How It Works · #3 of 4
Not All Stocks Are the Same — So We Don’t Treat Them That Way
A small-cap stock trading 50,000 shares a day behaves very differently from a blue-chip moving millions. gAInXalpha groups stocks by how actively they trade — and gives each group its own dedicated AI model.
The problem with one-size-fits-all forecasting
Most forecasting tools apply the same model to every stock in a universe. On the surface, that sounds efficient. In practice, it’s a significant compromise.
A heavily traded, large-cap stock moves in response to news, institutional flows, and broad market sentiment. A thinly traded small-cap reacts to very different forces — a single large order can move its price, analyst coverage is sparse, and price swings can be sharp and sudden. Teaching one AI to handle both is like asking a single doctor to specialize in both neurosurgery and pediatrics. Possible, perhaps — but not ideal for either patient.
Think of liquidity like traffic on a road. A major motorway (high-liquidity stock) has thousands of cars flowing smoothly — it’s predictable and well-studied. A country lane (low-liquidity stock) sees just a few vehicles a day — one tractor can change everything. You wouldn’t use the same traffic model for both. Neither do we.
What liquidity actually means
Liquidity is simply a measure of how easily a stock can be bought or sold. We use trading volume — the number of shares changing hands each day — as our main proxy. High volume means a stock is actively traded and easy to move in and out of. Low volume means it’s thinly traded, and prices can be more sensitive to individual transactions.
How we group stocks
Within each asset class — say, European equities — the gAInXalpha engine splits all covered securities into ten groups based on their liquidity. Each group gets its own AI model, trained specifically on securities with similar trading characteristics.
Why this is better for you
Academic research has shown that AI-based return signals are often strongest in less actively traded, harder-to-arbitrage stocks. A pooled model — one trained on all stocks together — tends to dilute that signal. By isolating each liquidity group, we let the model focus on what’s actually predictive within that specific trading environment.
The result: forecasts that are genuinely calibrated to the type of stock you’re looking at, not a generic average that satisfies no one.
The bottom line: gAInXalpha doesn’t apply the same formula to a FTSE 100 giant and a small AIM-listed company. By grouping stocks by how actively they trade and giving each group its own dedicated AI, we make sure every forecast is built on a model that actually understands the stock’s trading world.