Final Fantasy VIII

Level scaling

Updated 2026-09-13

Full spoilers.

This is the page the rest of this guide leans on. Four of the six mechanisms that make FFVIII guides disagree with each other live here, and all four come out of the same design decision.

The short version

FFVII stored an enemy’s HP as a number. FFVIII stores four coefficients and computes every stat from your level at the moment the fight starts.

So when two guides publish different HP for the same boss, neither of them is wrong and neither of them measured badly. They fought it at different levels.

The arithmetic

Every enemy record carries four bytes per stat. The formulas, read out of the shipped files:

HP  = floor(a×L²/20) + (a + 100c)×L + 10b + 1000d
STR = floor(( floor(L×a/10) + floor(L/b) − floor(floor(L²/d)/2) + c ) / 4)     cap 255
VIT = a×L + floor(L/b) + c − floor(L/d)                                        cap 255

VIT’s shape also covers Spirit, Speed and Evasion; Magic uses Strength’s. Every division truncates. HP is uncapped; everything else stops at 255.

Two famous figures reproduce from the coefficients exactly, which is the only reason any of the rest of this is worth printing:

What levelling actually costs you

Take the dinosaur in the Training Centre, which almost every player meets early and keeps meeting:

Its Strength is fourteen times higher at level 100 than at level 1. Your Strength does not do that. Character stats in FFVIII come overwhelmingly from junctions, and junctions come from spell stock, and spell stock has nothing to do with your level.

That is the whole case for staying low, and it is arithmetic rather than opinion. Across all 142 records the median enemy has fourteen times the HP at level 100 that it has at level 1, 33 of them have fifty times or more, and the worst offender multiplies by 474.

The fact that makes a low-level run practical

70 of the 142 enemy records pay zero experience and zero bonus experience. 48 of those 70 still pay ability points.

Every boss in FFVIII is in that group. Boss fights cannot raise your level at all.

Add up the ability points those 48 records carry and you get 1,314 — some of which are alternate versions of the same fight, so the number you can actually bank is a little lower, but the shape is not in doubt. You can fight every boss in the game, take every point of AP, and finish at level 7.

This is the part most published advice gets structurally wrong. “Don’t level up” reads as “avoid fighting”, and it does not mean that. Levelling in FFVIII is a product of random encounters only, which is exactly the category Diablos’ encounter-suppression ability exists to remove, and exactly the category the Card command sidesteps — carding an enemy ends the fight with AP and no experience.

The bookkeeping for anyone who wants the numbers: 72 records pay experience at all, and the biggest single award in the game is 240. 119 records pay AP.

Level bands: the part nobody documented

Each enemy carries three draw lists, three steal lists and three drop lists. Which one is live is decided by two bytes in the record — the level at which the medium band starts and the level at which the high band starts.

Published documentation calls those boundaries unknown and repeats 20 and 30 as folklore. They are stored per enemy, and across all 142:

• 135 enemies: medium at 20, high at 30. The folklore is right for 95% of the game.
• 6 enemies: 30 and 40 — Turtapod, Iron Giant, Behemoth, Torama, Imp, Elnoyle.
• 1 enemy: 35 and 45 — Ruby Dragon, the only one of its kind.

Those seven are not a curiosity. Six of the seven carry ultimate-weapon materials, which means the one place in the game where the boundary matters most is also the one place the folklore is wrong.

What actually changes at a boundary

Only 36 records in the game are level-invariant across all three tables. Everything else moves under you.

The dinosaur again, as the clean demonstration. Its draw list at low level is Fire and Thunder. At 20 it becomes Fira and Thundara. At 30 it becomes Firaga, Thundaga and Quake — a fourth slot that does not exist below 30.

That is the honest counter-argument to the low-level run, and it deserves stating as plainly as the case for it: levelling upgrades what enemies carry. Whether that trade is good depends on whether you junction what you draw, which is a judgement and not a fact.

And 52 enemies take something away

This is the half nobody writes down. A band boundary is not always an upgrade. 52 of 142 records hold an item in the low band’s drop table that is absent from the high band’s entirely. Not rarer. Gone.

