Running a markets scanner as a discipline

This is not financial advice. I am not telling you to trade, not handing you a strategy, not claiming Kairos, the scanner I run, makes money. No tickers, no returns, no do-this. The whole argument is about where a gut can learn and where it can't, and what you write down when it can't. If you came for a stock tip you came to the wrong essay, and I'm a little worried about you.

The case for trading on your gut is always the same case. A veteran develops a feel for the market. Twenty years watching the tape, and something in you learns to read it the way a doctor reads a chart across the room. It sounds right. It sounds like every other skill that gets better with time. It is the one place that sentence is close to false.

Here is what a feel actually needs to become real, and it isn't years. A gut only learns where two things hold. The world you're reading has to be regular enough to hold a stable pattern, and you have to have practised in it with feedback fast and clean enough that your brain can pin the correction to the thing that caused it. Kahneman and Klein, who spent years disagreeing about intuition and then agreed on this, called it the difference between a high-validity and a low-validity environment. A chess player has both. The stock market has neither, and Kahneman said so by name.

Take the three inputs a gut needs and watch the market withhold each one.

The pattern won't hold still. Whatever paid last cycle gets arbitraged away or flips. You are studying for a test whose answer key changes the week after you memorise it, so yesterday's lesson is not reliably today's rule, and the regularity a gut would train on keeps sliding out from under it.

The signal is buried in noise. Most of what a price does is nothing, just churn, and a run of wins can be pure variance. So the gut gets a gold star for a bad call and a slap for a good one, at random, often enough that the error signal it needs to learn from is drowned before it arrives.

And the feedback comes late and tangled. A discretionary call pays off weeks or months on, wrapped in the fifty other things that moved the market in the meantime, and by the time you learn how it went your memory has already quietly rewritten what you thought you predicted. The misses you'd learn the most from are the ones you no longer believe you made.

Put the three together and you get what Kahneman flagged as the low-validity case, where how sure you feel has almost nothing to do with how right you are. Experience in a room like that does not sharpen the gut. It laminates the original confidence under a stack of survivable stories, the wins were skill, the losses were the Fed, the tape, the guy who front-ran it, and not one of those stories ever puts the read at risk. Twenty years of that is one year of conviction, run twenty times, each pass sealing the last a little tighter. It feels exactly like learned intuition. It is confidence that was never once graded.

This is the same machine as the general case, pointed at the one environment built to defeat it. Elsewhere you can bolt a feedback loop onto a decision that doesn't come with one and drag it up toward the sky's honesty. Markets fight the loop the whole way. Which is the argument for building one anyway, deliberately, in writing, because the thing that feels most like expertise here is the thing you can trust the least.

A scanner run as a practice is not a smarter oracle. Kairos, the right moment as opposed to the clock, is a grand name for a spreadsheet that mostly tells me to do nothing. It doesn't know more than your gut. It substitutes, piece by piece, for the feedback the market refuses to hand over.

The environment won't hold a rule still, so you write the rule down first. One page, before any live call: the exact, checkable condition that counts as a hit, and what it does not count. If you can't write it before you see the chart, it isn't a rule. It's a mood with a ticker attached.

The grade arrives late and confounded, so you log the call before the outcome, with a threshold that can be proven wrong, and a date. Written after the fact it's a memoir. Logged before, it's the one artefact hindsight cannot revise, because it's already on the page in your own hand before you knew a thing.

The noise fools you trade by trade, so you review cold, on a calendar. Monthly, not after a win, because the urge to review only shows up after a win. Score every closed call against what you wrote, over the whole run, right, wrong, unresolved. The hit rate across many calls is the signal. The vivid last trade is noise, and it is loud.

And the whole place can turn without warning, so you size for survival, not for conviction. Position set by whether the practice outlives a run of being wrong, never by how sure the call feels, because "sure" here carries no information. Small enough that variance can't end you before the pattern turns, because in a world that keeps moving the first job is to still be here when it does.

