~?~

~?~

~?~

~?~

January 21, 2012 Mike Bostock Or, why random comparators are bad (in addition to being slow). This matrix diagram visualizes the probability of each element i being placed in position j at the end of a random shuffle. Red cells represent negative bias ; green cells represent positive bias . A good shuffling algorithm is unbiased ; you should see a uniform grey matrix with only a bit of white noise. A bad algorithm will show streaks in the matrix, indicating an un­even distribution. For more details, read my previous post on the Fisher–Yates shuffle . Also see screenshots of different browsers . Built with d3.js . Choose an algorithm: Or write your own: refresh auto-refresh Comments? Discuss on HN! January 21, 2012 Mike Bostock ...

January 14, 2012 Mike Bostock Say you had a fresh pack of cards: If you want to play a game of Texas Hold ‘em with friends, you should shuffle the deck first to randomize the order and insure a fair game. But how? A quick way of seeing an algorithm’s bias is a matrix diagram . A simple but effective way of doing this is to pull a random card from the deck repeatedly and set it aside, incrementally building a new stack. As long as you pick each remaining card from the deck with equal probability, you’ll have a perfectly-unbiased random stack when you’re done: Click to shuffle! But let’s say instead of a physical deck of cards, you wanted to write code to perform this same task with an in-memory array of n elements. Sounds straightforward (in part), but how would you pick a random remaining element from the original deck, exactly? One slow option—gotta start somewhere: pick a ran...

~?~