Ranked · Balance 27 · Steam build 26572818

What top-500 players pick: units and trainers

Unit presence and trainer picks of the leaderboard's top cohort, how the weekly shares are measured, and what the effects behind them can do for your runs.

Weekly cohort window: 10 Nov 2026 to 16 Nov 2026, Balance 27.

A trainer versus screen.

This page shows which units appear on the final boards of one group of ranked players, and which trainers that group picks. The group is the set of owners in the top 500 places of one pinned leaderboard snapshot. The page uses the finished runs those owners started in the weekly report window. It is not a count of active players, and it is not an MMR band at the time of each match. The title says "picks" because the report measures single units and trainer labels. It does not measure complete builds.

This edition uses the leaderboard snapshot taken at 2026-11-16 03:06:00 UTC. It counts finished ranked runs that started in 10 Nov 2026 to 16 Nov 2026, and the report calculated its values at 2026-11-16 21:41:45 UTC. The lowest MMR in the cohort of that snapshot was 4,363. That cutoff describes that one snapshot. It is not a fixed threshold you need to reach.

Who counts and what we captured

The block above always shows the source dates, or a notice when the source is stale or unavailable. Read it before you use any number on this page.

The snapshot lists 2,000 players, and the cohort holds 500 of them. Only 267 cohort members have observed finished ranked runs in the window; public history is partial, so other members may have played runs we did not capture. Together they contributed 4,582 runs, and 960 of those runs reached 10 wins (21.0%). The remainder holds 2,047 owners, 24,304 runs and 4,774 10-win runs (19.6%). Captured terminal boards cover 100.0% of cohort runs and 100.0% of remainder runs. Per-run MMR is empty for 4,582 cohort runs, so this report cannot place a run in an MMR band at match time.

How the groups are formed and checked:

  • A run belongs to the cohort when its owner sits in the top places of the pinned snapshot. The owner's rank at match time does not matter. The run must still have started inside the weekly report window.
  • The remainder holds every other eligible owner, including owners who do not appear on the leaderboard at all.
  • The report audits the split. It counts owners found in both groups, duplicate runs and runs whose outcome disagrees with their win total. The context block shows that audit.
  • The ordinary Meta view uses a different window and different denominators: Batomon rates count battle-round observations, and trainer rates count runs with a reported win or loss. Its numbers cannot replace the values on this page.

For comparison, the ordinary Meta window is currently 6 Nov 2026 to 19 Nov 2026. The cohort dates above come from the weekly report alone.

Unit presence on captured final boards

A unit's share is the number of finished runs whose captured final board contains that unit, divided by the captured non-empty final boards in that group. A run with two Dracana counts once. Each run has the same weight, so a member with many runs counts more than a member with few. A row appears only when the unit has at least 100 containing runs in the cohort and at least 100 in the remainder.

Final-board unit presence: top-500 snapshot cohort versus other owners, 10 Nov 2026 to 16 Nov 2026

Cyan: snapshot cohort. Gold: other owners. Counts and denominators appear in the table; these are not round win probabilities.

