How Damage Reduction Actually Stacks

They multiply, they don't add

A 30% reduction lets 70% through. A 20% reduction then takes a fifth of that 70% — not a fifth of the original.

0.7 × 0.8 = 0.56 gets through  →  44% reduction

Not 50%. And the gap widens with every source you add.

At the defaults — 30%, 20% and 15% — the real figure is 52.4% while adding them up says 65%. That's 12.6 points of defence you don't have.

Work in what gets through

Stopping 30% is awkward to combine in your head. Letting 70% through isn't.

Multiply the fractions that get past each source, then subtract from one at the very end:

0.7 × 0.8 × 0.85 = 0.476  →  52.4% reduction

Share of damage that still gets through is there for exactly this. Once you think in what gets through, the multiplicative behaviour stops being surprising.

The part that trips people up in both directions

Here's where most explanations stop, and where the real insight starts.

Yes — each further source adds fewer percentage points than the last, because it only acts on the shrinking remainder:

Reduction you already have	A 10% source adds
0%	10.0 points
44%	5.6 points
52.4%	4.76 points

So people conclude that stacking reduction is a waste. It isn't.

That same 10% source always multiplies your effective health by 1 ÷ 0.9 = 1.111 — in every one of those rows. It makes you take 90% of the damage you otherwise would, whether you had no reduction or half of it already.

It looks weaker on the tooltip and is worth exactly the same in hits survived.

That's why percentage points are the wrong unit for judging a defensive item, and effective health is the right one.

This chains into an effective health calculator

The two tools do one job each.

This one turns several sources into one honest figure. The Effective Health Calculator on this site turns one figure into hits survived and how far your health bar really goes.

So the output here is the input there. That's deliberate — it's why this tool doesn't report effective health itself.

Comparing two items, properly

Say you're choosing between a 12% reduction and a 15% reduction, and you already sit at 52.4%.

The tooltips say the second is a quarter better. The percentage points say otherwise: the 12% adds 5.7 points and the 15% adds 7.1 — still a quarter apart, but both far below what the tooltip implied.

Now look at effective health. The 12% gives 1 ÷ 0.88 = 1.136×. The 15% gives 1 ÷ 0.85 = 1.176×. Still roughly a quarter apart in the gain over 1.0.

The ratios survive; the absolute numbers don't. That's the useful lesson — comparing two reduction sources against each other is safe on any of these measures, but comparing a reduction source against anything else is only safe on effective health. An item offering 15% reduction and an item offering 20% more health are directly comparable only once both are expressed in effective health, and the reduction's "15" tells you nothing on its own.

Not every game does it this way

Multiplicative is the safe default, and most games use it. But some add sources within a category before applying the category, some cap total reduction, and some put armour and elemental resistance on entirely different curves that don't interact as you'd expect.

Your game's documentation or careful community testing is the authority. This tool assumes clean multiplicative stacking and says so rather than pretending to know your title.

Two practical uses

Valuing a duplicate. If an item grants a reduction you already have from elsewhere, enter it as a separate source. The second copy is worth a good deal less than the first, and this shows you by how much — which is the honest way to compare it against an item offering something completely different.

Deciding between health and reduction. They multiply together too, so doubling your health and halving the damage you take are worth exactly the same. That makes the cheaper option the better one at any given moment, and it's why builds stacking only one of the two lose to builds that keep both moving.

Leave unused sources at zero — a zero source multiplies by one and changes nothing.