Savanna Village Minecraft: 0 Records Say Savanna, 0 Say Savanna Village

How we counted, before you read any number on this page

Every figure below counts records, not worlds. That distinction decides the whole page, so here is the rule in full, before any of the counts.

What a record is. Forty-nine of them. Twenty rows in the coordinate table, each one a seed somebody wrote down after loading it, with a seed code, an edition, a version and whatever structures they found. Twenty-nine archive cards from 2011 to 2019, each one a seed code plus the terrain the person who found it described. Forty-nine records, and no invented entries among them.

What counts as saying savanna. The word itself, in either spelling, in either source, in any position on the record — a biome line, a terrain description, a note, a name. Zero records qualify. We did not count related words as savanna, and the related words are listed separately below so you can see exactly how far the dry-ground vocabulary goes.

What counts as a savanna village. Strictly this: the record has to name a village and attach savanna to it, or state outright that the village is a savanna village. Zero of the forty-nine do, and they fail on both halves of that sentence — zero name savanna at all, and zero name any village type.

Which column a word sits in. This is the part that changes the answer for plains. A row has a name, an edition, a version, a coordinate line and a description. A biome word in the name is a label the finder chose; a biome word in the description is an observation about the terrain. We keep the two apart, because the only plains in the table sits in a name and has no observation behind it.

The count: zero savanna, zero savanna villages, zero acacia

Two sources, one result each, listed in full. There is no third place these words could be hiding.

The bottom two rows are where the search term dies. A savanna village needs two things written on the same record, a village and a biome, and our set has two of the first and zero of the second. There is no overlap to report, because one side of the overlap is empty.

One correction worth making on our own numbers while we are here. The count table on our taiga page lists plains as zero rows out of twenty, and this table lists it as one. Both are true about different columns: that row was counting biome words written in the description text, and this one includes the name column, where row sixteen reads Plains Survival Island. We are flagging the difference rather than quietly keeping whichever number flatters the page, because the whole point of counting the same field twice is that the field has to stay the same.

The eight mentions of savanna on this site: all eight are sentences about absence

Before this page, the word savanna appeared eight times across the thirty-two pages this site had published, on three of them. That number is eight and the number of savanna records is zero, and the gap between them is the entire point.

Every one of the eight is a sentence about the word being missing. None of them is a savanna record, and none of them is even a sentence about a place. Repeating an absence on a fourth page adds words to a site and adds nothing to a dataset: the only honest way to raise this count from zero is to load a seed that has savanna on it and write the word down.

The same test applied to acacia is cleaner still, because it is zero in both directions. Zero in the records, zero in the prose, zero across thirty-two pages. If a single card had ever mentioned an acacia tree, that tree alone would have been enough to put savanna within reach of this page, and it never happened.

Plains: the one row, the one spawn, and the label we refuse to add

Plains is the paired word people search alongside savanna, and our set treats it almost the same way — one appearance, in a column that does not mean what it looks like it means. Two separate places carry it, and neither is a village type.

The trap is the middle row, and it is worth naming out loud. Row one does put the word plains on the record. It puts it on the spawn, at 0, 63, 0, in a phrase about oak and birch being within ten blocks. That is a fact about where you land. It is not a statement that the village is a plains village, and we are not going to quietly upgrade it into one — the same way we do not call the taiga card a taiga village spawn when the taiga is above the player’s head rather than under their feet.

One place on this site has been looser than the rest, and it should be named too. The FAQ on our archive page answers a question about finishing the game by saying the fastest recorded run started at “a plains village”. That is our own sentence and not a record: the record behind it says plains about the spawn and says nothing about the village. Both statements exist on the site today, and the count on this page follows the record rather than the sentence.

Five checks before you call a code a savanna village seed

Five checks, in the order we run them on a record before repeating anything off it. Each one is answerable from the record itself, in about a minute, without loading a world. The point is to separate a code that puts you in a savanna village from a code that has dry ground somewhere on it.

Check 3 is the one that fails everywhere, and it fails on a field rather than on a world — nobody who wrote these records down was filling in a village type, so the column is empty in all forty-nine rows and not merely in the ones we looked at. Check 4 is the one that separates seed 32 from a real observation: Plains Survival Island is a name, and the description under it says grass, trees, sheep and a spawner cave, which is a shape and an inventory rather than a biome. Check 5 is the one that decides whether a record is usable at all, and it is why the one row with a village on it is also the one row worth copying.

The five records, in full

First day on a grass island when the biome word is missing

Every record on this page is missing the same field, so the first day has to be planned without it. The order below is built from what the records actually say — grass, trees, sheep, spawners, caves — rather than from a biome label none of them carries.

1. Count the trees before anything else. The binding constraint on a small island is wood, not space. Seed 32 records a few trees on a small grass island; 2013 records seven and no passive mobs. Wood decides the first night, and neither row names a biome to distract you from that.
2. Take the sheep last, and take them alive. 32 is the only row in the table that records sheep on an island. Three wool is a bed, and a bed is a reset point — worth more on day one than the same sheep as food.
3. Do not plan a village around a spawn you have not seen. Zero of the forty-nine records name a village type, so no dry-ground record here tells you what you would be walking into. If you are on 2016000 the village is at -24, 64, 12 and you can walk to it; everywhere else, assume you are alone.
4. A spawner cave is a farm, not a threat, if you light it first. 32 records an underground cave system with a spawner and gives no coordinate for it. Clear the dark before you go looking, or the first trip down is the last one.
5. Write the biome word down when you get there. Forty-nine records, zero savanna, zero acacia. If you load a dry island and it turns out to be savanna, the single most useful thing you can add to this record is the word itself — that is the only way this count ever moves.
6. Check the edition and version before you blame the code. 2016000 is two different worlds on two editions, and 2013 may not generate at all past 1.13. If the terrain is not where the row says, that is the generator talking, not a broken seed.
7. Two characters is not a safe code. 32 is the shortest seed in the set. Copy it rather than retyping it, and confirm the edition before you load it.

Running a dry open spawn with other people

Open ground splits a group differently from forest, and the reason is in the records. Seed 32 puts sheep and a spawner cave on a small island with no coordinate recorded, which means everybody lands in the same place and then has to agree who eats and who keeps the wool before night falls — the group that settles it on the first day does not spend the second day arguing about it. Seed 2013 is the harder shape for a crowd: seven trees, no passive mobs, three wheat seeds, and a sand-only satellite island, so food is the bottleneck from the first minute.

If you want to run one of these with other people on Java or Bedrock, Shockbyte is a low-cost host that sets up in minutes — use code LAUNCH for 25% off your first month. If you sign up through that link, I may receive a commission at no extra cost to you. It does not change what you pay.