DracanaCohort 19.9%Others 15.7%OnsetraCohort 19.0%Others 13.3%DraconarchCohort 13.6%Others 10.4%SteamscuttleCohort 13.0%Others 17.7%SnapscaldCohort 11.9%Others 15.3%CoalemCohort 11.4%Others 7.0%VelocectCohort 11.4%Others 7.8%PuffloonCohort 11.0%Others 8.1%
Show the numbers as a table
Final-board unit presence: snapshot cohort versus all other owners, 10 Nov 2026 to 16 Nov 2026. At least 100 containing runs in each group. Run-weighted, not player-weighted. Shares and differences round separately from raw counts.
BatomonCohort shareOther shareDifference (pp)Cohort 10-win finishOther 10-win finish
Dracana19.9%
914 / 4,582
15.7%
3,820 / 24,304
+4.225.5%
233 / 914
19.3%
737 / 3,820
Onsetra19.0%
870 / 4,582
13.3%
3,224 / 24,304
+5.730.1%
262 / 870
27.6%
889 / 3,224
Draconarch13.6%
624 / 4,582
10.4%
2,535 / 24,304
+3.227.6%
172 / 624
28.6%
724 / 2,535
Steamscuttle13.0%
597 / 4,582
17.7%
4,290 / 24,304
-4.617.9%
107 / 597
18.0%
771 / 4,290
Snapscald11.9%
546 / 4,582
15.3%
3,721 / 24,304
-3.419.2%
105 / 546
20.9%
779 / 3,721
Coalem11.4%
521 / 4,582
7.0%
1,690 / 24,304
+4.420.3%
106 / 521
17.9%
303 / 1,690
Velocect11.4%
521 / 4,582
7.8%
1,891 / 24,304
+3.618.0%
94 / 521
16.3%
308 / 1,891
Puffloon11.0%
505 / 4,582
8.1%
1,961 / 24,304
+3.017.0%
86 / 505
18.9%
371 / 1,961
Final-board unit presence: snapshot cohort versus all other owners, 10 Nov 2026 to 16 Nov 2026. At least 100 containing runs in each group. Run-weighted, not player-weighted. Shares and differences round separately from raw counts.
BatomonCohort shareOther shareDifference (pp)Cohort 10-win finishOther 10-win finish
Dracana19.9%
914 / 4,582
15.7%
3,820 / 24,304
+4.225.5%
233 / 914
19.3%
737 / 3,820
Onsetra19.0%
870 / 4,582
13.3%
3,224 / 24,304
+5.730.1%
262 / 870
27.6%
889 / 3,224
Draconarch13.6%
624 / 4,582
10.4%
2,535 / 24,304
+3.227.6%
172 / 624
28.6%
724 / 2,535
Steamscuttle13.0%
597 / 4,582
17.7%
4,290 / 24,304
-4.617.9%
107 / 597
18.0%
771 / 4,290
Snapscald11.9%
546 / 4,582
15.3%
3,721 / 24,304
-3.419.2%
105 / 546
20.9%
779 / 3,721
Coalem11.4%
521 / 4,582
7.0%
1,690 / 24,304
+4.420.3%
106 / 521
17.9%
303 / 1,690
Velocect11.4%
521 / 4,582
7.8%
1,891 / 24,304
+3.618.0%
94 / 521
16.3%
308 / 1,891
Puffloon11.0%
505 / 4,582
8.1%
1,961 / 24,304
+3.017.0%
86 / 505
18.9%
371 / 1,961
Lamplet11.0%
502 / 4,582
12.2%
2,963 / 24,304
-1.216.9%
85 / 502
16.9%
501 / 2,963
Cinnabark10.9%
501 / 4,582
16.1%
3,910 / 24,304
-5.216.6%
83 / 501
19.0%
741 / 3,910

The difference column subtracts the remainder share from the cohort share. The result is in percentage points. It is not a relative lift. A share also tells you how often a unit was present. It does not tell you how likely a board with that unit is to win.

The 10-win column is the share of finished runs containing the unit that reached 10 wins. It does not count battle rounds, and it says nothing about one specific board, including the illustration further down this page.

Which trainers they pick

The four trainers below have catalog effects that change a run in different ways:

  • Monster Ranger: "Start with an Uncommon monster. Get another copy of that monster every 2 days."
  • Swim Coach: gains a random Water monster each day.
  • Twins: when monsters merge into level 3, you gain an exact copy with a sell value of 0.
  • Witch: the first shop rolled each day contains a Potion. The Witch guide covers that pool.

A trainer's share is the number of finished runs with that trainer, divided by all finished runs in the group. It does not use round observations. A trainer row needs at least 100 runs with that trainer on each side.

Trainer picks: snapshot cohort versus all other owners, 10 Nov 2026 to 16 Nov 2026. At least 100 containing runs in each group. Run-weighted, not player-weighted. Shares and differences round separately from raw counts.
TrainerCohort shareOther shareDifference (pp)Cohort 10-win finishOther 10-win finish
Monster Ranger9.8%
447 / 4,582
8.5%
2,069 / 24,304
+1.226.8%
120 / 447
26.6%
550 / 2,069
Swim Coach7.3%
333 / 4,582
6.2%
1,518 / 24,304
+1.030.6%
102 / 333
28.6%
434 / 1,518
Twins6.9%
318 / 4,582
6.8%
1,658 / 24,304
+0.121.1%
67 / 318
20.4%
339 / 1,658
Chemist6.9%
316 / 4,582
6.1%
1,488 / 24,304
+0.820.9%
66 / 316
24.3%
361 / 1,488
Witch6.2%
283 / 4,582
7.2%
1,739 / 24,304
-1.019.1%
54 / 283
14.5%
252 / 1,739
Painter5.6%
256 / 4,582
6.4%
1,550 / 24,304
-0.818.8%
48 / 256
16.6%
257 / 1,550
Gentleman5.3%
244 / 4,582
4.2%
1,014 / 24,304
+1.229.5%
72 / 244
25.1%
255 / 1,014
Egg Breeder5.0%
227 / 4,582
4.8%
1,173 / 24,304
+0.119.4%
44 / 227
19.8%
232 / 1,173
Treasure Hunter4.5%
208 / 4,582
4.5%
1,104 / 24,304
-0.019.2%
40 / 208
19.2%
212 / 1,104
Rich Lady4.4%
200 / 4,582
3.8%
926 / 24,304
+0.626.0%
52 / 200
18.9%
175 / 926

Monster Ranger accounts for 9.8% of cohort runs and 8.5% of remainder runs. Its cohort runs reached 10 wins at 26.8%, and its remainder runs at 26.6%.

