Machine Learning Signal (Assetick Model)
A model trained on 1.5 million DSE candles that estimates whether a share will beat the typical DSE stock.
Last updated:
Definition
The Assetick Model is a machine-learning classifier trained on more than 1.5 million historical DSE daily candles. For each share it estimates the probability that over the next five trading days it will outperform, roughly match, or underperform the median DSE stock.
How it is calculated
The target is market-relative: a share's 5-day return minus the median DSE return, with ±2% bands. Accuracy is judged by 'edge' — accuracy minus the share of the most common outcome — so the model is always compared with a simple baseline, and live predictions are recorded and checked.
How to read it
A probability is not a promise. The model's edge over the baseline is modest, which is realistic for a stock market. Treat it as one input alongside the chart, fundamentals and news.
For DSE investors
Predicting 'beats the market' rather than 'goes up' was the key fix that made the model useful on the DSE, where most shares move together with the market.
Frequently asked questions
How accurate is the Assetick Model?+
Its live edge over the baseline is tracked continuously and compared with its training holdout. Expect a modest, real edge — not certainty.
How often is the model retrained?+
Not daily. It is retrained only after enough live trading days show whether a new model is genuinely better.