fponticelli / dicerun-ts Goto Github PK

Consider the expression d6 / (d6 - 1). Note that the maximum value is 6 but that the cumulative probability of getting 6 or less is only 83%. This is presumably caused by results of NaN not being plotted.

We could:

• Explicitly show the probability of NaN
• Drop the NaN from the denominator, implicitly rerolling in the presence of NaN
• Completely abandon division as a feature since it is probably rarely used

The graphs for the dice rolls seem to break whenever the "bucket size" is more than one.

The graph starts normal and with expected results but then at some point in the sampling it will skip to the left and have the values around double the expected.

I have seen this happen primarily when rolling more than 10 d12s but also when combined with other dice pools in higher amounts.

A new idea has been floating about in D&D circles: the emphasis roll. You roll two d20 and take the one that is furthest from 10, rerolling draws. See https://homebrewery.naturalcrit.com/share/wyL-vuEIDrbC.

Is there a way to model this behavior using the existing expression language? Or do we need to add some special syntax for this?

If I ask dice.run for information about 2d51 I get a distribution ranging from 2 to to 102 with a mode of 52, which is correct. However, if ask dice.run about 2d52 it says that the lowest value bucket is 2-3, the highest is 206-207 and the mode is 104-105. The lowest bucket is correct but the highest and modal buckets are about twice what they should be.

Something similar happens with 3d34 (correct) and 3d35 (about double).

The problem gets worse for higher dice rolls. 10d100 should have a mode of around 505 but shows something like 5050-5059 as the mode.

So these seems to be related to the labeling of the distributions. The error increases linearly with the bucket size. This is supported by the fact that the random result generator is unaffected and correctly generates values in range.

dice.run uses cursor hover UI to tell the user the values and probabilities of each bar in the distributions. On iPad, and presumably other cursorless systems, those numbers are not visible.

Would it be possible to add tap-to-show functionality for those?