The sharpest case runs the other way and is the one this project got wrong first time:

Elnoyle is the game’s standard answer to “where do I farm Energy Crystals”. It is a correct answer above level 40 and a completely wrong one below 30, and no source mentions level.

Pointing the other way, the dragon on the plains outside Galbadia Garden has Dragon Fang in two of four drop slots below level 20, one slot from 20, and none at all from 30. A player who levels past 30 loses that farm without being told it existed.

The one lever you have over it

The bands key off the enemy’s level, and the enemy’s level tracks yours. The two abilities that shift an enemy’s level by five in either direction are therefore the only way to choose which table is live without changing your own party — and if you are sitting just under a boundary, or just over one, that is a real five-level nudge across it.

Stated honestly: the records show the boundary bytes and the stat formulas, and they show the level-shift abilities exist. That the shift is applied before the band is selected is an inference from how the fields are laid out, not something the files state. It is on the playtest sheets for exactly that reason.

The maximum-HP illusion

Guides publish a maximum HP for each boss. Those maximums are not maximums. They are the value at whatever level cap the writer assumed, and across one dungeon alone the assumed caps are 51, 55, 56, 57 and 58.

The clock-tower dragon in the final castle is published at 89,600 by five separate sources. That is its value at level 58. At level 100 it has 140,000.

The cleanest single example: one guide publishes a boss at 38,500 and three publish it at 60,000. 38,500 is that enemy’s exact value at level 57. It is not a disagreement at all.

This is distinct from ordinary level scaling, and the distinction is worth holding onto. Ordinary scaling is two observers at different levels writing down different numbers. This is guides publishing a ceiling that does not exist, and a later reader treating one of them as careless.

The control group

Three bosses in the final castle do not scale at all. Two sit at 10,000 HP and one at 30,000, flat from level 1 to level 100.

Their figures are the best-agreed numbers in that entire dungeon, across every source, with no dissent.

That is as close to a controlled experiment as this subject allows. Remove the scaling and the disagreements vanish. They were never about carelessness.

Fifteen records in the game grow by less than double across the whole level range, and the enemy with the largest flat HP — 30,000 at every level — is also one of the few whose published figures nobody argues about.

What this page cannot tell you

Enemy level is a property of the encounter, not of the creature. The record holds a curve that behaves identically whether the game feeds it level 12 or level 100. So the claims that certain areas run at a fixed level — the two hidden islands always at 100, one late dungeon’s trash always at 1 — are unsettleable from the enemy tables, and this guide flags them rather than deciding them. One source publishes a formula for how the encounter level is derived from your party average; it is the only source that does, and it is presented here as researched rather than settled.

Draw points are not enemy draw lists. They live in field script, which this project has not decoded, so nothing on this page applies to them.

Game Notes

I want to be careful about what the arithmetic does and does not prove, because the low-level run has a reputation as the clever answer and the clever answer is not always the right one.

What the files settle is narrow and solid: enemies scale, you do not scale with them, bosses cannot level you, and a great deal of what enemies carry moves at levels nobody wrote down. What they do not settle is whether you should care. A player who junctions a hundred of everything is over-equipped at any level, and the difference between an easy game and a trivial one is not really a difference.

The interesting thing here was never the advice. It is that this one design decision manufactures disagreement at scale.

Picture the forum thread about Elnoyle. One player says it is the Energy Crystal farm. Another says they have killed it thirty times and never seen one. Both are telling the truth. One of them is a few levels further along than the other, and there is no possible way for either to discover that this is the relevant variable, because the game never mentions levels in connection with drops and the enemy looks and behaves identically either way.

Multiply that by 106 records whose tables move, 101 whose draw lists move, and every boss in the game having no fixed HP at all, and you do not have a community that is careless. You have a community observing correctly and generalising from a sample of one playthrough, which is the only sample anybody had.

The part I did not expect, and the part that changed how I read every disagreement after it, is that this project made the same error. The first version of the Elnoyle section on the Esthar page said the crystal left the drop table at 40 and that the fix was to steal instead. The banding was read correctly off the ids; the direction was inferred, and inferred backwards, because the names had not been decoded yet. It is on that page, marked, rather than quietly fixed — the argument only works if it applies to us too.