390 entry models
7 years of research and development
up to 16 years of backtested data
1 platform that has them all.

Most systems hand you an alert and leave you to it. So you sit there all session with your finger over the trigger, taking the 3am entry model with the same conviction as the 10am one, and the market at 3am is not the market at 10am. It is thinner, it is slower, and the same setup that works at the open is noise overnight.

That is not a discipline problem. It is a missing piece of information, and the strategy already knows the answer.

A Legacy Window is a half-hour slot, New York time, where a strategy has actually earned its keep. Each is scored per weekday, on net R and on accuracy, and has to clear a floor before it counts as a window at all. Tuesday at 10:00 and Thursday at 10:00 are judged separately, because they are different days.

They are calculated from what has already happened, never from the session being traded. A window is a record, not a forecast.

It does not send you alerts. It does something better. An alert arrives whenever it arrives, which is usually while you are at work, asleep, or halfway through something else. Most people miss them even when they set them. So the alert becomes one more thing to feel guilty about.

A window is the opposite. It is known in advance. You can put it in your calendar on Sunday night, because the answer for Tuesday does not change on Tuesday.

Read the bottom two rows, because that is the whole argument. Same system, same sixteen years, same bracket. Traded indiscriminately it takes 14,377 trades to finish at -13.55R. Traded only inside its own windows it takes 569 and finishes at +78.57R.

The other 13,808 trades were not bad luck. They were work, commission and risk that paid nothing.

I've traded for seven years. During that time I went through dozens of mentorships. Some ideas helped. Most didn't hold up for me. I kept the useful parts, questioned the rest, and spent years trying to understand which market behaviours actually repeated.

I was also building things long before I was trading. I started teaching myself to program in sixth grade on a TI-83+, worked my way through a thick C++ book, and later studied video game design. When AI became publicly available, I saw a way to connect those two sides of my life.

The work started in a chat. I brought in the setups, rules and observations I had collected over years of trading. I challenged them, translated them into code, and tested them against historical market data. That grew into an AI-assisted research lab capable of generating strategy candidates, running backtests, comparing results, and showing me where an idea broke down.

In less than seven days, that lab explored roughly a hundred times as many candidate strategies as I had discovered on my own in seven years. It did work I could never have reviewed manually at that scale without fatigue, missed details, and a lot more time.

The surprising part was how much failed. We put idea after idea against the data. Some looked brilliant on a chart and fell apart when the rules were applied consistently. Others only held up in certain markets or at certain times. Out of that process, a small handful of core mechanics kept earning a closer look.

Those mechanics became the foundation for the models you see in the APES. We wrote the rules, coded the entries and exits, tested them against years of historical data, and put the results where you can inspect them, including the losses. The charting desk draws the entry models on the bars where they occurred. You can scroll back and examine what happened for yourself.

I was taught the same thing you were. Risk one to make three. A trade that risks more than it targets is a beginner's mistake. I believed it for years, and I passed on setups because the ratio didn't look the way I'd been told it should.

Then the lab let me test it instead of believe it. Thousands of strategies, the same rules applied every time, across years of data. The paradigm didn't hold. Reward to risk on its own tells you almost nothing. What matters is how often the target is reached before the stop, and those two numbers only mean something together.

That is why every strategy we test lands in one of two types.

Payoff strategies. Tight stop, far target. They lose often, and they depend on the occasional big winner to pay for the losing streaks in between. This is the one everybody is taught.

Accuracy strategies. The target sits closer than the stop. They are right most of the time, and each loss costs more than a single win pays. This is the one everybody is told to avoid.

Neither is better by default. Both can work and both can fail. Which one survives depends on the rules you are trading under, and that is the part nobody explains.

Think about what is actually being traded. A stock index is a basket of the largest companies in the country. Those companies grow their earnings, buy back their own shares, and are priced in a currency that loses value every single year. The index is built to drop the losers and keep the winners. Gold is priced in that same shrinking currency.

Every one of those forces pushes the same way, and they have pushed that way for decades. Shorting it means betting against all of them at once.

Try it yourself. Pull up a long-term chart of the Nasdaq, the S&P, the Dow, the Russell or gold. Zoom all the way out. Now tell me where a long-term sell would have worked.

You will find drops. Sharp ones. But they are one-offs, and every one of them was bought back and then some. A strategy that needs to catch the next one-off is a strategy that sits there losing while it waits.