Marvel: Darkhold Destiny forecast
Last updated · Scryfall data of Oct 27, 2026 · locked
Predicted Commander popularity for the 8 nonbasic cards of Marvel: Darkhold Destiny in Scryfall's data so far. The set releases April 9, 2027. This forecast assumes these new cards will be Commander legal when the set releases; Scryfall currently labels unreleased cards as not legal.
On past sets this model's scores had a 0.52 rank correlation with the actual popularity for commanders and 0.62 for all cards (1 is a perfect order, 0 is no relation).
Method
For more details see How the tests were run
Power
Power is a card's expected strength in a game, learned from older cards and separate from popularity. Three models see only what is known when a card is revealed: its rules text, type, mana value, colours and a parsed-effect score (lib/card_power.js).
Two read 17lands limited win rates of cards printed before the set; the third reads Cube Cobra draft Elo, which is how often drafters pick a card over others. A cheap language model (Haiku) also rates each card from its text alone, 0 to 10. The percentile from the three models and the percentile of that rating are averaged, so 0.50 is a typical older card.
Trained before 2024 and tested on 3,909 cards from 2024 on, the three models correlated 0.36 with win rate (mana value alone: −0.03), and adding the language model's rating raised it to 0.40 on a 500-card sample. The panel under a card lists the rules-text terms that moved the text model most.
Results
Each category is ranked separately. Model scores are percentiles among older cards, from 0 to 1. Power is separate from popularity (see Method). Select the Blend or Power heading to sort. Select a card to see its image and what raised or lowered its scores.
Commander candidates 3 cards
| # | Card | Bookmark | ||
|---|---|---|---|---|
| 1 | Gargantos, the EndbringerText model 0.98
Pairwise model 1.00
Power 0.87
| 0.99 | 0.87 | |
| 2 | Hela, Mistress of Darkness // Dark TutorText model 0.91
Pairwise model 1.00
Power 0.76
| 0.96 | 0.76 | |
| 3 | Magik, Illyana Rasputin // Darkchild, Demon UncagedText model 0.29
Pairwise model 0.78
Power 0.87
| 0.54 | 0.87 |
Other spells 4 cards
| # | Card | Bookmark | ||
|---|---|---|---|---|
| 1 | The DarkholdText model 0.80
Pairwise model 0.82
Power 0.64
| 0.81 | 0.64 | |
| 2 | Hela's ArcanumText model 0.60
Pairwise model 0.58
Power 0.68
| 0.59 | 0.68 | |
| 3 | Rise from the NetherText model 0.48
Pairwise model 0.58
Power 0.47
| 0.53 | 0.47 | |
| 4 | Penance StareText model 0.41
Pairwise model 0.33
Power 0.60
| 0.37 | 0.60 |
Lands 1 card
| # | Card | Bookmark | ||
|---|---|---|---|---|
| 1 | Temple of KhonshuText model 0.99
Pairwise model 0.95
| 0.97 | – |
Limits
4 of 8 cards still contain text the parser cannot fully interpret. Popularity scores predict how often a card is played, not how strong it is, and are not probabilities. Power is learned from limited win rates and draft picks, so it measures strength in limited play, not in Commander, and ignores synergy with a specific deck. On 3,909 cards from 2024 on, the models had a 0.36 rank correlation with win rate (0.40 with the language-model rating, on 500 cards); most of the spread in win rate is not predictable from card text. Popularity scores will be graded against real popularity after release.