#tradingmodel

Ever wondered how traders let algorithms do the work? Here’s the process broken down into simple steps you can actually follow:

💡 1. Start with an idea: Find a trading edge—momentum, mean reversion, or something unique to your market.

📊 2. Collect and clean data: Good data = good model. Use clean, complete, and reliable price and volume data.

⚙️ 3. Create trading rules: Turn your idea into clear, rule-based logic. When to buy, when to sell, when to stop.

🧠 4. Backtest your strategy: Run your model on past data to see how it performs. Watch out for overfitting!

💰 5. Manage risk: Set limits, manage position sizes, and always control your losses.

🚀 6. Simulate before going live: Paper trade first. See how your algorithm performs in real conditions.

🔍 7. Keep improving: Markets evolve—your model should too. Monitor, adjust, and refine.

This tumblr examines stock market trading models. It exists solely to satisfy my interests in investing, programming, and data analysis. The aim is to improve my knowledge in these areas through the investigation of models I find on Twitter, Tumblr, and anywhere else my browser lands. 

Keep this in mind: Anyone can write a blog on Tumblr. I do not have any magic beans. Do your own verification. Statistical analysis requires tools and methods of which I am only vaguely familiar (not to mention unqualified to undertake). 

Don’t trade based on anything you read here. And if you can’t follow that advice, be sure to only trade using a paper portfolio and not commit real monies. Conscientiously trading a paper portfolio for months before committing real money should be considered mandatory before trading based upon a model. Otherwise you are vulnerable to confirmation bias.

When I run across a model that piques my interest, I write code to implement the rules of the model (so far as I understand) and then attempt to quantify the performance of the model using data from sources like Yahoo Finance or the St. Louis Federal Reserve Bank FRED website.

My initial assumption is the model is wrong. My analysis attempts to quantify how and where the model is wrong.

In my experience, trading models usually fail in one of the following ways (additional bad outcomes omitted):

Cannot be verified (i.e., is no better than random)

Verifiable, but infeasible for real trading (i.e., trading costs and draw downs eventually absorb profits)

Fail in future (i.e., back testing bias / temporary correlations are in effect)

Other than long term dollar cost averaging in a 401(k)/IRA at a low cost mutual fund provider, such as Vanguard, I don’t know of a consistently profitable investment model.

I do believe there are profitable proprietary trading models. Hedge funds, high-frequency traders, and investment banks have advantages, such as market access and cost, which allow them to profitably trade using models.