Algo_Signals

The Four Headless Horseman

An AlgoTrader that allows the user to build / test / deploy their strategy without needing the coding background necessary to build from scratch

Stock / Crypto Screener (@Remy D initial design)

• Allow the user to select which market they want to choose (stocks, or crypto currency)
• Once the market is selected we will screen the market to come up with the most volatile stocks/cryptos for each trading day to put through the algo
• Choosing stock screener to filter stocks by:
  1. 20% premarket gain
  2. 3x relative volume over the last 10 days
  3. Breaking news or news article
  4. Number of outstanding shares (small, mid or large cap stocks)
  5. Call/Put ratio (maybe?)
  6. Market sentiment (stocks in the market gaining vs losing)

Determine what is classified as a “winning trade” for the buy / sell in our historical dataset

• Are we looking for short term trades that exceed the fees we would pay to take the trade? aka anything that is remotely profitable
• OR, are we wanting only “home run” trades of 5% or more within a certain time frame

Customer input technical indicators from a list (@Andrew initial design)

• The customer can select what technical indicators they would like the algo to consider when trading the stocks/cryptos
• These could be pulled through the “TA” package in Python or created by us as “modules” that can be called / plugged into the code
• The indicators would then be put into the dataset along with the stock historical price data

Entry / Exit Strategy (@Scott Slusher initial design)

• This would allow the user to select between a fixed entry / exit or a dynamic entry / exit based off of risk tolerance
  • Example:
  • Fixed:
    • Entry: Invest $1000 or 1,000 shares regardless of price or volume
    • Exit: A stop loss set at 5% below entry with an exit set 10% above entry
  • Dynamic:
    • Entry: Kelly Criterion based off a formula that takes in the algo’s P/L Ratio, Win %, Lose %, Capital available to trade, OR 10% of average trading volume over last 5 periods (whichever is smaller)
    • Exit: Based off the 2.5X ATR (Average True Range) of the stock for the given time period as a stop, then use support/resistant levels as “soft targets” for exits

Build Algo

• Basic structure will be provided (train - test - split, etc)
• User will need to determine if they would like to tweak the amount of layers in the neural network and how many neurons per layer
  • this can also be “optimized” per Remy with a for loop to run several variations to determine the best performance

Back Test

• Allow the algo to paper trade with live data (crypto 24/7/365, stocks during open hours or historical data from last trading day) to further confirm / deny if the algo is working

Show all available metrics

• Allow the user to determine if they would like to Deploy the algo that was created, and/or restart at the beginning and build another algo

This site is open source. Improve this page.