3 Ways Analyze Dice Probability Custom Polyhedral

You Built a Custom Dice System — Now You Can't Tell If It's Broken

Picture this: you're three sessions deep into playtesting your homebrew tabletop RPG. You designed a custom dice pool mechanic using a d14 for martial attacks and a d16 for spellcasting, with exploding dice on max rolls. Your players keep telling you the martial class feels "swingy" while the spellcasters seem consistently overpowered. You've rolled these dice maybe two hundred times by hand across those sessions, scribbling results in a notebook that now looks like a caffeinated seismograph.

Here's the uncomfortable truth: two hundred hand-rolled results, scattered across notebook pages, with no aggregation, no filtering, and no way to separate "attack rolls" from "damage rolls," is not data. It's noise. You cannot balance a game system on noise.

And that's exactly where most independent game designers hit a wall. You can feel that something is off in your custom polyhedral dice mechanics, but you can't prove it, and you certainly can't fix what you can't measure.

Why Hand-Rolling Fails for Probability Analysis

Let's be clear about something. There's nothing wrong with rolling physical dice during actual play — that's the whole point of the hobby. But when you're trying to analyze dice probability for a custom system, manual rolling has three structural problems that no amount of patience will solve.

The Sample Size Problem

To get a statistically meaningful picture of a single die's distribution, you need hundreds — ideally thousands — of rolls. For a standard d20, you'd want at least 1,000 rolls to start seeing whether your die is fair within a reasonable confidence interval. Now multiply that by every custom die in your system. If you're running a d14, a d16, and a d18 with exploding mechanics, you're looking at 3,000+ rolls minimum. Nobody has time for that.

The Recording Problem

Even if you had the time, human transcription is unreliable. You'll misread a 13 as an 18. You'll skip a row. You'll accidentally roll your d12 when you meant to grab the d14. And once your notebook is full of errors, your analysis is built on a foundation of sand.

The Aggregation Problem

Say you somehow roll 500 times and record everything perfectly. Now what? You've got a list of numbers. To turn that into a probability distribution, you need to count frequencies, calculate percentages, compare them to theoretical expectations, and identify outliers. Doing that by hand for multiple dice types is a spreadsheet nightmare waiting to happen.

This is why a free online roller with roll history isn't just a convenience tool — for a game designer, it's a probability analysis engine hiding in plain sight.

How to Analyze Your Custom Dice Using an Online Roller

Let's walk through this step by step. I'm going to use a concrete example from a real homebrew system I helped balance last year: a game called Tideforged that uses a custom d14 for weapon attacks and a d16 for spell attacks, with an exploding mechanic where rolling the maximum value lets you roll again and add.

Step 1: Define What You're Actually Testing

Before you roll a single die, write down the specific questions you need answered. Vague goals produce vague data. For Tideforged, the designer had three questions:

• Is the average attack roll for the martial class (d14, exploding) actually lower than the spellcaster (d16, exploding)?
• How often does the exploding mechanic trigger for each die, and does that frequency match the theoretical probability?
• What's the practical maximum roll for each die type across 1,000 rolls — i.e., what's the worst-case scenario a GM needs to prepare for?

Notice how specific these are. "Is my game balanced?" is not a testable question. "Is the average d14 exploding roll lower than the average d16 exploding roll across 1,000 samples?" is.

Step 2: Configure Your Custom Polyhedral Dice in the Roller

A good free online roller will let you specify custom die sizes — not just the standard d4, d6, d8, d10, d12, d20 set, but arbitrary face counts like d14, d16, d18, or even d24. This matters enormously because the exploding mechanic behaves differently on each die size.

Here's why: the probability of rolling the max on a d14 is 1/14, or roughly 7.14%. On a d16, it's 1/16, or 6.25%. That means the martial class's d14 should explode more often than the spellcaster's d16. But does the higher explosion frequency compensate for the smaller die size? That's exactly what your roll history will reveal.

