the return on different indexes.

Condition of satisfaction

• Assets have a domestic currency.
• Accounts have a base currency.
  • NAV is always expressed in base currency.
• When buying or selling an asset, execute the change between:
  • Account's base currency.
  • Asset's domestic currency.
• Import currency quotes.
  • In the model, allow for a spread.

Why?

• I want to test the Superthon and RSI filters using a smoothed version of the instrument quotes.
• LOESS and LOWESS are very widely used smoothing algorithms.
• LOWESS is the simplest of the two.

Condition of satisfaction

• Implement the LOWESS algorithm as another of the statistic calculations.
• Implement a market timing based on the candle filter and the LOWESS algorithm.

Additional indications:

• Fiitting a curve to data, aka lowess, aka loess
  • Explains what is LOESS/LOWESS, and why it works.
  • https://youtu.be/Vf7oJ6z2LCc
• LOESS/LOWESS Algorithm, of which seem to exist different versions, depending on when to apply the weights.
  • First regression without any weights, and then apply both W and G:
    • https://www.weisang.com/en/documentation/loessandlowessalgorithm_en/#
  • First regression using only the W weights for X elements, and then apply the G weights:
    • https://www.ime.unicamp.br/~dias/loess.pdf
• The inverse of a matrix:
  • Because it is needed to follow the matrix version of the LOWESS algorithm.
  • https://byjus.com/maths/inverse-matrix/
• The derivative of a matrix:
  • Because it is needed to follow the matrix version of the LOWESS algorithm.
  • https://youtu.be/e73033jZTCI
• The matrix version of the linear regression algorithm:
  • Because the LOWESS is just a variation of it.
  • https://youtu.be/K_EH2abOp00
  • Be aware that there is an error in the 3rd matrix differentiation rule (at 11:26) - where it says y = xA, it should say y = xTA.

Why?

• As of currently, there is too much aggregated code to effectively unit test the filters.
• I have doubts about the implementation of EMA, which is the base for MACD.

Condition of satisfaction

• Be 100% sure that the algorithm works, by comparing a unit test with some official documentation or tutorial.
• In particular, isolate the EMA calculation from the EMA filter.

Why?

• Because often the dividends are not available from the same source as the quotations. This is particularly true for indexes like IBEX35 or SP500

Condition of satisfaction

• Use the securities configuration file to point to the dividends information.
• The format of the dividends file is independent from the format with quotations.

Why?

• Because current vocabulary has become inconsistent.

Condition of satisfaction

• Don't change functionalities.
• Only change the name of main classes

Some ideas

• Data retrieved from http://finance.yahoo.com, or http://six-group.com, are Quotes. Some of the quotes represent indexes, other represent securities like funds, bonds or stocks.
• Therefore, the Stock class should be Quote, and the StockData should be QuoteHistory.
• Quotes have a name, but as not all quotes are securities, not all quotes have a ISIN (International Security Identification Number), therefore the isin property that is ubiquitous through the code should be name.
• The services that return quotes are all placed in the services/stock folder. It should be services/quotes.

Why?

• Because Stop Loss market timing implemented in #40 is not very satisfactory.

Condition of satisfaction

• Implement a market timing based on the RSI indicator.

Why?

Because time passes, and we need to keep up

Condition of satisfaction

• Upgrade angular core and cli to 15
• Upgrade the node versions of github workflow to 16 and 18

Technical hints

Use the official page for upgrade instructions:

• https://update.angular.io/?l=3&v=10.2-11.0

Apply the --force flag

Why?

• Because the Candlestick contains the open, close, highest and lowest price, that are provided by almost all quote providers.
• Because when you have stocks from different stock exchange sites and the bank holidays are not the same, then candlestick make it easier to merge, concatenate or extrapolate.
• Because Candlestick make easier to compute the dividends comparing the PR and TR versions of the same index, which is wanted for #13.
• Because sometimes the historical quotes are available only on weekly basis. A candlestick better represents what happened during the week that a single closing price value.

Condition of satisfaction

• Have a domain class called the Candlestick, containing the open, close, highest, lowest prices.
• Populate those values in all quote services.
• In HistoricalQuotes, fill all the gaps, so all instruments are present at all instants.
• Use the candles in the Superthon filter, and others.
• Keep all functions the same.

Why?

