Polyphony

Simple framework for portfoliio rebalancing and backtesting.

Version: Alpha

Goals and Use Cases

Synthetic and Alternative Data. Enable Monte Carlo simulations. Feed in simulated data to play out different scenarios. Use alternate pricing (e.g., historical Bitcoin prices) instead of newer BTC ETFs to conduct older backtests.
Generative Strategies. Take a strategy framework (e.g., 200-day SMA > current price) and test it against thousands of tickers.
Strategy Variable Optimizations. Optimize indicator variables by generating many combinations of inputs. For example, compare the performance of RSI 10-60 days > 60-80%.
Archiving and Backup. Save and store algorithms "offline" as a disaster recovery strategy.
Workflow Automation / Switchboard Building. Create Symphony JSON to use as a switchboard for multiple Symphony URLs. "Mute" branches instead of swapping BIL in/out to test different branches.
Observability, Debugging, and Tracing. See what assets will be chosen given a scenario, what branches were selected to generate an allocation, and what calculations were made for an indicator.

Version Alpha

Code is currently in alpha.

Focus right now is on consistency and trust of backtest calculations.

The code is changing frequently and will do so with little or no communication.

The code is ugly. Functionality is being layered in at the expense of code accessibility/usability to get a better understanding of workflows and how components interact with each other to help aid in a future redesign.

Example

import datetime
import yfinance as yf
import src.utils as utils
from src.allocate import allocate

# if current price of SPY is > 200d moving average of SPY, buy $SPY, otherwise buy $BIL
algo = ['ifelse',
  ['gt', ['now', ['asset', 'SPY']], ['ma', ['asset', 'SPY'], 200]]  ,
  ['asset', 'TQQQ'],
  ['asset', 'BIL']
]

# pick a trading day
date = datetime.date(2024, 2, 20)

# look up historical data data
price_data_start = utils.subtract_trading_days(date, 201) # need >= 200 days of data for SPY 200d MA
price_data_end = (date + datetime.timedelta(days=1))
price_data = yf.download('SPY TQQQ BIL', start=price_data_start, end=price_data_end, progress=False)

# get allocation for 2024-2-20
allocation = allocate(algo, date, price_data)

print(allocation)
# {
#   'TQQQ': 1.0
# }

Setup

Dependencies

pip install -r requirements.txt

Test

Run smoke tests

python smoke.py

Unit tests

python tests

Reference

Allocations

Allocation

A map of tickers and percentages (in decimals). The sum of percentages will always add up to 1.

{
  'SPY': 0.5
  'XLE': 0.5
}

Generate an Allocation

import datetime
import yfinance as yf

# define an algo
algo = ['wteq', [
  ['asset', 'SPY'],
  ['asset', 'XLE']
]]

allocation = allocate(algo, date, trading_data)

`allocate`

`preprocess`

Algo

An Algo is a serialized simple representation of a daily trading algorithm.

Simple Example

# Equal allocation between SPY, QQQ, and XLE
['wteq', [
  ['asset', 'SPY'],
  ['asset', 'QQQ'],
  ['asset', 'XLE']
]]

More Complex

# My Algo
# if the current price of SPY > 200d moving average of SPY then
#     allocate quality between SPY, QQQ, XLE
# otherwise
#     invest only in BIL
['group', 'My Algo', [
['ifelse', 
  ['gt', ['now', ['asset', 'SPY']], ['ma', ['asset', 'SPY'], 200]],
  ['wteq', [
    ['asset', 'SPY'],
    ['asset', 'QQQ'],
    ['asset', 'XLE']
  ]],
  ['asset', 'BIL']
]
]]

Primitive Objects

String 'foo' or 'bar'
Number 1, 2.3, or -0.4
Boolean true or false
List [] or ['one', 2, false]

Predicate Object

A Predicate is a List where the first entry is a name (String) and at least one additional Object which arg called arguments or args. A Predicate always returns an Object.

['op', 'arg1', 'arg2'] -> Predicate

Using a specific example

['asset', 'SPY'] -> Block

Core Predicates

Block
Indicator
Comparator

Block

A Block is an abstract allocatable collection of one or more assets.