Swim Coach accounts for 7.3% of cohort runs and 6.2% of remainder runs. Its 10-win rates are 30.6% in the cohort and 28.6% in the remainder.

Twins accounts for 6.9% of cohort runs and 6.8% of remainder runs, a difference of +0.1 pp.

Witch accounts for 6.2% of cohort runs and 7.2% of remainder runs, a difference of -1.0 pp. Its 10-win rates are 19.1% in the cohort and 14.5% in the remainder.

The cohort 10-win rates count finished runs in the weekly cohort window. They are not the trainer win rates on the Meta page, which use a different window and divide reported run wins by runs with a reported win or loss. To pick between the three trainers the game offers you, use Choosing a trainer.

What these units can do, according to the catalog

The report counts units. It does not record why a player bought a unit or which partners sat next to it. The uses below are plausible readings of the catalog effects. Examples use normal forms unless stated otherwise. They are not measured reasons for any preference, and they are not partner measurements. The Dragon guide, the Shield guide, the Ongoing guide and the Aegistruct with Stalagrove guide cover these ideas in depth.

On your board, front is to the right and behind is to the left, in the same row. Board order explains this.

Dracana: Charge to the immediate left

On each cast, Dracana charges the ally directly behind it by 1 second at level 1, 2 at level 2 and 3 at level 3. Dracana cannot receive Charge itself. Its own Damage is low (5 at level 1), so its value comes from the target. Charge advances the target's current cast progress. At the next check, any Charge beyond the cooldown is discarded. A slow caster behind Dracana therefore keeps more of each Charge than a fast one. The target must be able to receive Charge: Steamscuttle and Ironcore cannot, and neither can another Dracana.

In this edition Dracana is on 19.9% of cohort boards and 15.7% of remainder boards (+4.2 pp).

Onsetra: an Ongoing target to its left

Onsetra makes the ally behind it apply its Ongoing abilities one additional time at level 1. The shiny form affects adjacent allies instead. For normal Onsetra, no Ongoing ally directly behind means its ability adds nothing. Examples of Ongoing targets are Formiqueen in Pantra, and Cinnabark and Furnadon in Jinto. A normal Celestia disables Ongoing Batomon in its row on both teams, which also cancels what Onsetra repeats in that row.

In this edition Onsetra is on 19.0% of cohort boards and 13.3% of remainder boards (+5.7 pp).

Formiqueen gives adjacent Common allies +25% Cooldown Speed at level 1, so it only helps when Common allies sit next to it. Its cohort share is 9.0%, against 6.3% in the remainder.

Coalem: its start cast

Coalem has the On Battle Start ability "Trigger this." Triggering starts its full cast without resetting its cooldown progress. Coalem therefore casts once at the start, with 550 Shield and 20 Burn at level 1, then keeps filling its 15-second timer. Its level 4 cooldown is 5 seconds. Because the base timer is long, Charge and repeated start activations matter more for Coalem than for a fast caster.

In this edition Coalem is on 11.4% of cohort boards and 7.0% of remainder boards (+4.4 pp).

Draconarch: repeated adjacent start effects

Draconarch activates the On Battle Start abilities of adjacent allies when it casts. Every cast, Multicast repeats included, also adds cooldown to Draconarch itself for that battle: 6 seconds per cast at level 1, 3 seconds per cast at level 2 (Multicast 2, so 6 seconds per activation) and 2 seconds per cast at level 3 (Multicast 3, so 6 seconds per activation). A neighbour without an On Battle Start ability gains nothing. Coalem casts again each time. Emperooze costs $10, and each activation gives it "+60 Damage and +60 Heal permanently," so every repeat adds a permanent gain.

In this edition Draconarch is on 13.6% of cohort boards and 10.4% of remainder boards (+3.2 pp).

Emperooze has a comparable row this week: 2.3% of cohort boards against 1.9% of remainder boards.

Shield donors and converters

At battle start, Aegistruct increases its own Shield stat for the battle by the Shield stats of adjacent allies: 1x their total at level 1, 2x at level 2 and 3x at level 3, ignoring other Aegistruct. This stat increase is not an immediate team Shield grant. Aegistruct's later casts use the enlarged stat to give your team Shield. Stalagrove gains Damage for the battle when your team receives Shield: 15% of the amount at level 1, up to 45% at level 3. Stalagrove does not need to sit next to the Shield source.

In this edition Aegistruct is on 8.5% of cohort boards and 3.8% of remainder boards (+4.7 pp).

In this edition Stalagrove is on 7.9% of cohort boards and 4.0% of remainder boards (+3.9 pp).

