0. Five-second conclusion
Aggro Nightmare was a very good Set 9 launch choice in this test.
A local simulation screened 32 supported builds over 2,324 engine games. Aggro Nightmare ranked first in the supported subset with an 81.0% average and a 63.9% worst matchup. In actual play, the deck opened 5–0, including three wins against Witch, and the later seat-of-the-pants estimate remained around 70–80% overall, with Witch feeling close to an 80% matchup.
Then everybody else started playing Nightmare.
I wanted to hit Witch. I did not ask to fight my own deck. Nightmare is my kid, so seeing it on the other side feels like theft.
When the opponent is also using the default monster-like leader skin, it is not even cute. The mirror is bad for both mathematics and morale.
1. The AI call was directionally very good
The launch test started from 280 public deck submissions and compared 32 builds the engine could represent well enough. Total engine volume was 2,324 games.
| Rank | Deck | Average | Worst matchup |
|---|---|---|---|
| 1 | Aggro Nightmare | 81.0% | 63.9% |
| 2 | Tempo Forestcraft | 60.2% | 22.2% |
| 3 | Ramp/Face Dragoncraft | 59.7% | 27.8% |
That does not mean the real ladder win rate is 81%. The engine could not fully model every new archetype, and the popular Experimental/Sephie-style Witch decks were not represented well enough for a Witch-wide conclusion.
What the model got right was mainly the ordering and direction: “Aggro Nightmare should be the first deck to try.”
The useful workflow is not “AI produced a sacred percentage.” It is “AI narrowed the candidates, then ladder games checked the answer.”
2. The 91.4% result was Pirate Swordcraft-specific
A separate optimization tested 96 Aggro Nightmare candidates and four play plans against 18 fixed Pirate Swordcraft lists. Total engine volume was 5,904 games, including a 1,944-game unseen-seed holdout.
The baseline 40-card list scored 91.4%, with 92.6% going first, 90.1% going second, and 69.4% even against the worst opposing list.
Huge, but not a general ladder win rate. It is a targeted result against those 18 Pirate Swordcraft lists.
More interestingly, replacing one Dark Knight Marcha with one Baal dropped the test to 88.0%, 3.4 points below baseline.
So the lesson is not “91.4%, broken game.” It is:
The current 40 cards are probably not something to casually “improve.”
3. A real 70–80% PvP run is already excellent
Short samples are noisy, so a few sessions cannot prove a true 80% win rate. But if a player keeps landing around 70–80% over dozens of PvP games, the deck is clearly hitting the current field well.
Over 30 games, 70% is 21–9, 75% is roughly 22–23 wins, and 80% is 24–6. Ninety percent is absurd and likely includes substantial run-good.
On the evening of Aug. 27, GameWith placed Aggro Nightmare in A/Tier 2, described it as easy to pilot, and highlighted the Set 9 damage increase plus Suzy-buffed Storm finishers.
That is the interesting part: the deck is not universally labeled the single best deck, yet it can be extremely good for the opponents actually appearing in one player’s queue.
A tier list is global theory. Your queue is local weather.
4. Why the mirror is boring: it drags the win rate toward 50%
Suppose the deck wins 75% against non-Nightmare opponents and the mirror is roughly 50%.
| Nightmare share | Overall win rate |
|---|---|
| 10% | 72.5% |
| 30% | 67.5% |
| 50% | 62.5% |
Nothing happened to the deck. The player did not suddenly become worse.
Customers were replaced by competitors.
That is why three Nightmare mirrors in a row feel awful even when you win some of them.
Everyone should play strong decks. Preferably strong decks named Witch.
5. Why does a low-rated opponent have a complete deck?
Rating and collection size are different variables.
If the displayed number is Class Rating (CR), Worlds Beyond manages CR separately by class. A player can have extensive experience and a large collection while their Nightmare CR itself is not especially high.
So low Nightmare CR does not mean new player and does not mean small collection.
Launch day amplifies this. Established players rarely craft an entire deck from zero. They already own older staples and only need the new pieces.
GameWith’s Aug. 27 21:58 Aggro Nightmare list costs 18,090 to craft from scratch and includes three Garodeth vs. Zeth plus two Suzy.
To a new player, “low crafting cost” at 18,090 can sound like a practical joke. To a veteran who owns most of it, the real cost may be only a handful of missing cards.
