With one week left, shift from learning new material toward consistency. Identify the two or three weaknesses that still affect timing or accuracy and target those directly.

Use a mix of focused drills and a small number of realistic full runs. Avoid doing so many simulations that you become fatigued or start practicing carelessly.

The final day should be light: review the process, technical setup, and common mistakes, then stop early enough to be fresh. If useful, this guide on McKinsey Solve guide gives more detail.

1 week out I'd rather fix 2 weak areas than do 20 random sims.

full run every couple days + targeted practice sounds right.

don't learn brand-new tricks two days before.