KRILLION SCORING

Scoring

A Krillion answer is worth between 10 and 100 points, and the gap between those two numbers is the whole game. This page is every number behind it: the six tiers, how a prompt's answers get sorted into them, and how points turn into metres of water.

The six tiers

Every accepted answer sits in exactly one tier. The tier is a property of the answer, not of how fast you typed it or how many attempts it took.

The six tiers, their point values, and the descent each one buys.
TierPointsDepthWhat it is
🌟 One in a Krillion100400 mThe single designated answer for that prompt.
πŸ¦‘ Deep Cut85340 mTrue obscurity. Most of the bank lives here.
πŸ™ Rare60240 mGenuinely uncommon, but not unheard of.
🐟 Schooler30120 mSolid. Swims with the school.
🀑 Too Clever1560 mThe famously β€œobscure” pick everyone reaches for.
🫧 Plankton1040 mThe answer a room would shout in unison.

A round that times out has no tier and scores zero. It is not a seventh, negative tier β€” running the clock out costs you the round's points and nothing more.

How a bank gets cut into tiers

Each prompt's answers are written in awareness order, most obvious first, and the tiers are cut as shares of the bank rather than as fixed counts. A bounded category like doughnut chains and an unbounded one like cheeses cannot share an absolute cut-off.

Share of each bank per tier, with the clamps that stop small and large banks from distorting.
TierShare of bankMinMax
🫧 Plankton2.2%35
🐟 Schooler6.3%212
πŸ™ Rare11.6%424
πŸ¦‘ Deep Cuteverything elseβ€”β€”

Roughly four fifths of every bank is worth the full 85. That is deliberate. The hard part of a round is not being obscure, it is thinking of anything at all in 25 seconds, so the game pays for that and reserves its punishment for the three or four answers that arrive without any thinking at all.

On name a cheese, 169 answers deep, the cut lands at 4 Plankton, 11 Schooler, 20 Rare and 134 Deep Cut. Cheddar, Mozzarella, Parmesan and Brie are the floor. Every other cheese in the bank, including ones you would not call obscure, pays 60 or more.

Too Clever, and why it pays less than an honest guess

At most two answers per prompt are marked as traps, and a trap overrides whatever tier it would otherwise have sat in, paying 15. These are the answers that have a reputation for obscurity but are reached for constantly because of that reputation β€” the pick that feels like a deep cut and is in fact a reflex.

The budget is deliberately tiny, under 2% of a bank. A trap should catch a reflex, never punish knowledge: a thing you have to actually know is not a trap, however unusual it sounds.

Points into metres

One point sinks you 4 metres, and the result screen reports the metres because that is the number you feel. Seven designated gems would be 700 points and 2,800 m, the trench floor.

Nobody reaches the floor. Each gem is one specific answer out of a bank of 150-odd, seven times over, so a perfect dive is not a target β€” it is the ceiling that stops an excellent dive from looking complete.

The five bands

Your final depth lands you in one of five bands. They are pinned to places in the water column you actually watch slide past on the way down, and to how real dives score, rather than to round numbers.

Band thresholds, in points and in metres of descent.
BandFromDepthVerdict
Shallows00 mThe light was still finding you the whole way down.
Schooler160640 mPast the last of the light.
Thermocline2951,180 mA hundred atmospheres. Real depth.
Deep Cut4631,852 mDown where the water stops moving.
Trench6502,600 mThe floor. One in a krillion.

What the bands mean in practice

The percentile line

The results screen says you dived deeper than some percentage of divers. That number is currently a model, not a measurement β€” it is estimated from your score, because there is not yet a real distribution of finished dives to compare against. It is labelled as estimated in the code and it will be replaced by real numbers once there are enough of them. It is a plausible ranking, not evidence.