• Because it is a very well known provider of financial data.
• Because it provides SP500 and IBEX-35 data since inception.

Condition of satisfaction

You can download any data (for example: https://finance.yahoo.com/quote/%5EIBEX?p=%5EIBEX) from yahoo, save it in the appropriate folder, add it to the configuration file, and use it in a simulation.

Why?

• According to some authors, it is more efficient than MACD for indexes.
• Because it contains the same initial steps as the MACD, that is already implemented, so there is good opportunity of reusing code.

Condition of satisfaction

• Implement a Market Timing based on long EMA and short EMA.

Why?

My goal is to write a back-testing blog with plenty of simulation examples:

• Editing the pages as part of an application is long and painful, and it only makes sense if pages are dynamic.
• All pages are completely static, what is dynamic is only the simulation component.
• WordPress is the most widely used framework for blogs.

Condition of satisfaction

• Package the angular-backtest application as a WordPress pluging that enables using the <simulation> tags.
• Devise a simple way to make the quotes available.
• Devise a simple way to update.
• Write a short user manual to include the plugin.

Technical details

• https://laternastudio.com/blog/including-an-angular-7-app-inside-wordpress-laravel-or-any-php-page/

Why?

• To compare with USD.
• To reduce the currency risk.

Condition of satisfaction

• Find equivalent ETF in CHF.
• Simulate with different weights.

Additional information

• http://www.lazyportfolioetf.com/allocation/ray-dalio-all-weather/
• https://www.reddit.com/r/eupersonalfinance/comments/kuaq0o/all_weather_portfolio_in_europe/

Why?

• Because they've changed their formats

Condition of satisfaction

Yahoo and Investing can be downloaded again via

npm run refresh-quotes SPY AGG AUG

Why?

Pros:

• It looks nicer than charts.js
• It looks like they can handle a lot of data.

Cons:

• Is it paying?

Condition of satisfaction

• Implement a chart report that uses Highcharts.

Technical details

Follow this tutorial:

• https://www.highcharts.com/blog/tutorials/highcharts-and-angular-7/

Why?

• Because it is pretty popular.

Condition of satisfaction

• Have it implement the MarketTiming interface, so it can be used interchangeably to any other implementation.
• Test it with https://investsolver.com/calculate-macd-in-excel/

Why?

• Because it is nice.
• Because the Bear/Bull periods are difficult to spot when investing in bonds.
• Because Bear/Bull doesn't fit in any scale.

Condition of satisfaction

Display a vertical line with the mention Bear or Bull.

Technical details

See demo here:

• https://valor-software.com/ng2-charts/#/LineChart

Why?

Until now we've been downloading stock data from SIX, YAHOO and INVESTING but, besides being a time consuming process, they only offer historical data of currently listed stock:

• Until now we've been simulating on indexes or ETF tracking indexes, so it was not an important shortcoming.
• As soon as we simulate individual stock, we will incur into "survivor bias".
• To back test momentum strategy, we'll need hundreds of tickers, and downloading them individually is not practicable.

Condition of satisfaction

The idea is not to directly use data from Marketstack during simulation, because there is a limit of the monthly requests.

Instead, it is better to have the downloaded data stored locally:

• Having a file with all the tickers to be downloaded
• The script downloads the data for all tickers, and saves it into a folder:
  • If the file doesn't exist yet, create it for the ticker.
  • If the file exists, then append the new data.
• Use yahoo formatting, so it is directly possible to use the data in simulation.

Technical details

• Service is available at https://marketstack.com/
• Try to use server-side node.js?

Why?

We've implemented so far: the Candle filter, the EMA filter and the MACD filter:

• But are they any good?
• Which one is the best:

Also, we want to use the application, so we can see how handy it is.

Condition of satsifaction

Compare the three filters with the following indexes:

• Europe:
  • DAX
  • CAC40
  • IBEX
• Overseas:
  • FTSE
  • SP500
  • ASX
• Asia:
  • NASDAQ

Establish:

• The best configuration for each filter
• The best filter, overall.

Keep:

• Pages to record the work.

Why?

• Because it is said that a good entry indicator is not a good exit indicator, and vice versa.
• Because it looks like the MACD indicator is a good entry indicator.

Condition of satisfaction

• Make a Market Indicator that is the composition of two other market indicators.