Gym stops are not equally hard, and "hard" is not a feeling — the official campaign data publishes how often every stop is cleared, with a sample size and a confidence interval attached. This page puts those numbers next to the cores that actually cleared them, so you can plan a run instead of discovering a wall at Sabrina.
How to use the gym prep calculator
- Pick a run filter. The page opens on all difficulties because that is the largest sample — the line above the table prints the game count, the window and the rules version of the file it read. Difficulty 2 and difficulty 1 return smaller files that load faster on a slow connection, so the tool stores the projection it fetched and reuses it for an hour. Your choice is remembered the next time you open the page.
- Pick the stop you are standing in front of. Every stop in the table is listed, gyms, story stops and the final fights alike.
- Optionally type your lines (ids or names, comma separated). Cores that share units with your list get a N of your lines chip. This marks matching cores only — it does not re-sort the table, so the row order stays whatever Sort cores by says.
What the win rate numbers mean
The stop panel shows five things:
- Win rate — the share of attempts at that stop that were cleared.
- 95% confidence interval — the range the true rate plausibly sits in. A 34.7% win rate with a ±0.5 interval is a fact; a 100% rate over 20 games is a rumour.
- Games — sample size. This is the field to check before you trust anything else.
- First try — share of clears that happened on the first attempt. This is what decides whether you can take a stop early.
- Attempts — average tries per clear. The gap between this and "first try" is the cost of learning a stop.
The core table adds vs expected: the difference between the core's actual win rate and what the model expected given the units and items involved. A core at +20 is beating its own components; a core at −30 is a trap that happens to have a lot of games behind it.
What the gym data currently says
Using the window and the rules version printed above the table (all difficulties), a few numbers stand out and are worth knowing before you pick an order:
| Stop | Win rate | 95% CI | Games |
|---|---|---|---|
| Misty | 91.2% | 91 – 91.4 | 154,607 |
| Lorelei | 89.5% | 89.2 – 89.7 | 90,617 |
| Bruno | 77.7% | 77.4 – 78.1 | 97,721 |
| Koga (28) | 44.7% | 44.2 – 45.2 | 61,329 |
| Giovanni (30) | 24.6% | 24.2 – 25.1 | 65,182 |
| Marowak | 26.7% | 26.3 – 27.1 | 74,721 |
| Sabrina (20) | 22.6% | 22.3 – 23.0 | 88,155 |
| The fossil awakens | 24.3% | 23.9 – 24.7 | 76,611 |
The pattern is not "later is harder". Misty is a formality and Sabrina is a wall, and the wall is not about your level — it is about whether your board can survive a Psychic trade. And the two hardest stops in the whole file are not where the story puts them: the lowest clear rate in the window is the Rocket boss at 18.3%, ahead of Sabrina, so the file — not the story order — decides the run order. That is exactly the question the core table answers: filter to Sabrina's stop and sort by win rate to see the line-ups people actually clear it with.
Two warnings about reading the table too literally:
- Popular cores are not necessarily good cores. Cheap units appear in far more games, so a mediocre cheap core can outrank a strong expensive one on raw games. Sort by win rate and check
nbefore you commit. - The window moves. The campaign file covers a two-week window and is regenerated regularly; the tier list file is on a different, slower schedule. Numbers here are always the ones in the file at the moment you loaded the page, which is why the window and rules version are printed above the table.
Using it mid-run: gym order and stops
The tool is built for two moments:
Before the run. Look at the stop you plan to take fourth. If its first-try rate is low, plan to lose one fight there on purpose and spend the gold on the board instead of on a reroll you cannot afford.
At the stop. Type your three best lines. If any core in the top fifteen shares two of them, you are close to a line-up with a proven clear. If nothing shares even one, you are probably about to donate a life — go back to the tier list and check whether your lines place at all.
The gym order page covers the strategic side of this: which stops to take early, and why the order matters more than the individual fight.
Kanto Tactics gym FAQ
What is the hardest gym stop in Kanto Tactics right now?
By clear rate in the 31 October – 17 November 2026 window the lowest stop is Ladrón (24) at 18.3% over 43,701 games, then Sabrina (20) at 22.6% over 88,155, The fossil awakens at 24.3%, Giovanni (30) at 24.6% and Marowak at 26.7%. Misty clears 91.2% of the time, so the difficulty curve is not the story order. Because the window moves, check the live number on the stop panel before you rely on this — the page always prints the window it is using.
Why does the gym data have a run filter?
Because the campaign file is published per difficulty. Mixing them into one number would hide the fact that a stop is trivial on difficulty 1 and brutal on 2. The filter lets you see the sample that matches your run.
Should I prepare differently for difficulty 1 and difficulty 2?
Yes, at least once. Switch the filter and look at the same stop: if its win rate and first-try rate move a lot between difficulties, the stop is gated by enemy scaling and the answer is board strength; if they barely move, the stop is gated by a specific type or item matchup, and the core table will show which. The default all-difficulties view is the largest sample, not your run.
The gym stop I want is missing.
It should not be — the file lists 124 stops and the tool renders every one of them. If a stop genuinely is not in the list, it is not in the official table yet, and this site will not invent a number for it.
Why do some stops show a flag?
The publisher flags stops whose win rate deviates far from the expected value with a large enough sample (strong or weak). The flag is theirs, shown verbatim — it is not this site's opinion.
What does "expected" mean for a gym stop?
The publisher's own model output for a core at a stop, given the units, stars and items. The vs expected column is the residual. It is useful for spotting cores that punch above their components, and useless as an absolute measure of quality.
Can I trust a 100% win rate?
Only together with n. A core with 100% over 20 games has a confidence interval from about 79% to 100%; the same core over 400 games would barely be interesting. The interval column exists precisely so you do not have to do that in your head.
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.