Canada stock prediction (TSX): our best market, at ~53% on ~2,000 verified calls

I run Trading Agent, and I publish every prediction we make — winners and losers alike — at /predictions. A while back I wrote an early version of this article warning you not to trust our Canada number, because at the time it was a flattering ~56% sitting on only about 36 verified calls — far too few to mean anything. I told you to watch that sample grow, and that the headline would probably move once it did. It did move. We now have around 1,986 verified predictions on the Canadian market (the TSX), and the directional accuracy has settled at about 53.3%. That's down from the lucky early read — exactly the kind of regression to reality I warned you to expect — but it has held comfortably above a coin flip across a sample that's now large enough to take seriously. As of today, Canada is our single best market.

The number, with the denominator attached

Here is the honest framing, with the part most tools hide bolted right onto it: ~53.3% directional accuracy across ~1,986 verified calls. Not 53% on 36 rows that could vanish on the next hundred. Nearly two thousand independent, time-stamped, publicly logged predictions. That denominator is what turns "promising line on a chart" into "a real, measurable edge."

For honest context, our blended accuracy across all 16 markets is about 49.7% — essentially a coin flip. That number matters because it's the baseline you should judge every individual market against. Against a ~49.7% house average, Canada at ~53.3% is genuinely above-average for us, and it's why it currently sits at the top of our table. But let me be precise about what that gap is worth: roughly three-and-a-half points of directional edge is useful — the kind of statistical tilt that can compound when paired with disciplined position sizing and strict risk limits — and it is emphatically not a money printer. A 53% hit rate still means you are wrong nearly every other call. Anyone who tells you a number like this prints money is selling something. It buys you a small, real tilt in your favour, nothing more.

The horizon detail is where it gets more interesting. Our 7-day forecasts on Canada run near 57% — meaningfully stronger than the blended-horizon average, and our best single horizon in any market. Shorter and longer windows pull the overall figure back toward 53%. So if there's a sweet spot in this data, it's the one-week directional read, and that's the part of the Canada record I'd weight most.

One more honesty note that the whole brand rests on: our backtests look better than this. They always do. Backtests are run with the benefit of hindsight, clean data, and no slippage, and they reliably overstate live performance. The number that counts is the one on the public scoreboard, generated forward in time and scored after the fact — and that live number is the ~53.3% above, not whatever a rosier backtest would claim. When the two disagree, believe the live record.

Why Canada does relatively well (a mechanism, now backed by the sample)

In the old article I floated a hypothesis for why Canada might be genuinely good rather than lucky. With ~1,986 calls behind us, I'm more comfortable putting weight on it — though I'll still call it a structural argument, not a law of nature.

The TSX is a deep, liquid market dominated by large-cap financials and resource names — the big banks, the energy and materials majors. Two structural features plausibly help a momentum-and-volatility model like ours read signal instead of noise:

• Liquidity and clean price discovery. A continuously, heavily traded large-cap tape is where technical features behave; thin, gappy markets are where they break down. The TSX's depth gives our features something coherent to work with.
• Momentum-friendly large caps with trending sectors. Resource and financial cycles — energy, materials, rate-sensitive banks — tend to produce the kind of persistent directional moves that momentum features are built to catch, rather than the random chop that whipsaws them. It's no coincidence our other liquid, large-cap markets are also among our stronger ones.

I'm naming sectors here — banks, energy, materials — purely as structural examples of why the market behaves the way it does. None of that is a pick, a tip, or a list of tickers to go buy. The structural point is the point: this is a market whose shape happens to suit our features, and the live record now supports that read rather than merely hoping for it.

How I'm treating Canada right now

Confident in the measurement, disciplined about the magnitude. The sample is finally large enough that I'll say plainly we have an edge in Canada — but it's a ~53% edge, and I'd ask you to size it like one. Our model emits Bullish, Neutral, and Bearish directional reads on Canadian large-cap names, and those are descriptions of a model's probability lean — never instructions to buy or sell, never a guarantee, and never anything close to the "90% accuracy" fantasy other tools advertise. The right way to use a 53% directional signal is as one input among several: something you weigh against your own research, your risk tolerance, and sane position sizing — not something you bet the account on because it topped our table this quarter.

You can see exactly how we generate and verify every call on our methodology page, and the live scoreboard — Canada included, full sample size and every miss — is always at /predictions. The 36 calls I once told you to distrust grew into nearly two thousand, and the edge survived the journey down to a believable number. That's the honest version of a good result: not a headline that overpromises, but a real, modest, measurable tilt that earned its place at the top of our table the slow way.

See the evidence for yourself — download the full resolved-prediction dataset, read the live public self-audit (hit-rate confidence intervals, live-vs-backfill split), inspect every model card, or run the research tools on your own data. No hype, just the receipts.

---

This article is educational content about machine learning and market structure. It is not financial advice, not a recommendation to buy or sell any Canada-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).