Verifying Momentum Strategy

The goals of this project are:

1. Verify dual momentum by verifing both Absuolte momentum and in combination with relative.
2. Build set of reusable primitives

Steup

create a virtual environment

[WIP]

Required Modules

Required modules are in requirements.txt.

To install all required module run the following command:

pip install -r requirements.txt

Folder structure

• data\tickers folder includes all the tickers downloaded
• data\output includes saved results of verious runs. Format used is {tickername}_ABS_a{action}_lb{lookback size}.
  • Where ABS stands for the name of the algorithem, Absolute momentum, 'a' is the number of days between actions and 'lb' is the size of the lookback window. So SPY_ABS_a5_lb253, means SPY absolute momentum with action evaluated (and taken) every 5 trading days and the lookback window size is 253 trading days
  • for example, QQQ_ABS_a21_lb253
• each algorithem result csv file includes the following coloumns: Date | Close | Daily return | lookback window | ABS Close | RISK.
  • Date:
  • Close: (do we need to use adjusted close?)
  • Daily retun: is precentage change from previous close;
  • lookback window: is percentage change between close[i] / close[i- lookback]
  • ABS Close: is the absolute momentum result
  • RISK: shows action points and the result of the momentum evaluation

todo list

• should the algorithem use Adj Close or Close?

• compare algo to running using adj-close / close (return, MaxDD, )

• wrap absolute momentum algo in a method and in its own file

• create a notebook for basic experiments

• wrap all helper functions into module, that can be easily used from notebook and code

• chagne graph to show logaritmic values

• chagne calc_absolute_momentum to accept a flag using Close or Adj Close

• change save_symbole Abs Close to use algorithem naming conventin ABS_a{action}_lb{lookback}_start{start_date}_end{end_date}

• For each algo run, write a new line in log file, append to top,with results, including: returns; MaxDD; Sharp, etc...

• calc algorithem return, starting from a fix postion of a '100' to show relative of multiple assets

• test all algorithms to start from the same date. That is from the max lookback window (so they all start from the same starting point.

• consider using one file for all different permutation?

• add plot, to show spy vs. Abs + action points

• calc return, max DD (all up and per year); Alpha; Beta; Sharp; etc...

• run experiment on ABS Momentum across all permutation of action and lookback and compare results between different permutations action[1,5,10,21] X Lookback[1,5,10,21,63,127,190, 200,253]

• asstes to test: SPY, QQQ, AAG LQD

• add cash trigger (sma 200 S&P to exit, when do we return?)

• check the rate of change in the lookback window at crashes before/after... hypothesis for daily change rate or lookback is bigger before riskoff due to carsh

• consdier adding some sort of date/time to tickers folder of naming convention

• download list of all S&P tickers

• find top winners/ looser in S&P per action & window then compare returns

• run cash trigger on stocks (AAPL; MSFT; from the cash trigger book)

• create algorithem for relative momentum, compering

reading material

• https://github.com/matt-dong/Absolute-Momentum
• pandas has a rank function (yay) df.rank(axis=1) --> axis=1 means on columns