Engine Update — Prep Time, a bigger power scale, and why Mxyzptlk beats Superman
Engine Update — Prep Time, a bigger power scale, and why Mxyzptlk beats Superman
Three changes went live today. All of them came from the same problem: the engine was being asked to express things the numbers couldn't say.
1. Prep Time actually does something now
You've always been able to set Prep Time in Battle Options. From today it changes the fight.
The model is: everyone turns up with their standard kit. Batman has the belt. Iron Man is in the suit. Doom is armoured. That's the default, and it's what you'd expect from any matchup.
What preparation buys is different — it's the chance to study a specific opponent and build for them. That's the thing Batman is actually famous for. Not being rich, not having gadgets, but knowing who he's fighting before he fights them.
So the same matchup now gives two honest answers:
Batman vs Iron Man (MCU) — random encounter
Iron Man wins, decisively. The Mark 85's nanotech operates on a tier Batman's preparation and martial arts can't bridge. Technology beats preparation when the technology is the preparation.
Batman vs Iron Man (MCU) — Full Prep
Batman wins, decisively. Unlimited preparation lets him engineer a frequency-specific EMP that permanently severs the arc reactor, collapsing every one of Iron Man's technological win conditions.
Neither answer is a fudge. They're answers to different questions, and the settings decide which one you asked.
The flip side matters too: a random encounter now means what it says. Prep-based counters no longer appear in fights where nobody had time to prepare.
2. The power scale goes to 13
This is the big one.
A simple 1–10 style scale works well for most of a roster, and we started there too. The trouble is that the scale is roughly logarithmic — the gap from 1 to 2 is "human to superhuman", and 5 to 6 is "planetary to stellar". Which means by the time you reach the top, you're trying to fit universal → multiversal → conceptual → top-of-cosmology into the last two numbers.
It runs out exactly where the most-argued characters live.
The result was a bracket 10 containing MUI Goku, Whis and Jiren alongside the Living Tribunal, Zeno and Cosmic Armour Superman. Jiren and the Living Tribunal had the same number. That's obviously wrong, and it was distorting every top-tier fight.
So we've added three floors to the top of the building:
Bracket Class Examples
13 Supreme / Transcendent Living Tribunal, Zeno, Azathoth
12 Abstract / Metafictional Pre-Retcon Beyonder, Cosmic Armour Superman, Molecule Man, Lucifer, Featherine
11 Multiversal+ Anti-Monitor, Rune King Thor, The Spectre, Franklin Richards, Alien X
10 Universal / extreme cosmic MUI Goku, Whis, Beerus, Jiren, Broly
Nothing below bracket 9 moved. If you fight street-level or city-level characters, today changed nothing for you. This was entirely about giving the top end room to breathe.
Two things worth saying about how brackets work, because they get misread:
Bracket is not a ranking. It answers "what scale of entity is this?" — not "who wins?". Two characters at bracket 12 aren't equal; they're in the same class, and then abilities, resistances, bypasses and win conditions decide the actual fight.
We're being deliberately stingy with 12 and 13. If a character is arguably 11 or 12, they go in 11 until the feats prove otherwise. Otherwise in six months we'd have seventy characters at 13 and we'd have recreated exactly the problem we just fixed.
One side effect worth calling out: this gives characters like Thanos a proper ladder. Infinity Gauntlet at 10, Heart of the Universe at 11, Astral Regulator at 12 — three versions, three genuinely different classes of threat.
3. Why Mister Mxyzptlk beats Superman
Here's a fight the old engine got wrong, and it explains the third change.
Mxyzptlk is bracket 8. Superman is bracket 7. On raw scale that's nearly a coin flip — and everyone knows that's nonsense. Mxyzptlk does whatever he likes to Superman and always has.
The reason isn't power. It's that they're not operating on the same layer. Superman is in the story. Mxyzptlk is a fifth-dimensional imp who treats the story as material. Superman can punch anything inside reality; he has no answer to something standing outside it.
No amount of bracket adjustment expresses that. So characters now carry a frame:
In-universe — bound by physical law. Almost everyone.
