Madain Sari

Story mode. Written for a first playthrough and kept free of spoilers — it will not tell you what happens here, only what the game is asking of you.

Eiko’s village. There is no boss here and nothing can go wrong.

Save before you get here

Madain Sari is the only village in the game with no moogle save point. There are five moogles living in it and not one of them will save your game.

Four things to take

• Libra, a Stellazzio coin, behind the fountain — you have to walk round the back of it deliberately
• Tent, also by the fountain
• Ore and Phoenix Pinion in the room where Zidane and Dagger sit down

The Kirman Coffee

It is in the kitchen, among the cooking supplies.

This is one of three rare coffees an old man called Morrid wants, and you will be carrying it for a while — the quest does not pay out until much later. Take it now. The other two coffees are easier to lose than this one, but there is no reason to leave it sitting here.

Read the Eidolon Wall

It is optional and it is text, and it is the best writing in the village. Eleven eidolons painted on a curved wall with the summoners’ history written around them.

Nothing is hidden in it. Read it or don’t.

Some of this village is locked until later

If you have seen a guide mention a hidden room with a Survival Vest in it, or a moogle handing you an Exploda — those are real, and they are not available yet. You come back here after the Iifa Tree, and that is when the village opens up properly.

Do not spend twenty minutes looking for a room that is not there yet.

Game Notes

Madain Sari is where the game quietly stops being about the war.

Every town so far has had a function. Alexandria had a play and a queen, Lindblum had a festival and a war council, Burmecia had a massacre, Conde Petie had a locked door shaped like a wedding. This one has a kitchen.

There is nothing to do here. You collect a coffee for a man you have not met, read a wall nobody makes you read, and watch five moogles get assigned chores by a six-year-old. The most consequential thing that happens is a conversation.

It is the last quiet place in the game. What is coming after the Iifa Tree is not quiet at all, and this village will be one of the things it happens to.

Completionist mode. Full spoilers, in play order, with every permanent loss called out before you reach it. 2 entries below are marked where our sources disagree with each other — we have not picked a winner for you.

A field script can hold an item the visit cannot reach

This is a new problem, and it is not the one we have been managing.

Our first-visit scripts for Madain Sari carry six item awards: a Tent and the Libra coin at the fountain, an Ore and a Phoenix Pinion in the resting room, a Kirman Coffee in the kitchen — and also a Survival Vest and two Phoenix Downs in a hidden room, plus an Exploda and an Elixir on the path.

Publishers are unanimous that the last four belong to the return visit, after the Iifa Tree. Final Fantasy Wiki says the Survival Vest is in a store room “accessible after the party returns from the Iifa Tree”. Every walkthrough files the Exploda, the Elixir, the hidden room and the Scarlet Hair fight under the second visit.

So the awards sit in the first-visit script and are gated behind a story flag.

That is a different failure from the one we know about. The instruction-count problem makes our table say how many wrongly. This makes it say when wrongly — and a checklist that tells a reader an item is available in an area they are standing in, when it is not, sends them looking for a room that does not open. Over-counting wastes a tick box. Mis-dating wastes twenty minutes and makes the reader think they have broken something.

Rule out of it: an award’s presence in a field script is evidence about the room, not about the visit. Any area the player returns to gets its item list split against publishers before it is written. That is now three distinct ways a revisited location has bitten us — Lindblum’s refilling chests, Pinnacle Rocks’ two halves, and this.

We have split it: this page carries the four items you can actually take, and the other four go on the Eidolon Wall page for the return trip.

The village with no moogle, and a decode we had softened

Final Fantasy Wiki states that Madain Sari is the only town or village in the game with no moogle save point. Five moogles live here and none of them saves.

We inferred months ago that key item id 294 marks moogle save points — about forty sites, every documented one with a moogle. We softened that claim after finding counterexamples at the Treno Arch and Alexandria’s guardhouse, and wrote it up as “co-occurs with, mostly”, inferred rather than read.

There is no id 294 anywhere in Madain Sari’s scripts. Eleven fields, six item awards, no marker.

That is the first negative confirmation the decode has had, and negative confirmations are worth more than positive ones here. Positive hits only showed that 294 turns up where moogles are, which a loosely-correlated flag would also do. This shows it is absent from the one village the corpus singles out as having no save moogle — which is what you would predict if 294 marks the save point specifically, and not what you would expect from a flag that merely tends to follow moogles around.

It does not explain the Treno and Alexandria counterexamples, so the claim stays inferred and stays unpublished as a name. But it is materially stronger than it was this morning, and it got there from an absence.

Scarlet Hair splits the corpus on one digit

The boss is on the return visit, so the fight is written up there. The data problem belongs here, because it is the sharpest instance we have found.

Every other stat is identical across both camps — level 22, MP 5,865, 4,790 gil, zero experience, 9 AP, steal Ether and Poison Knuckles, drops a Tent. The only thing that differs anywhere is a leading 1 on the HP.

