I built Matador. It's and AI included smart trading platform. It gets signals from my external sources I feed into, then the trading platform trades based on a fine tuned trading plan and executes on it. There are a lot of guardrails and automations to assist. It started with me trading emotionally, now I am trading emotionally and unemotionally.
It's been a long journey. Version 1 started long long ago but it did not work. This was before AI had arms and legs. About two years ago, after blowing up another account clean, I rewrote our trading script and everything from scratch based on my previous learnings.
My biggest challenge, to beat Matador on performance. It's gotten way too powerful now and hard to admit but trading without it is difficult now.
My biggest lesson, backward testing is not the same as forward testing. And stay away on the FOMC days.
When you go to test your strategy in the wild you realize that it is a wild quant trading world out there and all the quants are competing against each other for the extra alpha. Some high frequency and some low.
Does it win all the time? No, not really. And the win rate is not above human average, but it has no emotions and does not care about getting the top or the bottom of the price action. It understands volatility and how to execute with a predictable long term strategy. Where it shines is it's decision making is fast, automated, and mechanical.
The signals come from different agents externally built to find alpha movers around specific criteria such as market cap, news, sector strength, market rotation, and partially from my predefined watchlist.
It trades options and stocks, and easily extendable to futures but I have not tested the futures or crypto.
All of the trading knobs are customizable as part of the automation fine tuning. Which options strategy, what expiration, how many contracts, the delta, which account, all adjustable for market conditions and risk tolerance and account size. One of my biggest mistakes developing these were setting the number of contract and expirations day input field close to each other and let the algos trade 20 contracts on a short 2 day expiration on a sideway day vs the other way. Position sizing is a guardrail in the automation now.
Options paper trading with real data is by far the most valuable tool for any quant dev. Testing and refining the algorithms till results are steady based on sensitivity of the signal momentum, trading discount, and all the other trade execution plans before going live is priceless. The amount of knowledge gained via paper trading with live data vs running against backtest archival data is an un-parallel advantage.
My new favorite feature is the automated ai research. Some traders trade everything. Most traders like to trade a specific set of stocks they follow and have a good understanding of. Active trading involves a lot of signal to noise research. Background agents give me the daily brief and trading ideas 2 hours before the market open. A short summary of what's hot and why. It does not replace active research but a blindspot mirror that keeps extra set of eyes and ideas to glimpse through.
The Claude AI agent behind the scenes goes and does the research based on my predefined criteria. I could refine the research within the existing framework for deeper research or any particular ideas I want to trade and the agents do the heavy lifting on doing the research and alpha scoring it for me based on the strength of the catalyst. The narrate mode is great, based on the OS voice tools, it is a little robotic but it is free. Maybe at some point add an open source ai reading agent that has a better voice but cannot beat free and fast right now. The research just generated ideas, how to trade is entirely based on the market conditions of that day.
There is always another trade. Position sizing is very important. Small wins add up. Wait for your repeatable setup. Trade your system. Take Profits. Leave a few runners.
Automations are the main advantage of the trading with Matador. Defined risk per trade else agent balances your position. Dont blow your entire account, 1/3 is enough. Defined trading principles or else.. you get the point. You have a gym trainer in trading.
Active Monitoring allows ai to learn on option price changes vs the strategy divergence. Right now ai does not make the incoming signal trade decisions and those are predefined with a manually locked strategy, but the recommendations will be forward tested based on the aggregate hopefully at some point. Earnings IV in this case, gives us some datapoints to explore. I might want to add some reinforcement learning for context awareness. The underlying data is what is useful for now at least for fine tuning stop losses.
Currently invite only Beta mode for experienced traders and quant developers with prior experience of algorithm trading. Trading is risky and this is not a financial advice.
Roadmap for V4 is pretty sweet. Stay tuned. Matador
