Raw Data Parsed Data Results
[{counts:[c0, c1, c2], boost:percentage_modifier, flat_boost:flat_modifier}]



Tested in FF, Chrome, and Safari. (Progress bar doesn't work in Safari.)

The two buttons run through different sets of possible rates. The full analysis runs through every integer -- the sparse analysis runs through 1-10, and then every other multiple of 5.

Enter the data as a JSON array of objects. Each object has two fields: counts and boost. counts is an array of the observed number of times [0, 1, 2, etc] of the items dropped, while boost is the bonus item drop you have running. An example, using Yiab's blooper data, is

[{counts: [6, 51, 337, 541], boost: 20},
{counts: [0, 2, 38, 179],   boost: 40},
{counts: [0, 0, 6, 105],    boost: 50}  ]
So the first line, for instance, states that 6 times no pixels dropped, 51 times exactly 1 pixel did, 337 times exactly 2 pixels dropped, and all three dropped 541 times -- and that all this was at +20 item drop.

Misc. Notes

A given set of rates is shown in the chart if the rate of belief is (a) more than 1% or (b) more than 10% of the most likely set's confidence.

Starting a new calculation will kill the old one, but you can also do that explicitly with 'kill worker'

Progress is pretty coarse grained, most useful if you've put in lots of separate data sets

The code uses eval to interpret the input, so you can do fancy stuff or crash the browser if you like

The parsing is updated anytime you leave the textbox -- the button is just a convenience.

Here are the original coffeescript files for the frontend and the algorithm.