1. Conclusion: do not only read the new-hire list; read the resignation list
Pack 9, Revenants of Azvaldt, releases on August 27, 2026. Its story is about a prison. Rotation, however, looks more like a corporate reorganization conducted by someone who discovered the “delete department” button.
New cards enter. Pack 3, Successors of the Omen, rotates out. Witch also receives a nerf to 正常の侵食 on the same day. Therefore the useful equation is not “new card power.” It is surviving shell - rotation losses - nerf losses + Pack 9 gains.
Our pure structural model ranks Nemesis 92 > Elf 90 > Witch 88. After mixing in a current all-cards-revealed external pre-release Tier list, the order becomes Elf 90.0 > Witch 88.6 > Nemesis 88.4.
The strongest consensus is not “Nemesis is definitely number one.” It is that Elf and Witch remain near the top under different assumptions. Nemesis has an excellent surviving AF shell, while Royal is the biggest upside risk because Pirate Royal is rated Tier 1 externally even though our conservative model scores Royal lower.
2. What actually changes on August 27
Rotation removes Pack 3, and several archetypes lose core engines at once.
- Elf: loses Kururu, weakening midgame board control.
- Royal: loses the core Treasure package including Octrice/Synthesizer-type pieces; Treasure Royal effectively disappears.
- Witch: Pack 3 rotation is manageable, but the
正常の侵食nerf reduces the old Magic Hand shell's tutor precision. - Dragon: loses Galmieux, including zero-cost AOE access and late-game damage. This is the star employee leaving with the shared-drive password.
- Nightmare: the Mode/Mixing shell loses its backbone and the late-game Lulunai & Vania-style finisher package.
- Bishop: Crest Bishop disappears with Marwynn, and Evolution Bishop also loses the Winged Lion Statue midgame board swing.
- Nemesis: Destruction Nemesis disappears with Lishenna and related destruction cards, but AF is a different shell and survives.
A dead archetype is not the same thing as a dead class. Nemesis is the clearest example: the Destruction department closes; the Artifact factory keeps running.
3. The 100-point model
The structural score has five components:
- Existing shell: 30 — how complete and powerful the Pack 8 foundation already was.
- Rotation resilience: 20 — whether the archetype keeps its engine after Pack 3 leaves.
- Pack 9 gain: 20 — whether new cards patch weaknesses or create a new win condition.
- Consistency: 15 — dependence on specific pieces, draw order, or fragile combo chains.
- Finishing power: 15 — how reliably the deck can actually end a game.
The “relative win-rate index” is not observed data. It is simply:
50 + 0.5 × (structural score - seven-class mean of 83.43)
This converts score differences into an intuitive scale. Calling 54.3% an actual Pack 9 win rate before Pack 9 exists would not be analytics. It would be fortune-telling with a spreadsheet.
4. Structural ranking
| Rank | Leader | Structural score | Existing /30 | Rotation resilience /20 | Pack 9 gain /20 | Consistency /15 | Finishing /15 | Relative WR index* |
|---|---|---|---|---|---|---|---|---|
| 1 | Nemesis | 92 | 25 | 19 | 19 | 14 | 15 | 54.3% |
| 2 | Elf | 90 | 25 | 16 | 20 | 14 | 15 | 53.3% |
| 3 | Witch | 88 | 28 | 13 | 19 | 13 | 15 | 52.3% |
| 4 | Nightmare | 82 | 23 | 13 | 18 | 13 | 15 | 49.3% |
| 5 | Dragon | 81 | 30 | 10 | 13 | 14 | 14 | 48.8% |
| 6 | Royal | 76 | 22 | 17 | 15 | 12 | 10 | 46.3% |
| 7 | Bishop | 75 | 24 | 10 | 16 | 13 | 12 | 45.8% |
* Not an observed win rate. A pre-release uncertainty of roughly ±3–4 percentage points is reasonable.
5. All seven leaders
Nemesis: Destruction closes, the AF factory gets an extension
Destruction Nemesis loses Lishenna and its supporting destruction package. AF, however, was already a strong independent shell. 弾哭の変貌アイズエデン (Eyes Eden) adds a 7-cost 5/5 that produces a Warden, destroys an opposing follower whenever an Artifact follower enters play, and repeats its Fanfare on Super-Evolution.