Set up two separate roll configurations: one for your d14 with exploding enabled, and one for your d16 with exploding enabled. Most online rollers that support roll history will let you label or tag your roll sessions, which becomes critical when you're comparing two different mechanics.

Step 3: Generate a Meaningful Sample

Roll 1,000 times for each die configuration. Yes, all thousand. This is where the online roller earns its keep — what would take you hours by hand takes seconds. Most rollers with roll history features can batch-roll in sets of 50 or 100, so you're looking at about ten batches per die type.

Do not skip this step. I've seen designers try to draw conclusions from 50 rolls, and the results are meaningless. With 50 rolls of a d14, you'd expect each face to appear roughly 3.57 times. The natural variance in such a small sample is enormous — a face might show up 7 times or 0 times, and neither result tells you anything about the die's true distribution.

At 1,000 rolls, each face of a d14 should appear approximately 71.4 times. Now you're in territory where deviations actually mean something.

Step 4: Export and Analyze the Roll History

This is the step most people skip, and it's the most important one. A free online roller with roll history gives you a raw data set — a list of every roll result, typically exportable as CSV or copy-pasteable into a spreadsheet. That data is your goldmine.

Here's what to do with it. Paste your 1,000 d14 results into column A of a spreadsheet and your 1,000 d16 results into column B. Then:

1. Calculate the mean for each column. For a standard d14 (no exploding), the theoretical average is (1+14)/2 = 7.5. For a d16, it's (1+16)/2 = 8.5. With exploding dice, the theoretical averages shift upward — the d14's expected value becomes approximately 8.07, and the d16's becomes approximately 9.07. Compare your actual means to these theoretical values.
2. Build a frequency table. Count how many times each face value appeared. Divide by 1,000 to get the empirical probability. Compare to the theoretical probability (1/14 ≈ 7.14% for each face of the d14, 1/16 ≈ 6.25% for the d16).
3. Identify the explosion rate. Count how many times the max value appeared. For the d14, you should see roughly 71-72 explosions out of 1,000 rolls. For the d16, roughly 62-63.
4. Find the practical maximum. Sort your results descending. What's the highest single roll in your 1,000-roll sample? This tells you the extreme a GM needs to be prepared for.

Step 5: Interpret the Results Honestly

When we ran this analysis on Tideforged, the results were surprising. The d14's empirical average across 1,000 rolls was 8.12 — within 0.05 of the theoretical 8.07. The d16 came in at 9.03, also very close to theory. So the dice themselves were behaving as expected.

But here's what the roll history revealed that the designer hadn't considered: the exploding d14 produced a maximum single roll of 27 (a 14, then another 13) in the 1,000-roll sample. The d16's maximum was 31 (a 16, then a 15). That's only a 4-point difference at the extreme end, which meant the martial class's "lucky moment" ceiling was nearly as high as the spellcaster's — even though the average was noticeably lower.

This explained the "swingy" feeling players reported. The martial class wasn't actually overpowered at the top end, but the gap between its average (8.12) and its maximum (27) was proportionally larger than the spellcaster's gap (9.03 to 31). The variance, not the average, was the problem.

Turning Analysis Into Game Design Decisions

Once you have this data, you can make informed changes instead of guessing. For Tideforged, the designer added a flat +2 modifier to martial attack rolls, which shifted the average without touching the ceiling — preserving the exciting explosion moments while reducing the day-to-day inconsistency players were feeling.

That's the real power of analyzing dice probability with a free online roller and roll history. It transforms a subjective feeling ("martials feel swingy") into a quantifiable observation ("the d14's variance-to-mean ratio is 1.87, compared to 1.71 for the d16") and then into a targeted fix.

You don't need a statistics degree. You don't need expensive software. You need a free online roller that supports custom polyhedral dice, tracks roll history, and lets you export results. Roll a thousand times. Count the frequencies. Compare to theory. The math is simple. The insight is enormous.

Your players will tell you when something feels wrong. Your roll history will tell you why — and exactly where to start fixing it.