Slice Crawl

Practice site for the puzzle: try your estimate() on as many random crawls as you like.

Goal: name the true best of the pizza shops in at least 95% of crawls, tasting as few slices (averaged over many crawls) as you can.

crawl.taste(i) buys a slice from shop i and returns its rating; estimate(crawl) returns the shop you believe is best. Everything runs here in your browser, and nothing leaves it unless you add a run to the leaderboard.

How your code is run

Type it here for one file of code, or Use my files to drop the .py / .ipynb files you work on elsewhere (or a .zip): the main file (the one defining estimate) can import the others. A notebook's code cells run top to bottom; lines starting with ! or % are skipped, and its own evaluate() calls run only 50 crawls, so loading stays quick.
Every function that takes a crawl shows up in the Estimator menu, so you can compare a few in one go. Name a run to tell versions apart: each row keeps the code it ran.

Anything callable works, even a method of an object:

class Racer:
    def __init__(self, z=2):    # runs once, when your code loads
        self.z = z

    def estimate(self, crawl):  # runs once per crawl
        ...

estimate = Racer(z=2.5).estimate

Your code

Starting Python...

Printed while loading

Run

Estimator

Run name Crawls Seed

Each crawl: 8 shops, bell-curve ratings. The same seed gives the same crawls, whatever the estimator.

Runs

Saved in this browser only. Click a row for its post-mortem and code, a name to rename it, a heading to sort. “Right” is ± 2 standard deviations; maybe means that range includes 95%, so run more crawls to tell. “Add baselines” runs the simplest estimators on the same crawls.

Post-mortem

Code for this run

Click a run above to go through its crawls one by one, with the hidden truth revealed.

Leaderboard

Every score here is our own re-run: 5000 fresh crawls of the official setup, on a seed nobody has seen. To add yours, use Add to leaderboard on a row of the Practice tab's runs.

#	Name	Method	Right	Avg slices	Median	90th pct	Posted