Insights
Methodology deep-dives, market commentary, and founder notes. Educational content — not financial advice.
NZX AI forecast: can machine learning predict New Zealand stocks?
Every global AI stock tool ignores the NZX — too small, too illiquid, not worth the engineering. As a New Zealand company, we don't. Here's an honest look at whether machine learning can forecast NZX-listed stocks — including the uncomfortable fact that, on roughly 1,823 verified predictions, our home market currently runs below a coin flip.
Australia stock prediction (ASX): the honest numbers
Our ASX model sits at ~51.5% directional accuracy on ~1,991 verified calls — a modest edge over a coin flip, not a money printer. Here's the resource-and-financials reason it's only modest, and why the live number is the one that counts.
India stock prediction (NSE): ~52% on ~1,964 verified calls — a real but modest edge, honestly framed
Our model hits ~52.0% directional accuracy on India (NSE) across ~1,964 verified predictions — above a coin flip, one of our better markets, with the 7-day horizon near 53%. An important compliance note: we are not SEBI-registered, so this is not directed at Indian retail investors.
Vietnam stock prediction (HOSE): a frontier market and an honest, modest edge
My model is ~51.5% on Vietnam (HOSE) across ~1,976 verified predictions — a modest edge, not a money printer, with real frontier-market and data-coverage caveats I own.
Indonesia stock prediction (IDX): the honest numbers and why my model lands below even
My model scores ~46.3% directional accuracy on Indonesia (IDX / .JK) across ~1,950 verified predictions — below a coin flip, and one of my weaker markets. Here's why thin liquidity, the rupiah, and foreign flows feed a momentum model noise it misreads.
Thailand stock prediction (SET): my single worst market, and why I'm telling you
Thailand (SET) is now my single worst market: about 43.8% directional accuracy on roughly 1,889 verified predictions — clearly below a coin flip. Here's the market-structure reason why, and why a number this low doesn't mean you can just bet the other way.
Malaysia stock prediction (Bursa): the honest numbers and why it's hard
Malaysia (Bursa) started as my worst, thinnest-data market — but with ~2,133 verified predictions it converged to ~49% directional accuracy, near coin-flip and mid-pack. Still sub-50, and here's the market-structure reason why.
Singapore stock prediction (SGX): the honest numbers and why it lands near a coin flip
My model scores ~49.1% directional accuracy on Singapore (SGX) across ~2,139 verified predictions — basically a coin flip, just under 50. Here's why REITs, defensive banks, and low momentum turnover keep it near even.
Hong Kong stock prediction (HKEX): the honest numbers and why it lands on a coin flip
My model scores ~49.8% directional accuracy on Hong Kong (.HK / HKEX) across ~1,950 verified predictions — almost exactly a coin flip, right at even. Here's why China-linked names, policy shocks, and Southbound flows keep it from finding an edge.
China stock prediction: the honest numbers and why it's a harder market
On our full public log, our model scores about 47.6% directional accuracy on China (Shanghai) across roughly 1,330 verified predictions — just below even. Here's the honest reason: it's a retail-driven, policy-sensitive market that's hard for a momentum model.
TW stock prediction with machine learning: the honest numbers and why it's hard
Most global AI stock tools ignore Taiwan entirely. We don't — and I'll be honest about the numbers: across roughly 1,560 verified predictions, our directional accuracy on TW stocks sits around 46.9%, below even and one of our weaker markets. Here's the real data, why Taiwan is genuinely harder to predict than markets like the US and Canada, and what that teaches you about market efficiency.
Korea stock prediction (KOSPI): our honest numbers on a retail-driven market
Our model lands around 49.7% directional accuracy on KOSPI — basically a coin flip, blended — on a now-large verified sample. The one relative bright spot: the 3-day horizon at about 54%. Here's why retail 'ant' flow and chaebol concentration wash out the technical signal.
Japan stock prediction (Nikkei / TSE): the honest numbers and why it's a coin flip
On a full public log of about 2,405 verified predictions, our model scores roughly 49.9% directional accuracy on Japan — almost exactly a coin flip. Here's the honest reason the Nikkei and TSE land right at even for our AI model.
UK stock prediction (FTSE): the 72% fluke collapsed to 51% — exactly as I warned
Our UK model once showed ~72% directional accuracy on a tiny sample. I told you not to trust it. Now, on ~1,976 verified calls, it has settled at ~51.2% — a modest edge, not a money printer. Here's the lesson, playing out in the open.
Canada stock prediction (TSX): our best market, at ~53% on ~2,000 verified calls
Our model now hits ~53.3% directional accuracy on the TSX across ~1,986 verified calls — currently our single best market, with the 7-day horizon near 57%. Above-average for us, but a disciplined edge, not a money printer.
US stock prediction with machine learning: where our model actually works
Our model hits ~52.6% directional accuracy on US large-caps across ~3,076 verified calls — a real, sample-backed edge, but a modest one, not a money printer.
Walk-forward validation — why most stock-prediction backtests lie
The 'tested on 5 years of data, returned 70% a year' pitch you see in finance Twitter ads is almost always smoke. Here's what's actually happening, why even a rigorous walk-forward backtest still flatters itself, and why we publish our live forward-tested record instead — gap and all.
Why most AI stock-picking tools are lying — and the one question that exposes them
Almost every 'AI stock predictor' advertises an accuracy number you can't verify. There's a single question that separates the honest tools from the marketing fronts — and most of the market fails it. Here's the question, why it works, and our own real numbers (including the markets where our model is below a coin flip).
Why I built Trading Agent — and why one person, not a team
There are dozens of stock-prediction tools out there. Most are either gambler bait or expensive enough that nobody you know can afford them. Trading Agent is my attempt at something in the middle — built by one person, in New Zealand, on a laptop.