The Flash Crash made one thing clear: you cannot compete on speed.

The speed problem is solved (against you) #

If you read the Flash Crash piece, you already know the punchline. The firms that made money on May 6, 2010 weren't sitting at home refreshing a screen. They had co-located servers and direct market access. They saw the order book collapsing microseconds before anyone else and got out.

That advantage hasn't gone away. It's gotten wider.

Algorithmic trading broadly accounts for somewhere between 60 and 75 percent of U.S. equity volume, depending on who's counting and what they include. High-frequency trading is a subset of that, but even the subset operates at speeds and budgets that retail can't touch. You are not going to out-speed them from your home office.

So if you can't compete on speed, what can you compete on?

One category of data that doesn't care about speed #

When a CEO buys shares of their own company with their own money, they have to tell the SEC. It's called a Form 4 filing.

Corporate insiders (officers, directors, anyone who owns more than 10% of a class of a company's equity securities) are required to report their transactions to the SEC within two business days. These filings are public and free. They're posted to EDGAR, the SEC's electronic filing system, where anyone can read them.

No license required. No data vendor subscription.

A CEO buying $2 million of their own stock on the open market is a different kind of signal than an algorithm executing a statistical arbitrage strategy in 40 microseconds. One tells you something about conviction. The other tells you something about speed.

You can't compete on speed. Conviction doesn't have a latency requirement.

Why this data matters #

When I started looking into this, I expected the academic research to be thin. It isn't. A Wharton study covering 22 years of data found that insider purchase portfolios outperformed the market by about 6% annually. Other studies put the number higher. The finding keeps showing up. Not always. Not reliably enough to be a trading system by itself. But consistently enough that I couldn't ignore it.

The reason isn't complicated. When someone who sees the internal numbers, the pipeline, and the board presentations decides to write a personal check for company stock, they know something the market doesn't. Sometimes they're wrong. But they're wrong less often than random.

The catch is that most insider transactions are noise. Stock grants, option exercises, automatic sales under 10b5-1 plans — those happen on schedule regardless of outlook. I had to learn to filter all of that out.

Open-market purchases are the signal. Nobody is forced to buy. When a CFO puts $500,000 of their own money into shares, that's a voluntary bet.

Why retail traders ignore it #

Form 4 data is boring.

It's not real-time. It doesn't fit into the dopamine loop of scalping or the narrative loop of technical analysis. Nobody is selling a course on "How to Read SEC Filings."

The filings themselves are XML documents. Structured, but not pretty. Parsing them requires knowing what fields matter, filtering out the noise transactions, and understanding context. Is this a cluster buy? Is the insider buying for the first time? How does the purchase size compare to their existing holdings?

Most retail traders don't even know EDGAR exists. The ones who do usually don't have the patience or tooling to monitor it systematically.

How I ended up here #

I spent years trying to build trading tools. Backtesting engines, signal generators, portfolio optimizers. Every approach hit the same wall: data costs.

Licensed market data kills your economics before you write a line of code. A basic historical equities dataset starts at hundreds per month. Real-time feeds are thousands. And the kind of data you'd need to build and distribute a useful product requires licensing agreements that a solo developer can't afford.

I kept circling back to the same question: what data is free, public, and actually worth something?

Form 4 filings kept being the answer. Public, free, filed within two business days, structured XML with consistent fields. No redistribution restrictions. No licensing problems.

The economics work because the SEC already made the data public. I'm not reselling someone else's feed. I'm reading what's already there.

What I built (and what I didn't) #

I built a pipeline that watches for Form 4 filings and surfaces the ones that look interesting. Open-market purchases by officers and directors, filtered from the noise of grants, exercises, and scheduled sales.

I should be transparent about what this is and isn't.

I don't trade. I don't have the capital to survive drawdowns, and I know enough about myself to know I'd be bad at it. I don't sell signals. I don't give investment advice. I don't look at the insider data and make personal trading decisions.

What I do is build plumbing. The Flash Crash taught me that market structure is plumbing, and plumbing is what I understand. So I built a system that monitors a specific, public, under-watched data source and makes it easier to pay attention to.

I built a tool that does this. It's at matchstick.trading if you're curious.

Open questions #

• Have you ever used insider filing data in your own analysis? What did you find?
• What public data sources have you found that are genuinely useful and free?
• If you've tried to build trading tools, where did you hit the data cost wall?