Three independent publishers on the low figure, one of them a long-standing boss-specific FAQ. Two on ours, one of which is Caves of Narshe — the fourth time CoN has matched our raw field where others print ten thousand less.

And in the same batch, Hilgigars at 8,106 is confirmed by six publishers with no offset at all — a boss, a scripted fight, zero experience, sitting one area earlier and perfectly clean.

So: the offset is real in the corpus, it is not a property of bosses as a class, and where a record has it the split runs between sources that read files and sources that describe fights. We went looking this batch for a structural marker separating the two sets and found one that matched ten hand-picked records perfectly and then failed on forty-one counterexamples the moment we ran it against the whole table.

Which is the rule working. After Ozma we do not get to publish a pattern that survived only the records we chose to look at.

Corrected after a second research pass. We first wrote that we would render 8,985 because “the low camp is the one with a controller in it.” That was wrong, and it is worth saying so plainly rather than quietly editing it.

The low camp is not playtest reports. It is narrative FAQs that print a stat line and no other granular numbers — and one of them, bover_87’s, prints 8,986, while a fourth GameFAQs boss FAQ prints 6,800, which nobody else supports at all. A cluster that cannot agree with itself to the last digit is not four independent observations; it looks like one old stat sheet copied forward with drift. Against that, Caves of Narshe and Game8 reproduce our entire stat block field for field — level 22, MP 5,865, 4,790 gil, 0 EXP, 9 AP — and land on 18,985.

So the honest position is that we do not know, and the reason we do not know is that both camps are reading rather than playing. The Armodullahan case had Jegged saying “deplete about 800 HP”, written from play. There is no equivalent here, and we reasoned as though there were.

The page renders the dispute rather than a number.

Poison Knuckles settle an argument from two areas ago

Conde Petie’s weapon stall sells Poison Knuckles for 5,000 gil. One publisher told you to buy them; two said not to, on the grounds that the character who uses them is not in your party yet.

You can steal Poison Knuckles from Scarlet Hair, here, shortly afterwards. Five publishers list it as one of his two steal slots, one of them giving a twenty-five percent rate.

So the two-against-one side was right, and for a better reason than the one it gave. It is not merely that the wielder is absent — it is that the weapon is free within an hour, from the fight that introduces him. Nobody in the corpus connects those two pages. It took our steal table sitting next to their shop list.

That is the cross-area blind spot again, in its most expensive form yet: 5,000 gil, at a point in the game where that is real money.

Three Explodas and three Elixirs that are one of each

Our path script carries six award instructions — three Exploda, three Elixir.

No publisher anywhere lists more than one of either. Game8 states “Exploda x1” and “Elixir x1” outright; every other guide uses the singular. And they are not chests at all — a moogle brings them to you, which is exactly the kind of scripted presentation that produces duplicate award paths.

Seventh area the instruction-count rule has caught. It is the most reliable thing we know about our own data.

There is a clean confirmation of the rule sitting right next to it, too. The second-visit script for the same room carries one Exploda and one Elixir — two instructions where the first-visit script has six. Same room, same items, different instruction count. That is about as direct a demonstration as the data is ever going to hand us that the count is an artifact of how a script is written, not of what is in the room.

Kirman Coffee is id 286

Every publisher that describes the kitchen puts a Kirman Coffee in it, among the cooking supplies. Our data awards unnamed key item 286 in that field, and nothing else is there.

It is one of three rare coffees for Morrid’s side quest, wanted by the end of disc 3. One publisher flags the other two coffees as missable and pointedly does not flag this one — which is a useful shape of evidence, since it means the question was in front of the author.

Not added to the named-ids table on this alone: 286 is outside the 300–316 run and we have no second constraint on it. Strong candidate, recorded as such.

The one we could not name

Our hidden-room script carries unnamed key item 258 alongside the Survival Vest and Phoenix Downs.

Publishers describe that room’s contents as the Survival Vest and Phoenix Downs and nothing else. The only candidate that surfaced is the Memory Earring, from a single ambiguous reading of one page that other readings of the same publisher contradict — most sources say Eiko simply hands it to you after the fight, with no room named.

Unresolved. It goes on the return-visit page as an open question, not as an item.

Game Notes

The Poison Knuckles thing is the one worth keeping.

Two areas back, three publishers had an argument about whether to spend 5,000 gil on a weapon. They argued about party composition, which is a reasonable thing to argue about and is not the point. None of them mentioned that the weapon is a guaranteed-ish steal from a mandatory boss fight an hour later, because the shop list lives on one page and the steal table lives on another and nobody owns the question that spans them.

We keep finding this. Stiltzkin’s eight stops, the Stellazzio ladder, the refilling chests, and now a shop price against a steal slot. The corpus is excellent at answering “what is in this room” and has almost no way to answer “should I buy this”, because the second question is never local.

That is not a criticism of guides. It is a description of what a route-shaped document can hold. But it is the clearest argument yet for the evergreen pages we have been putting off — the ones organised by system rather than by place. Everything interesting we have found in twenty-one areas has been at a seam.