There are a few types: asset, ifelse, wteq, filter, group

`asset`

['asset', String ticker]

Examples

# Example 1
['asset', 'SPY']

# Example 2
['asset', 'QQQ']

# Example 3
['asset', 'XLE']

`wteq`

['wteq', Block[] blocks]

Examples

# Example 1
['wteq', [
  ['asset', 'SPY'],
  ['asset', 'QQQ'],
  ['asset', 'XLE']
]]

# Example 2
['wteq', [
  ['asset', 'SPY'],
  ['wteq', [
    ['asset', 'QQQ'],
    ['asset', 'XLE']
  ]]
]]

# Example 3
['wteq', [
  ['asset', 'SPY'],
  ['ifelse',
    ['gt', ['rsi', ['asset', 'SPY'], 15], ['number', 80]],
    ['asset', 'BIL'],
    ['asset', 'TQQQ']
  ]
]]

`ifelse`

['ifelse', Conditional conditional, Block true_block, Block false_block]

Examples

# Example 1
['ifelse',
  ['gt', ['now', 'SPY'], ['ma', 'SPY', 200]],
  ['asset', 'BIL'],
  ['asset', 'TQQQ']
]

# Example 1 (annotated)
['ifelse',
  # conditional
  # current price of SPY > 200d average of SPY
  ['gt', ['now', ['asset', 'SPY']], ['ma', ['asset', 'SPY'], 200]],

  # true_block
  # if conditional true, use BIL
  ['asset', 'BIL'],

  # false_block
  # if conditional false, use TQQQ
  ['asset', 'TQQQ']
]

`filter`

['filter', Block[] blocks, FilterIndicator indicator, FilterSelect select]

Examples

# Example 1
['filter',
  [
    ['asset', 'SPY'],
    ['asset', 'QQQ'],
    ['asset', 'XLE']
  ],
  ['cr', 10],
  ['top', 1]
]

# Example 1 (annotated)
['filter',
  # list of Blocks
  [
    ['asset', 'SPY'],
    ['asset', 'QQQ'],
    ['asset', 'XLE']
  ],

  # filter sort
  # sort blocks by 10 day cumulative return
  ['cr', 10],

  # filter select
  # select top 1 of sorted blocks
  ['top', 1]
]

`group`

['group', String name, Block block]

Examples

# Example 1
['group', 'My Algo', 
  ['wteq', [
    ['asset', 'SPY'],
    ['asset', 'QQQ'],
    ['asset', 'XLE']
  ]]
]

Indicator

An Indicator runs a calculation on an asset and returns a Number.

There are a few types: now, car, ma, mar, number, rsi

`now` - Current Price

['now', String ticker]

Examples

# Example 1
['now', ['asset', 'SPY']]

`cr` - Cumulative Return

['cr', String ticker, Number window_days]

Examples

# Example 1
# 10d cumulative return of SPY
['cr', ['asset', 'SPY'], 10]

`ma` - Moving Average

['ma', String ticker, Number window_days]

Examples

# Example 1
# 10d moving average of SPY
['ma', ['asset', 'SPY'], 10]

`mar` - Moving Average Return

['mar', String ticker, Number window_days]

Examples

# Example 1
# 10d moving average return of SPY
['mar', ['asset', 'SPY'], 10]

`Number` - Number

Used in a Comparator as a fixed value.

['number', Number number]

Examples

# Example 1
# 10d moving average return of SPY
['number', 99.5]

# Example 2 (annotated)
['ifelse'.
  # lt (or <) only accepts an Indicator as arguments
  # `number` is used in cases when a comparison is made against a fixed number
  ['lt',
    # 10 day RSI of SPY < 90
    ['rsi', ['asset', 'SPY'], 10],
    ['number', 90]
  ],
  ['asset', 'SPY'],
  ['asset', 'BIL']
]

`rsi` - Relative Strength Index

['rsi', String ticker, Number window_days]

Examples

# Example 1
# 10d relative strength index of SPY
['rsi', ['asset', 'SPY'], 10]

Comparator

A Comparator compares two Indicators and returns a Boolean.

