Theory of Edge was not born from a business plan. It was born from two very different careers that ended at the same conclusion: an edge you cannot explain, measure and survive is not an edge.
Quoting both sides, managing inventory, living on edges measured in ticks, and on the days when they vanish.
Building search engines and models, and learning exactly where they lie.
Market making looks like the safest job in trading. You quote a bid and an offer, you earn the spread, you do it thousands of times, and most days are quietly green.
The edges are thin, a fraction of a tick per trade, and they never stay put. A model that worked last month has to be recalibrated this month, because the market, the competition and the flow have moved.
The real risk hides in the shape of the returns. Hundreds of small wins, then one fast market: a surprise headline, a gap, liquidity that disappears in a second. The inventory you were carrying moves against you faster than you can unload it, and a single session takes back what months had built. Every market maker knows this profile. The industry has a name for it: picking up pennies in front of a steamroller.
What he wanted instead was simple to say and hard to build: strategies whose risk is defined before the trade is taken, a stop on every position, ideas that rest on behaviour documented for years rather than on a statistical accident, and enough of them, uncorrelated, that no single day can decide the year.
Give a search algorithm enough freedom and it will always find a curve that goes up. Not sometimes. Always.
The other half of the team spent years building genetic strategy engines and machine-learning models. Financial data is mostly noise, and a flexible enough model will memorise the noise and call it a pattern. Every additional parameter is another way to fit the past; every additional backtest raises the odds that the best one is lucky. The models that look the most impressive on history are very often the first to fail on a live account.
Before running it: fewer free settings, fewer ways to fool yourself.
And never let that data choose anything during the search.
A strategy that only works at one exact setting is a coincidence. Keep the ones whose neighbours work too.
The more strategies you test, the higher the bar a result has to clear.
Put those two experiences side by side and the design of Theory of Edge follows almost by itself.
From the market maker: defined exits, and the conviction that a portfolio of modest, independent edges beats one brilliant strategy. From the engineer: a deliberately narrow search, documented strategies instead of invented ones, and a validation process designed to reject.
The engine never invents a strategy. You pick one of 17 behaviours documented in the literature, and it tunes a handful of settings on your own market: when to trade, how to enter, where the stop and the target sit, and at most a few filters. What survives is checked on data it never saw, stress-tested by Monte Carlo, and only then combined into portfolios. The whole process is on the Method page.
We built this tool for our own trading and we still use it on our own accounts. We do not sell signals and we do not manage money: you get the engine and the numbers, and the decisions stay yours.