The Plan

• We can let each scan run for a given number of iterations, and then switch to a different scan.
• If we return to a scan that ran beforehand, it will continue to run from where it stopped.
• We can time the scans to see when they solve each board in a random sample of the boards.
• We can use these statistics to construct a suitable meta-scan.
• Can you think of a good algorithm for doing that?