"How do you pick the weekly portfolio?" My answer was two words.
Every two weeks I sit down live with paid members and answer whatever they bring. Here is what came up this week, including the question that has the most surprising answer.
Every two weeks I run a live Q&A for Stats Edge Pro members. Open mic, no script, whatever they want to ask. This week’s covered a lot, and some of it is worth sharing with everyone, not just the people in the room.
Here are the highlights.
“How do you pick the stocks on the weekly list?”
My answer was two words: I don’t.
This is the part that surprises people most. There is no morning where I sit with a coffee and decide which names look good. The whole job is in building and testing the system. Once that is done, RealTest, the backtesting software, simulates the portfolio from the year 2000 to now and tells me exactly which orders to send. I translate that output into something readable and send it to members. I do not pick, I do not predict, I do not have a hunch. If I did, it would not be systematic trading anymore. The work is in the system. The list is just the system talking.
A system I spent about 100 hours on just failed.
I spent the better part of two months testing whether moving the swing systems to a daily cadence instead of weekly would improve them. More frequent alerts, more action. I tested it for weeks, hours a night.
It was no better. Within about one percent, before even counting the extra costs and slippage. So I am not doing it. That is not a fun thing to report, but it is the actual job. Being a systematic trader means you build things, you test them honestly, and a lot of them fail. The discipline is in being honest with the data instead of talking yourself into the version you wanted.
On using AI to build a trading system.
A member asked how to use AI to build a daily system. My honest advice: do not have AI do the whole thing. You will not know where the data came from or whether to trust it. Instead, buy a real backtester. RealTest costs around $300. Pay for real market data. A proper backtester handles the genuinely hard parts: tracking buying power in real time, dealing with delisted securities, all the computational machinery that has to be right. Then use AI to interface with it and help you code. I have AI work overnight on strategies for me, and I trust the output specifically because a real backtester sits underneath it. Build versus buy: buy the backtester, buy the data, and let AI be the assistant, not the engine.
Why the setups always look ugly at the time.
Pullback and mean reversion plays look uncomfortable when they trigger, because we are all taught to buy things that are going up. That instinct is also why a lot of discretionary traders struggle. It turns out that, especially in stocks and especially over shorter time frames, buying into weakness tends to work better than chasing strength. The system does not care that the chart looks ugly. That is the point of running a system.
The thing I am still working to fix.
The most common piece of member feedback right now is that running the systems takes more time than people expected. I hear it, and it is on me to solve, not on members to push through. I have been rebuilding the onboarding around exactly that problem, and there is more coming. If you have ever felt that way, you are not alone, and you are not doing it wrong.
That is a slice of what a member Q&A looks like. The next one is in two weeks. Members can ask anything, and I answer it on the spot.
If you want to be in that room, sign up! If you are not ready, that is fine. These highlights will keep coming, and the free letters stay free.
Michael
Michael Nauss, CMT, CAIA, CDMS. Founder, Stats Edge Trading.