Analyze decks and scores

What you’ll accomplish

Analysis answers a different question than readiness. Readiness asks: can I build this deck with my collection? Analysis asks: does this deck have a healthy structure and playable opening patterns? By the end of this workflow you can read both layers, spot when they disagree, and convert the signal into a concrete deck edit.

Before you start

• A target deck created and populated — see Build decks.
• A working sense of your format’s metagame. Coverage scores are calibrated against format meta; without that context the numbers float.

Steps

1. Start with the coverage radar

   Coverage is a structural read of why the deck should work. Five dimensions, each scored 0–4:

   • Access — how reliably the deck reaches its engine card
   • Conversion — how reliably access becomes payoff
   • Protection — how well the deck plays through interaction
   • Answer density — how well it answers opposing threats
   • Recovery — how well it keeps playing after the first exchange

   The radar’s shape tells you the deck’s style. A combo deck looks like a spike on access and conversion with low answer density. A control deck looks square — moderate on everything, high on answer density and recovery.

   A score is directional. Use it to guide decisions, not to replace testing.

   Screenshot pending /img/docs/analyze/analyze-coverage-radar.png

   A radar with one collapsed dimension is informative. A perfectly square radar usually means the model couldn't tell what the deck is trying to do.

2. Cross-check with simulation

   Simulation samples opening hands (10,000+ by default) and reports observed rates: starter access, extender access, interaction density, playable hand rate, dead hand rate, engine-online rate.

   When simulation agrees with coverage, the structural read is trustworthy. When they disagree — coverage says “high access,” simulation says “starter access 38%” — believe the simulation. The structural model has context boundaries; the simulation is empirical.

   Screenshot pending /img/docs/analyze/analyze-simulation-metrics.png

   A 'good' starter rate depends on the deck. 60% is great for control, mediocre for combo.

3. Read structural risks separately

   Risks aren’t part of the score. They’re warning flags about how the deck behaves in edge cases:

   • Brick risk — too many uncastable hands
   • Single point of failure — entire game plan hinges on one card resolving
   • Normal summon dependence — collapses to any normal-summon negate
   • Extra deck dependence — folds to extra-deck disruption
   • Graveyard dependence — folds to graveyard hate
   • Going-second fragility — needs to go first to function
   • Low follow-up risk — runs out of gas in grind games

   A deck can have acceptable scores and still carry a risk that matters in your local environment. If your locals all play Droll & Lock Bird, “graveyard dependence” stops being theoretical.

   Screenshot pending /img/docs/analyze/analyze-risks-list.png

   Risks are context-dependent. 'Severe' in one meta is 'acceptable' in another.

4. Identify the bottleneck and act

   The action loop is small:

   1. Identify the weakest coverage dimension.
   2. Confirm in simulation that the dimension’s metric is actually low.
   3. Read findings and risks for the same dimension — they often suggest a direction.
   4. Open the Compare and improve flow for candidate swaps.
   5. Save the new list as a deck version before applying the change.

   Screenshot pending /img/docs/analyze/analyze-bottleneck-callout.png

   The suggestion is a hypothesis to test in a new version — not a directive.

What success looks like

You can name your deck’s structural weakness in one sentence, point to the simulation metric that confirms it, and have a candidate edit lined up as a new version ready to test. The full analysis page is a tool for diagnosis, not for grading — if you walked away with a number and no next action, you stopped reading too early.

Common pitfalls

• Chasing a universal “good score.” Different deck styles have different expectations. Combo needs high access and conversion. Control accepts lower conversion if answer density and recovery are strong. Blind-second needs payoff conversion and answer density.
• Ignoring the deck-style inference. Pegasus tags an inferred deck style from card composition. If it labels your combo deck as “control” the analysis is being graded against the wrong rubric — override the inference in deck settings.
• Acting on coverage without simulation cross-check. A 4/4 structural score that the simulation reports as 28% starter access is a deck the model misread. The simulation always wins that argument.

Going deeper

The five dimensions and their weight tuning are documented in Scoring model. The simulation methodology — sample size, opening-hand definition, mulligan rules — lives in the same reference. The format-specific calibration corpus is YGOPRODECK tournament meta lists, joined to current format banlist state.

For why a structural score and a simulation rate can diverge, see Core concepts.

For the field-by-field reference of the analysis surface — every coverage dimension, every simulation metric, every risk type — see Quick Analysis (reference).

Next steps

Readiness and acquisition Turn the analysis into a buy plan.

Compare and improve decks Test the candidate swap against the current version.

Manage deck versions Keep a record of why each change happened.