UK stock prediction (FTSE): why I don't trust our own 72% number yet

I run Trading Agent, and I publish every prediction my model makes — winners and losers — at /predictions. Most founders writing about their own product would lead with their best-looking number and let you assume it means something. I'm about to show you our best-looking number and then spend the whole article telling you not to believe it. That's not false modesty. It's the single most honest thing I can say about how this model works, and the UK is the perfect place to say it.

The number that looks amazing — and almost certainly isn't

On UK-listed equities (the FTSE, traded on the London Stock Exchange), my model's verified directional accuracy is about 72%, across roughly 36 verified predictions.

Seventy-two percent. If I put that on a billboard, I'd probably sell a lot of subscriptions. A coin flip is 50%, so 72% sounds like a genuine, sizeable edge — the kind of number that makes you think "this thing actually predicts the market."

It almost certainly doesn't. Not at 36 data points. Let me explain why I refuse to trust my own headline figure, because this is the most important thing in the article.

Thirty-six rows proves essentially nothing

Here's the uncomfortable maths. With only ~36 verified calls, the random swing around the true accuracy is enormous. To make the intuition concrete: imagine a model with no skill whatsoever — a literal coin flip, 50% true accuracy. Run 36 flips and it is entirely ordinary for that coin to land 22, 24, even 26 times right purely by chance. Twenty-six out of thirty-six is 72%. A skill-less coin can produce my exact headline number on a sample this small without breaking a sweat.

So when I see 72% on 36 rows, the responsible interpretation is not "my UK model is brilliant." It's "I cannot yet distinguish this from luck." The confidence interval around 72% on 36 samples is so wide it comfortably includes outcomes from "genuinely good" all the way down to "no better than a coin." I'd bet that as the sample grows into the hundreds, that 72% drifts down — possibly hard. It would not surprise me at all to watch it collapse toward 45–50% once we have a few hundred verified UK calls. That's not pessimism; that's what regression to the mean looks like.

Contrast it with the one market I do trust. Our US figure is about 54% across roughly 830 verified predictions. That's a far less exciting number — barely above a coin flip — but 830 samples is deep enough that the 54% is very unlikely to be a fluke. A small edge measured over a large sample is worth infinitely more than a large edge measured over a tiny one. Our blended accuracy across all 13 markets is around 46% on ~2,248 verified predictions, which tells you the honest baseline. The UK's 72% is sitting above every other market we cover — and that, paradoxically, is exactly why I distrust it. Outliers on small samples are usually noise wearing a nice suit.

This is the whole reason I built the brand around radical honesty. The industry standard is to find your flukiest, best-looking market and frame it as proof of skill. I wrote about that pattern in why most AI stock-picking tools are lying. Here I have a 72% in hand — the exact kind of number a less honest operator would weaponise — and I'm telling you to ignore it. If you only remember one sentence: do not trust our UK number yet.

The market structure (a real footnote, not the headline)

There are genuine reasons the FTSE behaves differently from other indices, and they're worth understanding — but they're secondary to the sample-size point, so treat them as context, not as a defence of the 72%.

The FTSE 100 is unusually value- and dividend-heavy, and it is one of the most globally exposed major indices in the world. Most of its large constituents — the oil majors, the miners, the global banks, the consumer-goods giants — earn the majority of their revenue abroad, not in Britain. The practical consequence is that the index often trades on the pound and on global macro as much as on the UK economy itself. A weaker pound can lift the index even on a gloomy domestic day, because overseas earnings translate back into more sterling. There's also a lingering Brexit-era re-rating overhang that has kept UK valuations structurally cheaper than US peers for years.

All of that makes the UK a genuinely different animal for a momentum-and-pattern model like mine. But notice what it does not do: it does not rescue 36 data points. Even if every structural story above were perfectly true, I'd still need a few hundred verified calls before I could claim the model has a real edge here. Interesting structure plus a tiny sample still equals "unproven."

What I actually do with a 72% I don't trust

I label UK signals Bullish, Neutral, or Bearish — never Buy or Sell — and when I mention an index-level name as an example, I'm using it to illustrate structure (globally-exposed, sterling-sensitive large-caps), not pointing at anything to trade. On the UK specifically, I treat the headline accuracy as provisional and weight it accordingly. A number I can't separate from luck is not a number I'd build a decision around, and I'd urge you not to either.

So the honest UK story is almost the inverse of a sales pitch: I have a great-looking figure, and my professional judgement is to distrust it until the sample catches up. If you want to see exactly how these calls are generated, scored, and verified, it's all on the methodology page, and the live, unedited record — UK included — is always at /predictions. Watch the UK number over the next few hundred predictions. If it holds, I'll be thrilled and I'll say so. If it regresses, you'll have seen it happen in the open. Either way, you'll have watched real data settle the question instead of trusting a billboard.

---

This article is educational content about machine learning and market structure. It is not financial advice, not a recommendation to buy or sell any UK-listed or other security, and not directed at any individual's circumstances. Trading Agent is a quantitative research tool operated by WU Capital Limited (New Zealand).