Meta — operates on the narrative layer. 17 characters, hand-checked.
Absolute — the authors. One-Above-All, The Presence. Unchanged, still above everything.
A meta fighter has frame priority over an in-universe fighter — unless that opponent has a demonstrated ability to reach the layer they're operating on.
That exception matters, so here it is working. Mxyzptlk and Sentry are both bracket 8. Frame priority would hand it to Mxyzptlk — except Sentry has reality warping and molecular disassembly, which the engine recognises as sufficient reach into that frame. So it stands aside, the fight resolves on normal terms, and Sentry wins it on his own win conditions.
Certain forms of conceptual erasure, soul interaction, domain effects, reality manipulation, narrative manipulation and toon force can provide that reach, where the character has demonstrated the necessary feats. What you can't do is punch your way there — which is why Goku, two brackets above Mxyzptlk, still loses.
Meta vs meta resolves normally. Living Tribunal (13) still beats Cosmic Armour Superman (12), because when both are on the same layer, scale and win conditions decide it again.
This also settles Cosmic Armour Superman vs Anti-Monitor, which used to come out as a draw. Anti-Monitor is enormously multiversal — but CAS is a Thought Robot built out of narrative, fighting on a layer Anti-Monitor can't reach.
What this doesn't change
The engine still decides every fight, and the AI still only narrates the result it reached. Every verdict shows its full reasoning chain — power brackets, attribute clashes, ability counters, which win conditions were tested and which failed and why.
If you disagree with a result, the useful thing is to disagree with a specific value. "Kill reach should be 7, not 5" is an argument we can settle. That's the whole point of building it this way.
Known and being worked on
Goku's forms are compressed — Base and Super Saiyan currently share a bracket, and the whole ladder from Base to MUI spans too few steps. Next on the list.
Anti-Monitor's classification is under review after the frame change.
Frame reach is currently pass/fail, not graded. A character with a minor soul technique gets the same recognition as one who unmakes souls wholesale. Adding a threshold is on the list.
A few characters are still carrying tags from before these systems existed.
Spot something wrong? Use the Flag button on any result. Several of today's fixes started as flags — including the Prep Time one.
Three changes went live today. All of them came from the same problem: the engine was being asked to express things the numbers couldn't say.
1. Prep Time actually does something now
You've always been able to set Prep Time in Battle Options. From today it changes the fight.
The model is: everyone turns up with their standard kit. Batman has the belt. Iron Man is in the suit. Doom is armoured. That's the default, and it's what you'd expect from any matchup.
What preparation buys is different — it's the chance to study a specific opponent and build for them. That's the thing Batman is actually famous for. Not being rich, not having gadgets, but knowing who he's fighting before he fights them.
So the same matchup now gives two honest answers:
Batman vs Iron Man (MCU) — random encounter
Iron Man wins, decisively. The Mark 85's nanotech operates on a tier Batman's preparation and martial arts can't bridge. Technology beats preparation when the technology is the preparation.
Batman vs Iron Man (MCU) — Full Prep
Batman wins, decisively. Unlimited preparation lets him engineer a frequency-specific EMP that permanently severs the arc reactor, collapsing every one of Iron Man's technological win conditions.
Neither answer is a fudge. They're answers to different questions, and the settings decide which one you asked.
The flip side matters too: a random encounter now means what it says. Prep-based counters no longer appear in fights where nobody had time to prepare.
2. The power scale goes to 13
This is the big one.
A simple 1–10 style scale works well for most of a roster, and we started there too. The trouble is that the scale is roughly logarithmic — the gap from 1 to 2 is "human to superhuman", and 5 to 6 is "planetary to stellar". Which means by the time you reach the top, you're trying to fit universal → multiversal → conceptual → top-of-cosmology into the last two numbers.
It runs out exactly where the most-argued characters live.
The result was a bracket 10 containing MUI Goku, Whis and Jiren alongside the Living Tribunal, Zeno and Cosmic Armour Superman. Jiren and the Living Tribunal had the same number. That's obviously wrong, and it was distorting every top-tier fight.
