Your model doesn't watch the news

Remember this?

February 2020. Quant vol-selling strategies were still printing. VIX crept higher — to the algorithm, another "sell the spike" setup. Parameters met. The model sold.

The traders who survived that month weren't running better algos. They looked at infection rates in Wuhan and realized the regime had shifted beneath them.

This is what systematic trading gets wrong. Models assume the future rhymes with the past. 95% of the time, that's true. The machine harvests theta more consistently than you ever will.

But that other 5% is where careers end.

When Oil skew blows out because of a real military escalation, the model sees "elevated skew = sell premium." The human sees "this tail isn't kinetic — it's prohibitive. Stay out."

That distinction doesn't live in any backtest.

I spent 20 years on institutional desks. The most sophisticated risk decision I ever witnessed was a senior trader hitting pause. Not adjusting. Not hedging more. Just stepping aside.

No-trade is a decision too. Sometimes it's the best one you'll make all month.

The algo handles the known unknowns. For everything else — the stuff that hasn't happened yet — the only hedge is a human with the humility to say "I don't know where this goes."

That's not breaking the model. That's giving it the context it never had.