Trading systems built in the open and tested the hard way.

We call them the L systems. Each one is a book: a set of trading ideas, which we call legs, that run side by side on one account and share its money. A leg is a complete, self-contained rule: when to enter, how much to risk, and when to get out. Some legs hold a position for minutes, some for several days; one holds overnight, and one sells short on days that look weak. A book is more than the sum of its legs because the legs do not all earn at the same time.

Everything the systems see is built by us from raw one-second prices; we do not use a charting package. The books read many charts at once: candlestick charts at one, five, fifteen and thirty minutes, and brick charts at several sizes. A brick forms only when the market has moved a fixed amount, so a quiet hour may draw two bricks and a wild one forty; a brick chart has no clock. One family of bricks is adaptive: its size is reset each morning to the recent character of the market, so a calm week and a violent one draw a comparable number of bricks a day. Different legs read different charts, and some read several.

From those charts the software computes a large vocabulary of readings, and the logic of the books is built from more than a thousand distinct conditions drawn from that vocabulary. Which conditions, in which combinations, is the firm's property and is not published.

Behind the books sit three things that matter as much as the books themselves. A simulation engine that runs every idea through a realistic account, with limits on how much can be deployed at once, real trading costs, and a start from every year since 2020, so a result that only works from one lucky start date is seen for what it is. A research discipline: a test is written down before it is run, every result is judged against random look-alikes, and every campaign goes into a ledger, including the many that failed. And a live stack that runs the books on a practice account with the broker, decision by decision, so what the simulation says and what the machine does can be compared every morning.

We trade both, and we keep separate books for each, because the two instruments are good at different jobs. The choice follows the goal of the leg, not a preference.

An option turns a small move in the index into a large move in the option, and the most you can lose on a bought option is what you paid for it. When a leg has a strong directional entry and a short horizon, that shape is exactly what it wants. The price of an option also depends on things other than the index, above all how volatile the market expects to be, and an option loses value as time passes, so options suit legs that act decisively and get out.

A future moves point for point with the index, costs a fraction of a point to trade, and trades almost around the clock. When a leg needs to hold through the close and manage a position overnight, or to sell short, or to size up and down in small steps, a future is the cleaner tool. The Micro E-mini, one tenth the size of the E-mini, lets a book grow in small, exact increments.

There is good general information about both instruments in the public domain, and we will not repeat it here. What is ours is the decision of which book runs which instrument, and why.

A trading day is a pipeline. Prices arrive once a second. Our own software turns them into several sets of bricks and several intervals of candlesticks, all built by us, in parallel, from the same stream, and into the readings the rules use. Each leg decides for itself and asks the shared account for room. If there is room, an order goes to the broker; a stop goes with it. Everything the machine decides, and why, is written to a log. After the close that log is read beside what the simulation would have done on the same day, so that every difference is explained or flagged the same afternoon. Anything found wrong in the machinery is fixed and deployed before the next open; what a system trades is never changed overnight, only through a new version that earns its place.

A new book earns its place in stages. It first runs for days in a dry run: making every decision, placing nothing. Then it moves to a paper run on a practice account with the broker, at the sizes the simulation used, where real orders meet real prices and real fills. That is where the firm is today, and where it plans to stay for some time, confirming that every part behaves the way the simulation said it would before any real money is involved.

A trading record is easy to flatter. Pick the start date, count trades instead of days, quote the drawdown from the kindest year, and almost anything looks good. We hold our own systems to a stricter reading, and the fairest way to show you that reading before we show you our results is to apply it to something everyone can check: the S&P 500 fund itself, bought and held. Every number below is computed by our own code from our own stored daily prices, as of 2026-09-18, price only (dividends are not included).

The firm started with a limit every discretionary trader hits: a person can follow one or two ideas at a time, and only during the hours they are fresh. Years of rules, written down and tested by hand, could not be traded at once, consistently, by one pair of eyes. The only way to run them all, every day, the same way, is to give them to a machine.

Building that machine taught us something we did not expect. The hard part is not the code. It is the honesty: a simulation will tell you what you want to hear unless you build it to argue back. So the work became as much about measurement as about trading — real prices instead of modelled ones, random controls for every idea, a written record of every test, and a live practice account that reports what actually happened. We would rather know a system is modest than believe it is brilliant.

Here is the trap that shaped how we work. A backtest will readily surface trading logic that was profitable in the past. Most of it is not skill. Try enough rules on enough years and some will look wonderful by luck alone, and a luckily profitable rule looks exactly like a good one on paper. So we never judge a rule by its own profit. For every idea we build twenty random look-alikes: rules that trade just as often, on days chosen at random, with the same size and the same exits. An idea has to beat at least nineteen of the twenty, in every period we test, before it is even a candidate; and a candidate still has to be written up, run through the full account simulation from every start year, and then prove itself on paper. Most ideas do not get past the look-alikes. The ledger of the ones that did not is as valuable to us as the register of the ones that did.

The goal is a set of books with a verified track record: simulated, then on paper, then real, with the three compared in view of the people who follow this work. Everything else follows from that.

The members area holds what this site describes in outline: the register of systems and each system's page, with results labelled simulated, paper or real; the ledger of every research campaign, including the failures; and the documents prepared for the firm's own partners. Different guests see different rooms.

If you have been invited, use the link you were sent. If you have not, there is nothing to apply for.