There are a few types: gt, gte, lt, lte

`gt` - > Greater Than

['gt', Indicator lhs, Indicator rhs]

Examples

# Example 1
['gt',
  ['ma', ['asset', 'SPY'], 10],
  ['ma', ['asset', 'SPY'], 60],
]

# Example 1 (annotated)
# 10d moving average of SPY > 60d moving average of SPY
['gt',

  # lhs (left hand side)
  # 10d moving average of SPY
  ['ma', ['asset', 'SPY'], 10],

  # rhs (right hand side)
  # 60d moving average of SPY
  ['ma', ['asset', 'SPY'], 60]
]

`gte` - >= Greater Than or Equal To

['gte', Indicator lhs, Indicator rhs]

Examples

# Example 1
['gte',
  ['ma', ['asset', 'SPY'], 10],
  ['ma', ['asset', 'SPY'], 60],
]

# Example 1 (annotated)
# 10d moving average of SPY >= 60d moving average of SPY
['gte',

  # lhs (left hand side)
  # 10d moving average of SPY
  ['ma', ['asset', 'SPY'], 10],

  # rhs (right hand side)
  # 60d moving average of SPY
  ['ma', ['asset', 'SPY'], 60]
]

`lt` - < Less Than

['lt', Indicator lhs, Indicator rhs]

Examples

# Example 1
['lt',
  ['ma', ['asset', 'SPY'], 10],
  ['ma', ['asset', 'SPY'], 60],
]

# Example 1 (annotated)
# 10d moving average of SPY < 60d moving average of SPY
['lt',

  # lhs (left hand side)
  # 10d moving average of SPY
  ['ma', ['asset', 'SPY'], 10],

  # rhs (right hand side)
  # 60d moving average of SPY
  ['ma', ['asset', 'SPY'], 60]
]

`lte` - <= Les Than or Equal To

['lte', Indicator lhs, Indicator rhs]

Examples

# Example 1
['lte',
  ['ma', ['asset', 'SPY'], 10],
  ['ma', ['asset', 'SPY'], 60],
]

# Example 1 (annotated)
# 10d moving average of SPY <= 60d moving average of SPY
['lte',

  # lhs (left hand side)
  # 10d moving average of SPY
  ['ma', ['asset', 'SPY'], 10],

  # rhs (right hand side)
  # 60d moving average of SPY
  ['ma', ['asset', 'SPY'], 60]
]

Dev Environment

Setup

asdf

Download and install asdf https://asdf-vm.com/guide/getting-started.html

python

Using asdf add python plugin

asdf plugin-add python

Using asdf install python (via .tool-sersions file)

asdf install

Create venv

python -m venv venv

Load venv

source venv/bin/activate

Install python dependencies

pip install -r requirements.txt

Run

Make sure venv is active

source venv/bin/activate

Run

python run.py

Test

pytest

howardr/polyphony

Polyphony

Goals and Use Cases

Version Alpha

Example

Setup

Dependencies

Test

Reference

Allocations

Allocation

allocate

preprocess

Algo

Primitive Objects

Predicate Object

Core Predicates

Block

asset

wteq

ifelse

filter

group

Indicator

now - Current Price

cr - Cumulative Return

ma - Moving Average

mar - Moving Average Return

Number - Number

rsi - Relative Strength Index

Comparator

gt - > Greater Than

gte - >= Greater Than or Equal To

lt - < Less Than

lte - <= Les Than or Equal To

Dev Environment

Setup

asdf

python

Run

Test

`allocate`

`preprocess`

`asset`

`wteq`

`ifelse`

`filter`

`group`

`now` - Current Price

`cr` - Cumulative Return

`ma` - Moving Average

`mar` - Moving Average Return

`Number` - Number

`rsi` - Relative Strength Index

`gt` - > Greater Than

`gte` - >= Greater Than or Equal To

`lt` - < Less Than

`lte` - <= Les Than or Equal To