So we've added three floors to the top of the building:
Bracket Class Examples
13 Supreme / Transcendent Living Tribunal, Zeno, Azathoth
12 Abstract / Metafictional Pre-Retcon Beyonder, Cosmic Armour Superman, Molecule Man, Lucifer, Featherine
11 Multiversal+ Anti-Monitor, Rune King Thor, The Spectre, Franklin Richards, Alien X
10 Universal / extreme cosmic MUI Goku, Whis, Beerus, Jiren, Broly
Nothing below bracket 9 moved. If you fight street-level or city-level characters, today changed nothing for you. This was entirely about giving the top end room to breathe.
Two things worth saying about how brackets work, because they get misread:
Bracket is not a ranking. It answers "what scale of entity is this?" — not "who wins?". Two characters at bracket 12 aren't equal; they're in the same class, and then abilities, resistances, bypasses and win conditions decide the actual fight.
We're being deliberately stingy with 12 and 13. If a character is arguably 11 or 12, they go in 11 until the feats prove otherwise. Otherwise in six months we'd have seventy characters at 13 and we'd have recreated exactly the problem we just fixed.
One side effect worth calling out: this gives characters like Thanos a proper ladder. Infinity Gauntlet at 10, Heart of the Universe at 11, Astral Regulator at 12 — three versions, three genuinely different classes of threat.
3. Why Mister Mxyzptlk beats Superman
Here's a fight the old engine got wrong, and it explains the third change.
Mxyzptlk is bracket 8. Superman is bracket 7. On raw scale that's nearly a coin flip — and everyone knows that's nonsense. Mxyzptlk does whatever he likes to Superman and always has.
The reason isn't power. It's that they're not operating on the same layer. Superman is in the story. Mxyzptlk is a fifth-dimensional imp who treats the story as material. Superman can punch anything inside reality; he has no answer to something standing outside it.
No amount of bracket adjustment expresses that. So characters now carry a frame:
In-universe — bound by physical law. Almost everyone.
Meta — operates on the narrative layer. 17 characters, hand-checked.
Absolute — the authors. One-Above-All, The Presence. Unchanged, still above everything.
A meta fighter has frame priority over an in-universe fighter — unless that opponent has a demonstrated ability to reach the layer they're operating on.
That exception matters, so here it is working. Mxyzptlk and Sentry are both bracket 8. Frame priority would hand it to Mxyzptlk — except Sentry has reality warping and molecular disassembly, which the engine recognises as sufficient reach into that frame. So it stands aside, the fight resolves on normal terms, and Sentry wins it on his own win conditions.
Certain forms of conceptual erasure, soul interaction, domain effects, reality manipulation, narrative manipulation and toon force can provide that reach, where the character has demonstrated the necessary feats. What you can't do is punch your way there — which is why Goku, two brackets above Mxyzptlk, still loses.
Meta vs meta resolves normally. Living Tribunal (13) still beats Cosmic Armour Superman (12), because when both are on the same layer, scale and win conditions decide it again.
This also settles Cosmic Armour Superman vs Anti-Monitor, which used to come out as a draw. Anti-Monitor is enormously multiversal — but CAS is a Thought Robot built out of narrative, fighting on a layer Anti-Monitor can't reach.
What this doesn't change
The engine still decides every fight, and the AI still only narrates the result it reached. Every verdict shows its full reasoning chain — power brackets, attribute clashes, ability counters, which win conditions were tested and which failed and why.
If you disagree with a result, the useful thing is to disagree with a specific value. "Kill reach should be 7, not 5" is an argument we can settle. That's the whole point of building it this way.
Known and being worked on
Goku's forms are compressed — Base and Super Saiyan currently share a bracket, and the whole ladder from Base to MUI spans too few steps. Next on the list.
Anti-Monitor's classification is under review after the frame change.
Frame reach is currently pass/fail, not graded. A character with a minor soul technique gets the same recognition as one who unmakes souls wholesale. Adding a threshold is on the list.
A few characters are still carrying tags from before these systems existed.
Spot something wrong? Use the Flag button on any result. Several of today's fixes started as flags — including the Prep Time one.
👁 10 views · 💬 0 replies
Join the discussion — it's free!