That is exactly the kind of upgrade a mature deck wants: do not replace the win condition; patch the defensive hole. This is why the structural model gives Nemesis 92.
The counterpoint is important: Game8's all-cards-revealed pre-release table places Evolution AF Nemesis in Tier 2. So the model may be overvaluing inherited stability versus the explosive new shells.
Elf: Kururu leaves, Magachiyo arrives holding a flamethrower
煙管の咎人マガチヨ is a 3-cost 2/2. Its Fanfare deals 4 to one enemy follower; at Combo 3 it deals 4 to all of them; Super-Evolution gives it Storm.
In plain language: the board clear can also run at your face. That compresses defense and finishing into one slot. Elf loses Kururu, but Pack 9 gives the combo shell a new way to regain tempo and turn that tempo into lethal pressure.
External pre-release evaluation also puts Combo Elf in Tier 1. Elf is therefore the cleanest “model + outside opinion agree” candidate.
Witch: Magic Hand gets nerfed and is transferred to the laboratory
The 正常の侵食 adjustment hurts the old Magic Hand shell's ability to find specific followers. Then Pack 9 introduces 万術の咎人セフィー (Sefy).
Sefy can Fuse cards, spend 2 PP to produce a 没入の実験体, produce two more on Fanfare, and grant a Crest on Super-Evolution that gives the first Specimen each turn Storm. After enough Specimens have entered, later ones gain +3/+3. Published sample builds also combine this with surviving Magic Hand pieces and 旧き天晶カルギデンスラ, preserving very high burst ceilings.
So Witch is not simply “nerfed.” It is reorganized into a new R&D division with a working prototype. Both GameWith and Game8 already publish Experiment Witch lists, and Game8 rates it Tier 1 pre-release.
Nightmare: the old Mode shop closes; an Aggro shop opens next door
The old Mode/Mixing structure takes a serious rotation hit. That would normally push the class down. Yet the all-cards-revealed external list rates Aggro Nightmare Tier 1.
This means Nightmare should be modeled as a class changing speed bands, not preserving its old identity. Our 82 may therefore be conservative.
Dragon: the Pack 8 champion loses the employee who did three jobs
Ramp Dragon starts with a 30/30 existing-shell score because its Pack 8 structure was excellent. Then Galmieux rotates. Losing zero-cost AOE and late-game damage simultaneously reduces how easily Dragon can develop large threats while cleaning the board.
External pre-release evaluation currently puts Ramp and Face Dragon in Tier 3, matching the model's direction. “It was Tier 1 yesterday” is not a valid reason to buy the stock today.
Royal: the model is cautious; Pirates kick open the boardroom door
Treasure Royal disappears with Pack 3. Rally survives much better. Our model gives Royal only 76 because a brand-new Pirate shell is difficult to score from inherited data.
But Game8's all-cards-revealed pre-release table places Pirate Royal in Tier 1. Royal is therefore the largest blind-spot candidate in this model. If Pirates are as coherent as the first lists suggest, Royal can jump several places immediately.
Bishop: demolish two buildings, then start construction
Crest Bishop disappears, and Evolution Bishop loses a major midgame board tool. New Pack 9 cards can add recovery and control tools, but the class has more structural replacement work than the leaders above it. External pre-release evaluation also places Evolution Bishop in Tier 3.
6. Estimated Pack 8 → Pack 9 movement
| Leader | Estimated change | Main reason |
|---|---|---|
| Elf | about +10 | Combo gets removal plus finishing pressure |
| AF Nemesis | about +10 | Strong AF shell gains Eyes Eden defense |
| Nightmare | +5 to +6 | Old shell dies, new Aggro shell is highly rated |
| Witch | 0 to +2 | Nerf largely offset by Specimen engine |
| Royal | 0 to +8 | Conservative model; Pirate Tier 1 could mean large upside |
| Bishop | -1 to -4 | Crest deletion + Lion Statue loss |
| Dragon | -8 to -10 | Galmieux loss is unusually expensive |
Dragon is a downgrade. Nemesis is a departmental reorg. Witch changes job titles. Royal launches a startup. Elf hires someone who appears to have brought a legally questionable flamethrower.
7. Cross-checking external Tier lists — and correcting an earlier interpretation
An important correction: earlier in this conversation, a GameWith detailed ranking was treated as if it were already a Pack 9 pre-release ranking. Rechecking at the August 27 00:05 JST snapshot shows that GameWith's main Tier article still primarily displays the Pack 8 Chronicle of Destiny environment. It should not be cited as Pack 9 live data.
