Dicing Dangers

Dice Maths

Comparing a bonus, a reroll, and advantage

Three ways of making a character better at a skill, priced against each other on the same twenty-sided die. They are not interchangeable and the gaps are large.

Once a game decides that training should improve a character’s chances, it has three standard tools. It can add a number to the target, let the player roll twice and keep the better result, or let the player roll again after a failure. Each is a modest-sounding benefit, and only one of them behaves predictably across the whole range.

This note prices all three on the same object: a fair twenty-sided die, with every face equally likely. Advantage in the free rules is defined as two d20s with the higher one used (D&D Beyond — Free Rules: Rules Glossary). Against a roll-over target that amounts to succeeding if either die does; in the roll-under form used in this note, the matching rule keeps the lower die, which also succeeds if either die does.

Setting up the comparison

Take a roll-under target of T, so the plain chance is T/20. Then:

  • Bonus. Add b to the target: chance becomes (T + b)/20, capped at 100% and floored at 0%.
  • Advantage. Roll twice, succeed if either roll succeeds. The failure chance is ((20 − T)/20)², so the chance of success is 1 minus that.
  • Reroll on failure. Roll once; if it fails, roll again and use the second result. The failure chance is ((20 − T)/20)², identical to advantage.

That last line is the first surprise. Rerolling on failure and rolling twice up front are the same probability when the second roll is a fresh, independent roll. The two rules feel different — one commits you to a result, the other only rescues a bad one — but the arithmetic cannot tell them apart. The number is the same, whatever the argument at the table.

The numbers

Target Plain +2 bonus +4 bonus Advantage / reroll Advantage as a bonus
3 15% 25% 35% 27.75% about +2.6
6 30% 40% 50% 51% about +4.2
9 45% 55% 65% 69.75% about +5.0
12 60% 70% 80% 84% about +4.8
15 75% 85% 95% 93.75% about +3.8
18 90% 100% 100% 99% about +1.8

The “advantage as a bonus” column is the gain in chance divided by 5 points per face: for a target of 9, (69.75% − 45%) ÷ 5 = 4.95, printed as about +5.0. It is an approximation of what a flat bonus would give, which is why the table prints “about”.

What the table shows

Three facts fall out of the numbers.

Advantage is worth the most in the middle. At a target of 9, advantage adds 24.75 percentage points, which is worth about a +5 bonus, close to its high point (a target of 10 gives exactly +5.0). Near either end of the range it collapses. At a target of 18, advantage adds nine points, worth under +2, because the character was already likely to succeed and a second roll has little left to rescue.

Bonuses do the opposite. A +4 bonus is worth exactly twenty points everywhere it applies, until it runs into the ceiling or the floor. That makes a bonus the tool for a designer who wants a predictable, quotable number: “trained characters are better by four” is a true sentence at every target.

The top saturates. A character at target 18 with a +2 bonus is already at 100%, and a +4 bonus adds nothing more, so the rulebook’s promised +4 is quietly worth only 10 points there. A flat bonus has this problem in a visible form; advantage never reaches 100% on its own (99% at target 18).

Picking one on purpose

Some design advice follows directly.

  • Use bonuses when the game’s fiction is about degrees of training and the sheet should read as a list of competencies.
  • Use advantage when the game wants situational luck to matter — the difference between a careful approach and a rushed one — and when you are content for it to matter most in the middle of the range, which is where most contested rolls live.
  • Use reroll on failure when the fiction is about a character who gets a second try, and accept that on paper it is advantage with more dice handling.

What not to do is treat them as interchangeable currency. A rulebook that offers “either +2 or advantage” as an equal choice is offering +2 against roughly +5 in the middle of the range, and +2 against roughly +2 to +3 at the edges (+2.6 at target 3, +1.8 at target 18). The choice is not balanced; it is a disguised preference for where the game should be lucky.

One more wrinkle

Advantage also changes the spread of the outcome, not just its average. Taking the higher of two d20s makes the standard deviation smaller than that of one d20 (5.77, see OpenStax for the formula; the higher of two d20s works out to about 4.7 by the same method). A player who takes advantage is buying reliability as much as success, and a table that wants dramatic swings should notice that the rule is reducing them.

That is a sentence a designer can act on: if your game’s pleasure comes from wild turns, price advantage highly and use it sparingly. If your game’s pleasure comes from watching preparation pay off, advantage is the cheapest way to buy that feeling, and a flat bonus will disappoint you in the tails.

Method

Each row is computed from the target T: plain T/20, bonus (T + b)/20 capped at 100%, advantage or reroll 1 − ((20 − T)/20)². The table figures and the standard deviation of the higher of two d20s were checked by listing all 400 ordered pairs of two d20 rolls. The Advantage definition was read on 2026-10-31 on the page listed below.

Sources

Written for: a player or designer choosing between a flat bonus and a second roll who needs to know what each one is actually buying.


Sources