Steamscuttle, Cinnabark and Furnadon: Jinto triggers

Steamscuttle charges adjacent Fire and Water allies. Cinnabark gains additional Shield stat based on the enemy's Poison stacks. Furnadon gains Burn for each different Trinket you own. All three are Jinto units.

Steamscuttle is on 13.0% of cohort boards and 17.7% of remainder boards (-4.6 pp).

Cinnabark is on 10.9% of cohort boards and 16.1% of remainder boards (-5.2 pp).

Furnadon is on 7.3% of cohort boards and 10.3% of remainder boards (-3.1 pp).

A lower share does not make a unit weak. The report does not say why these players chose other units.

An illustrative Pantra board

This board is a proposal for explaining effects. It is not a top-500 comp, it is not an observed complete lineup, and it is not an optimal board. All six units can appear in a Pantra shop. The board leaves out Onsetra on purpose: nothing here gives Onsetra an Ongoing ally to repeat. High-share units do not all belong on one board.

  1. DraconarchLv 1
  2. CoalemLv 1
  3. DracanaLv 1
  4. PebblerLv 1
  5. AegistructLv 1
  6. StalagroveLv 1
A proposal to explain effects (Pantra), not observed data, not a top-500 comp and not an optimal board.

Each slot has a role from the catalog:

  • Dracana (top right): charges Coalem, the ally directly behind it, on each cast.
  • Coalem (top middle): casts at battle start through "Trigger this," which gives Shield and applies Burn, then keeps filling its long timer.
  • Draconarch (top left): touches Coalem. Each Draconarch cast activates Coalem's start ability again. Its other neighbour, Pebbler, has no start ability.
  • Aegistruct (bottom middle): touches Coalem above it, Pebbler to its left and Stalagrove to its right. All three carry a Shield stat: 550, 20 and 60 at level 1.
  • Pebbler (bottom left): a $15 Shield caster that touches Aegistruct. Its On Cast increases its own Shield stat by 15 for that battle.
  • Stalagrove (bottom right): turns each Shield your team receives into Damage for that battle. It also donates its own Shield stat to Aegistruct.

Shield absorbs direct Damage before team HP. While you have Shield, it also reduces Burn, Poison and Shock damage by 15% before it absorbs them. For more Shield placements, read the Aegistruct with Stalagrove guide.

What you can use if you are not in the cohort

The data does not show that copying these picks wins runs. Use it as a list of units worth reading, then check each offer against your own board:

  1. Check the price and the shop rank. Draconarch costs $60 and Onsetra $50. The highest rarity in the normal base odds is Rare at Shop Rank 4–5, Super Rare at 6–8 (Coalem, Aegistruct) and Legendary from Rank 9 (Draconarch, Onsetra). Shop rank rises by one each day, and effects such as Rich Lady or Apex Bait add rank. Some trainers, Items and events grant units outside these odds. Check your own trainer's effect before you count on one.
  2. Bridge with cheap pieces. In Pantra, Pebbler ($15) and Emperooze ($10) already fit the triggers above; both are Pantra units, so a Jinto shop offers them only through an effect that adds foreign species. Gentleman removes Common and Uncommon offers from day 4, which closes that bridge.
  3. Do not force an engine before you hold its trigger. Draconarch needs adjacent start abilities. Onsetra needs an Ongoing ally behind it. Dracana needs a target behind it that can receive Charge. Aegistruct needs neighbours with a Shield stat. Formiqueen needs Common allies next to it.
  4. Check your region. Normal offers use your region's roster, unless an effect adds foreign species, such as Smuggler or the Witch's Portal Potion. Steamscuttle, Cinnabark and Furnadon belong to Jinto. Formiqueen, Coalem, Aegistruct and Stalagrove belong to Pantra.

This report has no Item or Trinket data for the cohort, so it cannot recommend one. To turn these picks into a habit, follow the ranked climb plan, and read How to read stats before you compare a share with a win rate.

Dates, refresh and limits

  • Snapshot rank is not historical rank. A member may have played some runs in the window at a lower rating. Tied ratings at the edge of the cohort do not make the cutoff a permanent threshold.
  • Members are not all active. Only matched members contribute runs, and public history and captured final boards are partial.
  • The groups are disjoint. The remainder holds every eligible owner outside the cohort, including owners who are not on the leaderboard.
  • This export measures less than a full build. It has no cohort compositions, partner pairs, Item outcomes or Trinket outcomes. The Comps page uses its own sample and window, and none of its rates is a top-500 result.
  • Refresh timing. Cohort values change when updated report data merges. The normal schedule is Monday 07:20 UTC; an explicit audit refresh can occur later and uses the visible calculation timestamp. Daily site rebuilds redisplay that source and do not recalculate it. When the report is older than eight days, the comparison hides, and a failed new report does not fall back to an older cohort.

A share describes the runs of the past report window, and it is not a forecast for the next week.