6. Around 23:00, the meta did not look like pure Nightmare
The supplied live X search snapshot around 23:00 included a 12-win Experimental Witch report, discussion of Sephie Witch, a comment that the new format seemed to contain many viable decks, and a Midrange Zerael Nightmare showcase.
This is not usage-rate data. A few seconds of a live timeline cannot tell us “Witch is X%” or “Nightmare is Y%.”
But it is enough to reject the idea that the whole server had become one Nightmare deck.
Three mirrors in a row therefore look more like launch-day popularity plus short-term matchmaking variance.
Please redirect that variance toward Witch.
7. “VS-kun” works even if nobody remembers his real name
The full card name is Garodeth vs. Zeth, an 8-cost 8/8 Nightmare Legendary from Set 9.
While in hand, at the end of your turn, if your leader is at 12 health or less, its cost drops by 1.
On play, choose one mode:
- Gain Storm and deal 2 damage to your own leader.
- Gain Ward and deal 8 damage to every enemy follower.
Suzy makes the joke worse: on evolve, she can give a follower in your hand +3/+0.
Practical memory aid:
“The card with two guys fighting gets cheaper in hand and then an 8/8 suddenly runs at your face.”
“VS-kun” is sufficient field terminology.
8. The real mystery: why do two people fighting each other suddenly Storm a third party?
Official lore says Garodeth and Zeth are master and disciple who oppose each other. Garodeth pushes forward against his master, while Zeth tries to stop his disciple.
Excellent. “VS” makes sense.
Then the match starts.
Garodeth: “I will defeat you!”
Zeth: “Come!”
Bystander: “A dramatic master-student duel.”
Both: “…wait, there is an enemy leader over there.”
STORM
Bystander: “WHY ME?”
The artwork shows two people fighting. Mechanically they are one follower. Choose the Storm mode and they temporarily settle their differences to punch an unrelated third party for eight.
Please resolve the “VS” portion of the card first.
9. My VS-kun is a partner. Your VS-kun is counterfeit
Card games have a universal emotional rule.
Your strong card: reliable friend. Opponent’s identical strong card: fraudulent nonsense.
My VS-kun: “Good. Hit that guy.”
Opponent’s VS-kun: “Why do YOU own that?”
The text box is identical. The emotional card is not.
Nightmare also has a strong “this is my class” feeling, so a mirror looks like an intruder using your toys. Add a default monster-style leader skin on the other side and the visual dissonance gets even worse.
This is not a balance issue. It is an ownership-and-aesthetics issue.
10. The good news: mirror experience is starting to matter
After several games, patterns become readable:
- when the opponent enters realistic Storm range,
- when pushing them below 12 health helps discount their VS-kun,
- when board control matters more than face damage,
- how much health must be reserved for a hand-buffed finisher,
- and who can reach lethal first.
A mirror naturally pulls toward 50%, but a player with more mirror experience can still push above it.
The goal does not have to be “learn to love mirrors.”
You are allowed to hate them and become good at them anyway.
11. Final conclusion: the AI found a hit, which is exactly why everyone else should stop using it
The launch-night story is simple.
AI: “Aggro Nightmare looks strongest in the supported test.”
Player: “Fine, I will build it.”
Ladder: “Witch is getting eaten. This deck feels like 70–80%. Good craft.”
Internet: “Nightmare is strong.”
Matchmaker:
Nightmare
Nightmare
Nightmare
Player: “All of you, go play Witch.”
That is normal card-game ecology. Strong deck appears. Players move to it. Mirrors rise. The deck’s free win-rate edge shrinks. Then anti-meta decks arrive.
The logic is fine. The emotion is not.
Nightmare is my kid, so stop putting it on the opponent’s side. VS-kun is my card, so stop playing counterfeit copies. And if Garodeth and Zeth are already fighting each other, settle that dispute before both of you Storm an innocent third party.
Sources
- https://revenants-of-azvaldt.shadowverse-wb.com/ja/
- https://gamewith.jp/shadowverse-wb/503018
- https://gamewith.jp/shadowverse-wb/497197
- https://gamewith.jp/shadowverse-wb/573786
- https://game8.jp/shadowverse-beyond/794169
- https://gamewith.jp/shadowverse-wb/516869
- Local reports:
SET09_ROTATION_SIMULATION_20260827(2).md,SET09_AGGRO_NIGHTMARE_VS_PIRATE_ROYAL(2).md - Community observation: supplied X live-search snapshot around 23:00 JST on 2026-08-27; not a usage-rate dataset.
