Update: The Rules Keep Growing, the Portfolio Doesn’t (Yet)
It’s been about a week and a half since the last update, which is either not long enough to expect much or exactly long enough to notice that not much happened. Both are true. Here’s where things actually stand.
Quick correction to the premise first, since the title’s already carrying it: this isn’t actually a $100 experiment anymore. I added a second $100 on July 14, less than a week into the whole thing. Honestly, not for a great reason. I added it somewhat impulsively, half expecting that by this point the agent would be running through candidates fast enough that $100 would already feel like a cap on the story rather than the story itself. That has not happened. As you’ll see below, the actual pace has been the opposite of that. So the account is funded like I was gearing up for a much higher-volume version of this than the one I’m actually getting, and I think that gap is worth naming rather than quietly updating the numbers and moving on.
What’s changed since last time
One new trade. On July 13, the agent opened a second position in UUUU (Energy Fuels) on a two-insider cluster, CEO and Chairman both buying on the open market within a day of each other. Twenty dollars in, same as GEHC before it.
Since then: a run of sessions where nothing got bought, and that’s mostly the point of this update.
- KMX has now been blocked three sessions in a row. It’s a genuinely strong-looking candidate, a five-insider cluster including the CEO, but the rule requires each insider’s Form 4 verified directly against SEC filings, not against a news aggregator’s summary of them. Two of the four are confirmed straight off SEC.gov as of this week. Two aren’t, despite real effort to track them down. So it stays blocked. That’s not a bug. That’s the rule doing the thing I asked it to do.
- A congressional cluster in AT&T looked clean right up until it didn’t. Bipartisan, no red flags, and then it turned out the company reports earnings two days out, which is inside the blackout window the rules set specifically so we’re not opening a position right before a number that could move the stock either way. Caught it, skipped it.
- A congressional cluster showed up in SpaceX. Two members bought in mid-June, right after the IPO, one of the largest in market history. The cluster technically clears the rule as written. I didn’t take it. SpaceX has been trading for about five weeks. There’s no real history to evaluate a company like that against, and “a rule technically permits this” isn’t the same as “this is what the rule was built for.” Flagged it instead of forcing a decision either way.
Portfolio value is $197.15 against $200 contributed. Down about half a percent net, against a shadow QQQ benchmark that’s basically flat over the same stretch. Not a story either way yet. Just a number.
The part where I’m not sure how I feel about this
I knew going in this wasn’t going to make me rich. Twenty dollar positions were never going to. But somewhere in the last week and a half I noticed I was spending more time editing rules than making decisions, and that started to bother me in a way I didn’t expect.
I said it plainly in a session this week: I’m not sure this is going anywhere. Not the money specifically, and not the trade count specifically, and not the rulebook specifically. All of it, together, at once. Two trades and a growing SOP in twelve days is a real ratio, and I don’t think I’d decided yet whether that ratio means “building something durable” or “organizing around the thing instead of doing it.”
I don’t have a tidy answer to that yet. I think that’s honest to report rather than resolve into something cleaner than it is.
Cutting Signal C’s FOMC leg
This is the concrete thing that came out of sitting with that discomfort instead of ignoring it.
Signal C was built to catch cases where a prediction market’s implied odds on a macro print, CPI, payrolls, a Fed decision, disagree meaningfully with an independent read on where things actually stand. CPI has one: the Cleveland Fed publishes a running inflation nowcast, so there’s a real number to check the market’s pricing against. Payrolls has one too, off the jobless claims trend. FOMC never did. There was never a defined anchor for what a rate decision “should” be priced at, independent of the prediction market’s own price. That gap sat quietly in the rules for six weeks because nothing forced the question.
This week something did. There was a real spread between two prediction markets on the July FOMC meeting, one showing meaningfully higher odds of a hike than the other, driven by some oil and geopolitical noise. First real live test of that part of the signal. And I had no honest way to say whether it was a real divergence worth trading or just noise, because there was nothing solid to measure it against.
I could have made up a number. Picked some percentage-point gap and called it the threshold. I didn’t, for the same reason I didn’t invent a materiality threshold for CPI when I first built this thing: a number pulled from nowhere doesn’t stop being arbitrary just because it’s sitting inside a rulebook. It just gets to look official while it’s wrong.
So FOMC is out. CPI and payrolls stay. If FOMC comes back, it comes back because there’s a real anchor for it, not because I got impatient watching Signal C sit at zero trades since it was added.
Next steps
Being straight about this part too: I don’t fully know yet, and I’d rather say that than manufacture a clean plan.
The real options on the table:
- Hold the SOP still for a while. No new rules unless something actually breaks, and just let the existing ones run for a month, so the project isn’t also a rules-writing project.
- Simplify on purpose. Fewer signals, a lower verification bar, more trades and more noise, in exchange for actually seeing the strategy play out instead of watching it get stuck at the gate.
- Let it be the slow thing it’s turned out to be, and write that as the real story instead of waiting for a dramatic one that might not come.
- Decide this isn’t the project I want to keep running, which is also a legitimate answer twelve days in.
No decision yet. Whichever way this goes, it’ll be in the next update, made in real time rather than smoothed out after the fact.
Leave a Reply