Learn / Market intuition
Answer page / market intuition
Topic cluster / Signal and portfolio designBecause probability is not expected value. A 60 percent signal can still be bad if the payoff is poor, the model is miscalibrated, the costs are too high, or the trigger only appears in the noisiest parts of the market.
What to remember
People hear '60 percent probability' and think 'good trade.' But a signal is only useful if the payoff, costs, and calibration make the trade worthwhile. Directional correctness by itself is not enough.
A strategy can be more likely right than wrong and still lose money if it pays too much when it is wrong, trades too often, or acts on overstated confidence.
Probability outputs often feel precise, which makes them dangerous. The number on the screen may come from a model that is poorly calibrated, trained on the wrong base rate, or mapped to trades with weak payoff asymmetry.
Instead of asking 'is the signal above 60 percent?' ask 'does the signal in this range produce enough net edge after costs and drawdown to justify trading it?' That is the question the backtest and calibration work should answer.
This is the same discipline Alphora pushes everywhere else. A pretty score is not enough. You need real net edge, honest execution assumptions, and a decision rule that improves the strategy instead of just sounding confident.