Compare and improve decks

What you’ll accomplish

Compare is an analysis tool first and a raw diff tool second. By the end of this workflow you can pair decks meaningfully, read coverage deltas in strategic terms, and accept or reject suggestions based on whether they solve a problem you actually have.

Before you start

• At least two deck states to compare — either two versions of the same deck (see Manage deck versions) or your deck plus a benchmark list.
• Recent analysis on both — without an analysis run, compare has nothing to compare structurally beyond the card list.

Steps

1. Pick the right pair

   The comparison’s value depends on the pairing. Useful pairings:

   • Your v1 vs your v2 — testing a specific change
   • Your list vs a benchmark meta list — measuring distance from a known-good build
   • Two builds of the same archetype — comparing engine variants
   • Budget version vs full-power version — quantifying what the budget costs you

   Useless pairings: two unrelated archetypes (the comparison is just “everything is different”); your deck vs a deck from a different format (the meta context doesn’t translate).

   Screenshot pending /img/docs/compare/compare-pick-decks.png

   Comparing apples to apples gets you signal. Comparing apples to oranges gets you noise.

2. Read coverage deltas first

   The coverage delta visualizes the 5-dimension change as before/after pairs. Watch for:

   • Access up + dead-hand down — clean improvement, the change worked
   • Conversion up + brick-risk up — a real tradeoff, decide if you accept it
   • Protection down + answer-density up — also a tradeoff, common in side deck reads
   • Recovery changes — important in grind-game matchups

   A delta with everything green is suspicious — usually means you added a card without giving anything up, which is rare in deck-building. Look for what got worse.

   Screenshot pending /img/docs/compare/compare-coverage-delta.png

   Green-only deltas are usually the model not seeing the cost of the change.

3. Use takeaways to interpret the delta

   Takeaways translate the numerical delta into strategic prose:

   • “starter density increased — opening hand quality up”
   • “payoff brick load decreased — fewer uncastable hands”
   • “interaction density improved — better matchup vs combo”
   • “recovery got weaker — grind games harder”
   • “the compared build uses a different non-engine package”

   Read takeaways before the raw card diff. The diff tells you WHAT changed; the takeaways tell you WHY it matters.

   Screenshot pending /img/docs/compare/compare-takeaways.png

   If the takeaways and the delta seem to disagree, the takeaways usually have the better read.

4. Treat suggestions as diagnoses

   The suggestions panel proposes swaps. A good suggestion is a diagnosis paired with a remedy:

   • Problem: low starter density (38%)
   • Suggestion: add a role-compatible searcher
   • Tradeoff: cut a low-impact payoff or support card

   A suggestion without a stated problem is just an opinion. If you can’t see the diagnosis, don’t accept the swap — read the analysis for the dimension first.

   When you accept a swap, save the result as a new version with a note (Manage deck versions).

   Screenshot pending /img/docs/compare/compare-suggestions.png

   A suggestion you can't justify out loud is one you shouldn't accept.

What success looks like

After a compare session you have a saved new version with a one-line note explaining the change, both versions have fresh analysis, and the takeaways read coherently — you can say what the change does in one sentence without referring back to the card diff.

Common pitfalls

• Reading raw diff first. It biases your interpretation toward “cards changed” thinking. Takeaways first; diff as evidence.
• Comparing decks with no analysis run. The compare view degrades to just card diff — you lose 80% of the value.
• Accepting suggestions without a stated problem. Suggestions without diagnoses are anonymous opinions. Make the model show its work.
• Comparing a brewing experiment against a tournament-tested list. You’ll always lose the comparison and learn nothing. Pair against something at the same maturity level.

Going deeper

Coverage deltas and the suggestion model are documented in Scoring model. Benchmark lists are sourced from YGOPRODECK tournament meta (see Data sources and affiliation) and refreshed weekly.

For the cadence of compare → version → re-analyze, see Recommended workflow.

For the field-by-field reference of the compare surface — pair picker, coverage delta, takeaways panel, suggestion grammar — see Compare (reference).

Next steps

Manage deck versions Save the result of your compare session with intent.

Organize inventory Loop back to inventory after a deck change — staples may have shifted.

Recommended workflow Lock in the cadence so compare doesn't become a one-off.