After 6 months of live trading alerts, my algo has an 84% win rate and I'm improving that with chart formations and data analysis.
At the end of last year I started running my algorithm on live data and trading the alerts with much success.
After 6 months here are the stats:
- Win Rate 84% * Average Win 3% * Average Loss -6.8% * Average Hold 3.5 Days * Performance Simulator 995.4% Return * I've gained 72% manually trading the alerts * I started auto trading paper in Alpaca and gained 33% in 1 month
My algorithm capitalizes on the Momentum Effect by updating a list of high performing stocks every 2 weeks and monitoring RSI for oversold - a mean reversion strategy. I split the stocks into 3 categories based on their momentum and performance.
I also started overlaying as much data into each alert based on dynamic information at the time of the alert. This data includes things like:
- Time of day * Price * RSI * News Sentiment * Analyst Price Target * Analyst Sentiment * Support and Resistance Level - Sentiment * Chart Formation - Sentiment * Volume - PACE Direction, RVOL * Beta * Distance to price target * Distance to man reversion * PEG Ratio * Performance thresholds achieved * Stock Score that considers 8 data points
I store all this data and then use AI to perform statistical analysis on all of the alerts and their outcome to determine how to best modify the strategy as well as place a rating on each alert to instruct the algorithm on how to trade - which alerts to favor, etc.
The data analysis allows me to extract this kind of data to tweak how the algorithm works moving forward.
Essentially, I've greatly reduced the amount of data from THE ENTIRE STOCK MARKET to only THE DATA OF MY ALGORITHM. With a smaller set of data, I have insight into very similar numbers, but can separate outliers or focus on where Alpha lies.
Let me know what you think of this approach of using statistical analysis on Algorithmic data.