Lesson Context Market Feed
SPY $776.34 -0.20%
QQQ $731.07 -0.14%
AAPL $305.93 +0.22%
MSFT $495.40 -0.30%
NVDA $225.16 -0.06%
XLK $190.01 -0.40%
Module 03 · Checklist

Questions that keep optimization honest

These filters help distinguish productively refining a strategy from merely rescuing it.

Metrics to watch weekly
  • Expectancy and trade quality trend.
  • Drawdown duration relative to plan.
  • Exposure concentration by symbol or strategy family.
Signs you should not optimize yet
  • You cannot explain what specifically changed in behavior.
  • The review is driven only by recent pain or recent euphoria.
  • The proposed adjustment rewrites the thesis instead of refining it.
Strong review outcomes
  • Leave the strategy unchanged and keep monitoring.
  • Tighten capital rules without altering the signal logic.
  • Retire the bot because the thesis no longer earns desk space.
In this lesson
  • Metrics to watch weekly
  • Signs you should not optimize yet
Try it — overfitting

Tune a strategy hard enough and it fits the past perfectly — then falls apart on new data. Push the tuning up and watch the out-of-sample curve diverge from the in-sample one.

In-sample
—
Out-of-sample
—
Overfit gap
—
Optimization warning

Do not optimize to erase normal drawdown.

← Previous lessonCore Concepts