What each chart measures, and how the numbers behind it are worked out. Every probability on the site comes from the same place: two dice have 36 equally likely outcomes, and everything else follows from counting how many of those 36 produce a given result.
Pick the value of each die as the roll happens. The moment both dice have a value, the next row appears on its own — there's no button to press between rolls. Every chart updates live as you go.
If you just want to see the tools working, use Auto Roll in the top bar. It throws random dice at up to 10× speed and fills the session for you.
Every chart has a dropdown above it. View All fits the whole session into the visible width, so there's no scrolling however long you play. Last 50 shows only the most recent 50 rolls at a readable spacing, which scrolls sideways.
On the running-total charts the maths never restarts — the window moves the viewport, not the counting, so a hardway tally still shows the session total. The one exception is the Sum Distribution, which is a snapshot rather than a running total, so Last 50 genuinely re-counts the last 50 rolls and rescales the bell curve to match.
No. Everything lives in your browser tab for as long as it's open. Refreshing or closing the page clears the session, and nothing is uploaded anywhere. The Reset button in the top bar clears it deliberately — it asks for a second tap to confirm.
Roll sum & running average
The white line is each roll's total, so it bounces between 2 and 12. The gold line is the running average: every sum so far added together and divided by the number of rolls. That's why it swings wildly at the start and settles as the session grows — each new roll counts for less.
The dashed line sits at 7, the average of all 36 outcomes and also the single most common total. The shaded bands are session markers, not predictions: crossing above 8.5 flags OVERBOUGHT, dropping below 5.5 flags OVERSOLD.
Snapshot, not a running total
Each bar counts how many times that total has actually come up. The gold curve is what the maths expects, in the same units, so the two are directly comparable.
Of the 36 outcomes, exactly one makes a 2 and one makes a 12, while six make a 7 — that's the bell shape. The expected count for any total is rolls × combinations ÷ 36. Over 180 rolls you'd expect about 30 sevens and about 5 twos.
4 of 36 outcomes
A hardway is both dice showing the same face on 2, 3, 4 or 5 — hard 4, hard 6, hard 8 and hard 10. Rolling 4 the hard way (2-2) is not the same as rolling it 3-1, which is why this is tracked separately.
The orange line steps up by 1 each time one lands and holds level otherwise. Four of the 36 outcomes qualify, so the dashed expected line climbs at rolls ÷ 9. Above it the session is running hot on hardways; below it, cold.
6 of 36 outcomes
Counts sevens as they land. Six of the 36 outcomes total 7 — more than any other number, because there are more ways to make it: 1-6, 2-5, 3-4 and each of those reversed.
The dashed line climbs at rolls × 6 ÷ 36, or one seven every six rolls on average. "On average" is doing real work there: long gaps between sevens are completely normal, and the two lines converge over hundreds of rolls rather than dozens.
16 of 36 outcomes
Counts rolls landing on a field number: 2, 3, 4, 9, 10, 11 or 12. Those cover 16 of the 36 outcomes, so the dashed line climbs at rolls × 16 ÷ 36 — just under half.
Worth noticing: the field covers seven of the eleven possible totals but still loses more often than it wins, because the four it misses (5, 6, 7 and 8) are the most common ones. Counting totals is misleading; counting combinations is what matters.
1 of 36 outcomes each
Snake Eyes is 1-1 and Boxcars is 6-6. Each is exactly one of the 36 outcomes — the rarest results on the table — so both share the same dashed expected line, climbing at rolls ÷ 36.
Expect both lines to sit flat for long stretches and then step. Crossing back and forth over the expected line, sometimes by a wide margin, is what a 1-in-36 event looks like over a normal session.
2 of 36 outcomes each
Ace-Deuce is 1-2 and Yo is 5-6. Unlike a pair, each of these can arrive two ways — either die can show either face — so each covers 2 of the 36 outcomes, exactly double the chance of Snake Eyes or Boxcars.
The shared dashed line climbs at rolls × 2 ÷ 36, roughly one every 18 rolls.
Single die faces, 1–6
These count individual die faces, not roll totals. Every roll contributes two faces, so a 60-roll session has 120 dice in the denominator. The percentage shown is that face's share of all dice seen, next to the 16.7% you'd expect if everything ran perfectly even.
Ties are shown together — if three faces are level on the low count, all three appear as cold.
Streaks of 2 or more
Expand any total to see how often it repeated immediately. A "streak" means the same total landing on consecutive rolls — a single isolated roll isn't one, so the distribution starts at length 2.
The cards are colour-coded by how likely the total is, which makes the comparison easy to see: back-to-back sevens turn up regularly, while back-to-back twos are genuinely unusual.
It's where the count would sit if the dice landed exactly according to the odds — nothing more. It is not a target the dice are pulled toward, and it isn't a forecast of the next roll.
Being below the line doesn't mean the result is "due". Dice have no memory of what they've already done, and the gap between actual and expected tends to grow in absolute terms as a session lengthens, even while it shrinks as a percentage.
Almost certainly not — the rarer the event, the more dramatic this looks. A 1-in-36 result can easily be at double or zero times its expectation after 100 rolls, and that's ordinary variance rather than a fault.
The charts that settle quickly are the common ones: the running average and the field line usually track closely within a couple of hundred rolls. Snake Eyes and Boxcars can look wild for a very long time.
No, and it isn't built to. Every chart here is a record of what has already happened. No arrangement of past rolls changes the odds of the next one, because each roll is independent of every roll before it.
It's a way to see variance clearly and understand how the maths behaves over a real session — not a betting system.
Give it more rolls. At 10× speed a few thousand rolls take a couple of minutes, and that's roughly the scale at which the lines start hugging their expected values. Fifty rolls tells you very little — which is itself one of the more useful things the tool demonstrates.