A tier list for an autobattler is only as good as its sample. This one is not an opinion piece: every row below is a line-up that actually placed in the ladder, with its average placement, top-4 rate, win rate and the number of games behind it. The tiers come from the game's own published thresholds, and the table updates from the live data file each time you load the page.
How the tiers are assigned
The publisher groups line-ups by average placement (avg), using three offsets from a reference average. The letters themselves are already in the file — the table reads tier from the API and never re-derives it.
| Tier | Average place | Rule |
|---|---|---|
| S | 3.70 or better | avg < ref − 0.5 (below 3.71) |
| A | 3.71 – 4.00 | ref − 0.5 ≤ avg < ref − 0.2 |
| B | 4.01 – 4.40 | ref − 0.2 ≤ avg < ref + 0.2 |
| C | 4.41 or worse | avg ≥ ref + 0.2 |
The reference average ref is 4.21 in the current file, so the boundaries land on 3.71, 4.01 and 4.41 — each row starts where the previous one stops, with no gap and no overlap. In this window that is 168 S, 127 A, 198 B and 658 C line-ups; C is by far the largest group, which is what a bell curve around the reference average should look like.
Two things follow from that table, and both matter more than the letters:
- The letters are relative, not absolute. A 4.05 average is an A today because the reference is 4.21. If the ladder gets stronger, the same board becomes a B without changing anything. Compare rows inside one load, not across weeks.
- 8 players per lobby means 4.5 is the middle. Average placement below 4.5 is better than breaking even; 4.21 is the reference, not 4.5, because the file uses its own sample's mean.
Reading a row
- avg — average finishing position. Lower is better. This is the single number the tier is derived from.
- top4 — share of games finishing in the top four. The most stable of the three rates, because it does not care whether you were first or fourth.
- win — share of first places. Noisy on small samples: a core with 120 games can swing 3 percentage points on one lucky session.
- n — games in the sample. Read this before you believe anything else in the row.
- units / items — what the core is made of, behind the row: each unit with its star level, how often it appears, and the items players build on it. Two rows with the same average but different items are different answers to the same question.
When two rows disagree with your intuition, n is usually why. A 2.9 average on 80 games is a rumour; a 3.6 average on 40,000 games is a finding.
How to use it without fooling yourself
- Find your units first. Expand a row to see its composition. If your board shares three units with an S row, you are one item away from a known-good core.
- Then find your items. A core that needs a specific finished item is a different plan from one that works with components; check which you can actually build before you commit.
- Check the tier against the sample. A-tier rows with tiny
nare placeholders until the next window; S-tier rows with tinynare interesting, not instructions. - Do not chase the top of C. The bottom half of the table is mostly the same few boards with different items; the useful information there is what not to build.
The table opens on the first 100 line-ups and loads 100 more per click, and every row expands in place to its units and items. So you can also read it backwards: pick the item you can build and see which cores use it. That is the fastest route from "what do I have" to "what should I aim for".
What the tier list cannot tell you
- It cannot tell you what to do in the shop. Ladder aggregates describe finished boards, not the path to them. The beginner guide covers the economy side.
- It cannot tell you which stop to take. Campaign performance is a different file and a different page — see the gym order page and the gym prep calculator.
- It cannot tell you whether a core is legal on your board right now. Levels and shop odds decide that; a tier row is a board you could have, not one you have.
- It is a rolling window. The aggregates end earlier than the campaign file, so a patch that lands inside that gap shows up here first as noise and later as a shift.
FAQ
Why is C the biggest group?
Because the tiers are offsets around the mean. Most boards sit within ±0.2 of the reference average, and that band is B and C by construction — C alone holds 658 of the line-ups in this window. A tier list where most entries were S would mean the thresholds were set wrong, not that the ladder is easy.
What counts as a "line-up" here?
A core: a specific set of units at specific star levels with specific items. Two entries that differ by one item are two line-ups, which is why the table is long and why the item column matters.
Should I copy the S tier exactly?
Only if you can actually build it. An S row usually assumes star levels and finished items you do not have at level 6. The realistic use is to take the shape of the top rows — which types, which carries — and let the shop decide the rest.
Why do the numbers sometimes change between two loads?
The data file is regenerated as the ladder plays. This page asks the feed every time you open it and replaces the table only when a newer answer arrives; if the fetch fails it keeps the current table, its window and the Try again button on screen. The reference average ref is printed in the file, so a shift in the boundaries is visible rather than silent.
Does win rate or average placement matter more?
Average placement, for two reasons: it is the number the tier is derived from, and it is what the ladder rewards. Win rate is the tie-breaker when two rows have the same average — a core that wins slightly more often is usually the one with a higher ceiling.
Where does this data come from?
meta.json in the game's public data feed at einoow.com/api/meta/kanto/, read directly by your browser. The site never re-derives a tier letter and never models a fight: if a row looks wrong, the aggregate is wrong, and the version stamp on the page tells you which file you are looking at.
Comments
No comments yet. If a number here disagrees with your run, say so in the box above — corrections get folded back into the page.