Here is a result that looks like a bug and isn't: switching on an ability that only ever adds a hit can make the calculator report less damage.
Punishing reads: if you retain any critical successes, you can retain one of your fails as a normal success instead of discarding it. A fail becomes a hit. Free damage, surely. Yet with the right supporting rules it costs you damage, and the reason is worth understanding because it generalises well beyond this one ability.
You roll 1 critical, 0 normals, 1 fail. Your weapon is 3 damage on a normal, 4 on a critical. You have Rending ("if you retain any critical successes, you can retain one of your normal successes as a critical success instead") and one FailsToNorms conversion — some rule that modifies a fail into a normal success.
| Line | How it resolves | Result | Damage |
|---|---|---|---|
| Take Punishing | Punishing retains the fail as a normal. That dice is now retained, so Rending can't promote it, and FailsToNorms has no fail left to work on. | 1 crit, 1 norm | 7 |
| Decline it | FailsToNorms modifies the fail into a normal, which stays promotable. Rending then upgrades it to a critical. | 2 crits, 0 norms | 8 |
Same dice, same rules, one damage apart — and the line that uses the optional ability is the worse one.
Kill Team distinguishes retaining a dice from changing one, and a dice can only be retained once. Punishing says "retain one of your fails as a normal success" — so the moment you take it, that dice is spent in the retention sense. Rending asks to "retain one of your normal successes as a critical success instead", and there is nothing left to retain. The full breakdown of which rules lock a dice and which don't is on the Retained vs Modified Dice page.
That alone would make Punishing merely weaker than it looks. What makes it actively harmful is the second ingredient.
Punishing and FailsToNorms both want the same resource — a fail — and they produce different goods from it:
With one fail on the table, they are rivals, and Punishing gets there first. It converts your only fail into the lesser of the two possible normals — and in doing so strands the promotion, which had no other target. You didn't lose a hit; you lost the upgrade on one.
Which means the paradox needs all three of these at once:
| Roll and rules | Take | Decline | Better |
|---|---|---|---|
| 1c, 1f — Rending + FailsToNorms 1 | 7 | 8 | decline |
| 1c, 1f — NormsToCrits 1 + FailsToNorms 1 | 7 | 8 | decline |
| 2c, 1f — Rending + FailsToNorms 1 | 11 | 12 | decline |
| 1c, 2f — Rending + FailsToNorms 1 | 11 | 8 | take |
| 1c, 1f — Rending only | 7 | 4 | take |
| 1c, 1f — nothing else | 7 | 4 | take |
The fourth row is the one that proves the diagnosis. Add a second fail and the conflict disappears — Punishing takes one, FailsToNorms takes the other, Rending promotes the second one's output, and taking Punishing wins comfortably (11 against 8). Scarcity was doing the damage, not Punishing itself.
The last two rows are the ordinary case. When nothing else wants the fail, Punishing is exactly the free hit it appears to be.
Punishing is worded "you can retain", so it is a choice, and a player who saw this coming would simply decline. The calculator used to take it unconditionally, which is why ticking the box could push the damage number down.
It now resolves both lines through the remaining steps and keeps whichever ends better, the same way it already handled the Mystic Scry choice. Ties keep the retention, so every roll where nothing else competes for the fail behaves exactly as before. The decision is made in the dice step, once, as at the table — never adapted to how the damage later turns out.
An ability that is free in isolation is not free when it consumes something another rule wanted. Punishing doesn't take damage away from you directly; it spends a scarce fail on the lower-value conversion and leaves a promotion with nothing to promote. Any time two rules draw on the same pool of dice, "always use it" stops being safe advice — which is exactly why the rules bother to say can.