Almost every game that has armour converts it to a damage reduction percentage with a curve of the same shape:
reduction = armour ÷ (armour + K)
where K is a constant the designers pick. It is a good curve. It never reaches 100 per cent, so no amount of armour makes you immortal, and it behaves sensibly at zero. It also produces a tooltip that gets misread constantly.
What the tooltip shows
With K set to 100, the first 100 armour buys 50 per cent reduction. The next 100 buys only another 16.7. The next 200 buys 13.3 more. The numbers get smaller and smaller, and every player who reads them concludes that armour has diminishing returns and stops buying it.
That reading is correct about the percentage and wrong about the thing the percentage is for.
What survivability does
What you actually care about is how much raw damage you can absorb before dying. That is your effective health: the health pool an unarmoured character would need to survive the same incoming damage. Work it through and the two formulas collapse into something much simpler:
effective HP = health ÷ (1 − reduction) = health × (1 + armour ÷ K)
That is a straight line in armour. No curve, no diminishing anything. With 1,000 health and K of 100, every single point of armour adds exactly 10 effective health, from the first point to the ten thousandth. The effective HP calculator puts the two columns side by side, and watching one of them bend while the other does not is the quickest way to internalise it.
Why both are true
There is no contradiction. Reduction is the fraction of damage removed, and as it climbs towards 1 there is less fraction left to remove, so of course it slows. Effective health is the reciprocal of what remains, and the reciprocal of a shrinking quantity grows without limit. Halving incoming damage doubles your survivability; halving it again doubles it again. The percentage steps get smaller precisely because each one is doing more work.
So the practical advice is the opposite of what the tooltip encourages. Armour does not get worse as you stack it. The only reason to stop is that health and armour multiply, which means the cheapest improvement is usually whichever of the two you have less of.
Two things that do break it
First, this applies to percentage reduction, not flat reduction. If your game subtracts a fixed amount per hit rather than a percentage, armour is enormously valuable against many small hits and nearly worthless against one large one, and none of the above holds.
Second, effective health is per damage type. Run the calculation separately for physical and magical ratings, because the lower of the two is what kills you, and a character with 9,000 effective health against one and 1,200 against the other does not have 9,000 effective health.
Feed the post-mitigation damage into time to kill to see how long that actually buys you, and for the attacking side of the same arithmetic there is the piece on damage thresholds.