Game8, however, updated an Experiment Witch article at 21:29 on August 26 with an all-cards-revealed pre-release table:
- Tier 1: Pirate Royal / Experiment Witch / Aggro Nightmare / Combo Elf
- Tier 2: Evolution AF Nemesis / Last Words Nightmare / Rally Royal / Spell Witch
- Tier 3: Ramp Dragon / Face Dragon / Evolution Bishop
The overlap with our model is clear: Witch and Elf high; Dragon and Bishop low. The disputed zone is Nemesis versus Royal.
8. Consensus-adjusted score
A structural model tends to underprice brand-new archetypes. To reduce that bias, assign the best external archetype in each class Tier 1 = 90, Tier 2 = 80, Tier 3 = 70, then calculate:
consensus score = structural score × 0.70 + external Tier score × 0.30
| Rank | Leader | 70% structural + 30% external Tier | Relative WR index* | Read |
|---|---|---|---|---|
| 1 | Elf | 90.0 | 53.4% | High in both the model and external pre-release evaluation |
| 2 | Witch | 88.6 | 52.7% | Specimens create a new win condition |
| 3 | Nemesis | 88.4 | 52.6% | AF shell is strong, external Tier is more cautious |
| 4 | Nightmare | 84.4 | 50.6% | Aggro rises even after Mode disappears |
| 5 | Royal | 80.2 | 48.5% | Pirates are the biggest possible model blind spot |
| 6 | Dragon | 77.7 | 47.2% | Losing Galmieux is an HR disaster |
| 7 | Bishop | 73.5 | 45.1% | Must replace Crest and Lion Statue losses |
* Still not an observed win rate. It is a comparison index centered around 50%.
This changes the headline from “Nemesis first” to Elf > Witch ≈ Nemesis > Nightmare > Royal > Dragon > Bishop.
That is a better pre-release conclusion. When new evidence arrives, update the number. If you start bending the evidence to protect the number, congratulations: your spreadsheet has become a religion.
9. What should you build?
Only one deck
Experiment Witch or Combo Elf is the safest evidence-weighted choice. Both score well structurally and externally.
You already own AF pieces
AF Nemesis remains attractive. Reusing a mature shell and adding a few Pack 9 pieces is lower-risk than crafting an entire new archetype on minute one.
You want the new-toy experience
Pirate Royal is the interesting bet. It is the class our inherited-data model understands least and the external list rates most aggressively.
You want to keep playing old Ramp Dragon unchanged
Be careful. Removing Galmieux and swapping two or three cards is the card-game equivalent of deleting the department head and declaring “business as usual.”
10. How this prediction can fail
Five things can reorder the table quickly:
- Undiscovered card combinations, especially inside new shells such as Pirates.
- First/second-player splits that a single average score hides.
- Metagame targeting: the best deck can lose percentage points simply because everyone brings hate for it.
- Execution difficulty: theoretical ceiling and human ladder win rate are not the same variable.
- Deck refinement: day-one 40-card lists and week-one 40-card lists are often effectively different decks.
After release, the useful dataset is usage rate, overall win rate, first/second win rate, matchup matrix, stabilized lists, mulligan behavior, and difficulty. A few thousand matches are far more useful than twenty screenshots of someone going 8-0.
11. Final verdict and FAQ
Pre-release overall favorite: Elf. Second: Witch. Nemesis is essentially tied for the next slot. Nightmare and Royal are the upside new-shell candidates. Dragon is a clear decline from Pack 8. Bishop currently has the hardest rebuild.
Was the earlier “AF Nemesis #1” conclusion wrong?
The pure structural model still puts it first. The overall conclusion changed because we added a newer external all-cards-revealed evaluation. Updating the conclusion when the evidence set changes is the point.
Is 53.4% a real expected win rate?
No. It is a normalized comparison index. Pack 9 live ladder data did not exist at the snapshot.
What is the safest day-one craft?
If maximizing evidence rather than novelty, Experiment Witch / Combo Elf. If you already own most AF pieces, AF Nemesis is also economically sensible.
Pack 9 is about a prison. The scariest mechanic may still be mass layoffs disguised as Rotation.
