Model performance & accuracy
Our forecast engine is a logistic-regression model trained on all-time PSX history and retrained every day on the previous day's realized outcomes. Here is how it actually performs — measured, not marketed.
How to read these numbers
The accuracy figure below is directional hit rate: how often the signal matched the direction the stock subsequently moved. It is measured out-of-sample — on data the model had never seen when it made the call — not on the history it was trained on. That distinction is the whole point. Any strategy can be tuned to look brilliant on the past, because the rules were chosen knowing what happened; that is overfitting, and it is the main reason promising models disappoint in live use.
Set your expectations accordingly. On a problem this hard, an honest directional hit rate lands modestly above a coin flip. It is not going to look like 90%, and if you see a site claiming that, they are either measuring on training data or not telling you the truth. A model that is right somewhat more often than chance is genuinely useful as one input among several — it is not a machine that tells you what to buy.
A hit rate is also not a return. Being directionally right slightly more than half the time says nothing about the size of the moves you caught versus the ones you missed, and it excludes brokerage, taxes and slippage — which, as our getting-started guide explains, are what quietly separate a good-looking strategy from a profitable one.
Per-stock records below are shown only where there are at least 20 resolved predictions. Below that, a high percentage is noise: a stock the model called correctly four times out of five has told you almost nothing about its fifth call.
Held-out hit-rate
59%
vs 58.57% for always guessing the commoner direction (+0.43 pp)
Training samples
276,776
all-time price history
Features
17
technical + relative-strength
Last retrained
6 days
ago
Live forecast record
We have started recording every forecast as it is issued, and grading it against the price that actually printed once its ten-day window closes. Nothing is shown here until those first calls mature. Until then the figures above are the honest ones we have: accuracy measured on history deliberately held back from training.
Daily learning curve
Backtested hit-rate after each training run — the model keeps learning from new outcomes.
53 training runs · baseline for a coin-flip is 50%.
Current signal mix
Strongest per-stock track records
Directional accuracy on each stock's own backtested forecasts (≥20 resolved).
How to read these numbers
Forecasting individual stocks is hard. A directional hit-rate a few points above 50% is a real, repeatable edge on live market data — we deliberately do not inflate it. The headline figure is measured across thousands of backtested signals; the per-stock track records show where the model reads price action best. This is an analytical tool, not investment advice — see our methodology and disclaimer.