That's the whole claim. The scanner is not a better prediction. It's the graded, pre-registered, survivable loop that a gut in this environment structurally cannot build for itself, written down on the outside where you can't spin it.

There's a real case on the other side, and skipping it would be cheating.

Great discretionary traders exist. Soros, Druckenmiller, Paul Tudor Jones, decades of returns no coin-flip explains. I'm not going to stand here and tell you their edge was luck, because I can't prove that and it isn't true. Two things, though. You are naming the handful who won a game that millions sat down to play, and in a low-validity field a few people win big on real edge plus variance and are impossible to pick out in advance from the confident many who quietly blew up. And look at what the winners actually preach. Stops. Position limits. Cut the loser, and cut it now. Druckenmiller and Jones sermonise about risk discipline precisely because raw feel is not to be trusted. They are the argument for the written loop, not against it. Their edge, wherever it comes from, does not ship to your account when you admire it.

Second counter, and it's the strongest. A rigid system encodes yesterday's structure and keeps trading it straight into a regime that broke it, while a human at least notices the world changed and steps back. True. Fully. A frozen system is exactly that blind, and regime change is where systems die. Which is what the cold monthly review is for, that's the appointment where you notice the regime turned and change or retire the rule, and what the survival sizing is for, that's what keeps a regime break from ending you before the review comes around. Systematic here does not mean never intervene. It means intervene on a schedule, off a written record, not off the adrenaline of a drawdown at 2 a.m. The failure mode the counter names is un-reviewed rigidity, which the discipline forbids on the first page. The undisciplined gut has the identical blindness and none of the record to catch it with.

Third, and this one's fair enough that the discipline is built to beat it. "Systematic" often just means a strategy tuned until it fit the past perfectly and predicts nothing, an over-fit curve that passes for rigour. Bailey, Borwein, López de Prado and Zhu wrote the formal version of this, backtest overfitting and what they bluntly called financial charlatanism, and it's real. So the loop is stricter than the backtest. Rules written before the data, not tuned to fit it. Tested out of sample. Graded live going forward, where the real test lives. Sized so an over-fit strategy bleeds slowly instead of detonating. Over-fitting is an argument for a harder discipline, not for trusting a gut that over-fits its own past in silence and keeps no log at all. The un-logged read is the worst over-fit there is. Fit to memory, and never once scored.

None of that hands you back your feel. The strongest counter narrows the claim and leaves it standing.

There's a base rate worth sitting with, and I'll use it for exactly what it's worth and no more. It measures professional fund managers, not you at your laptop, so treat it as weather, not proof. S&P keeps a scorecard, SPIVA, of active managers against the plain index they're paid to beat. Over fifteen years, not one of the twenty-two US stock-fund categories had most of its managers come out ahead. Close to nine in ten trailed. These are people with terminals, teams, and clean data, and the market still mostly wins. That doesn't prove the calibration gap. It just tells you that the room where confidence is cheapest is the room where being resourced and being right come apart hardest.

The practice, then. Not a stock tip. Five moves that make a scanner a discipline instead of a hunch with extra steps.

Write the rule first. Before any live call, the exact condition that triggers a hit, on one page, and what it excludes. Can't write it before the chart? It's a mood.

Log the call before the outcome. Timestamped, with a falsifiable threshold, what would prove this wrong, by when. This is the only artefact hindsight cannot rewrite.

Review cold, on a calendar. Monthly, not after a win. Score the whole sample against what you wrote. The hit rate is the signal; the last vivid trade is not. This is also where you catch a regime that turned.

Size for survival, not conviction. Set by whether the practice outlives a bad run, never by how sure you feel, because sure is uninformative here. Small enough that variance can't end you.

Grade the process, not the month. Judge whether you ran the loop honestly, not this month's P&L, because noise means good process loses and bad process wins over any short stretch. And the one question that closes it: of your last twenty calls, how many did you write down before the outcome, with a threshold, and then actually go back and score? If the answer is near zero, you don't have a system. You have twenty stories, and you're the one they're for.