How ScryCheck scores Commander decks
Understanding your power level, bracket estimate, and confidence interval
Our goal
The best games of Commander happen when everyone is on a level playing field. That’s what the Rule Zero conversation is for — but it only works when players honestly understand how powerful their decks are.
Brackets are a great starting point, but they only check whether you have specific cards. A deck can stay under the Game Changers limit while packing premium cards in every other slot — and brackets alone won’t catch that.
ScryCheck evaluates every card in your deck across five scoring vectors, identifies combos, detects themes, and shows you exactly how every point is calculated.
The result is an honest power level assessmentthat gives your pod a common language — so nobody shows up with an unfair edge.
Overview
ScryCheck analyzes Commander decks using a multi-stage pipeline that evaluates card quality, synergy, combo potential, and mana efficiency to produce a power level rating from 1–10.
Every score contribution can be traced back to specific cards and their roles in your deck — nothing is a black box.
How it works
When you submit a deck, it passes through a deterministic multi-stage pipeline. Each stage feeds into the next, building up to a final power level and bracket estimate.
Card parsing
Deck list is parsed, commanders are identified, and card names are normalized against Scryfall data.
Card rating
Each card is evaluated using a layered rating stack: hand-tuned overrides for high-impact cards, LLM-generated category tags for broad coverage, land-type rules, and heuristics derived from oracle text.
Vector scoring
Cards contribute points to five scoring vectors: Speed, Consistency, Interaction, Mana Base, and Threats. Each vector reflects a distinct dimension of deck strength.
Commander synergy
If your commander has a dedicated profile, cards that align with their strategy receive synergy bonuses. Over 100 popular commanders have profiles.
Combo detection
The deck is checked against Commander Spellbook's 100K+ combo variant database. Detected combos are classified by outcome type and card count, then factored into the Threats vector.
Theme & archetype detection
75+ themes (tribal, keyword, and package-based) and 16 archetypes are identified. Your primary archetype influences how vectors are weighted in the final score.
Mana color balance
Card color spread, mana symbol demand, and land production are analyzed to identify fixing gaps and color coverage ratios.
Composition analysis
Mana base quality, card quality distribution, and mana curve efficiency are evaluated. Significant deficiencies in any of these areas cap your maximum power level.
Strategy ceiling
A structural cap is computed from four signals: combo redundancy, commander engine status, commander independence, and infrastructure density. Decks earn headroom above a neutral base — this is what separates a cEDH shell with a theme-first payoff (Bracket 4) from one with a metagame-first payoff (Bracket 5).
Power level calculation
Vector scores, combo weight, card quality, and the strategy ceiling combine into a final power level. A confidence interval reflects how much uncertainty is present in the rating.
Bracket assignment
Hard gates — cards with outsized game impact — set minimum bracket floors independent of power level. Power level can push the bracket higher, but never below the hard gate minimum.
Power level scale
ScryCheck rates decks on a 1–10 scale with decimal precision. Your rating reflects overall deck strength across all five vectors — not just one axis like combos or mana. The ± marginshows the confidence interval: it widens when there’s meaningful uncertainty (many unrecognized cards, weak archetype signal, boundary proximity) and narrows with strong coverage and a clear deck identity.
Top-end competitive Commander: maximal efficiency, compact wins, and robust protection across all axes.
Near-cEDH construction and pacing; viable at cEDH-adjacent tables but not top-tier meta.
Fast, consistent, and resilient; can present and protect strong wins in stronger pods.
High slot efficiency and tight construction; strong execution without the full cEDH profile.
Well-tuned synergy shell with consistent lines and solid fundamentals.
Reliable game plan with intentional card choices; moderate speed and interaction.
Meaningful upgrades and cleaner synergies, though consistency gaps remain.
Comparable to stock precons: clear plan, but slower mana and lower card efficiency.
Core idea exists, but ramp, draw, and interaction still need major refinement.
Little strategic cohesion; mana and card quality are very inconsistent.
Commander brackets
ScryCheck estimates your deck’s Commander Bracket based on the official WotC guidelines. Brackets reflect both what’s in your deck and how powerful it is overall.
How bracket assignment works:
Certain high-impact cards act as hard gates — they set a minimum bracket floor regardless of overall power level. Power level can push your bracket above that floor, but never below it. A deck with no hard gates is bracketed purely by power level.
Power level vs. marker-only reading
WotC’s bracket guidance focuses on specific card markers (Game Changers, combos, mass land denial). ScryCheck uses those markers too — but also factors in overall power level, which can push a bracket higher than the markers alone would suggest.
A common example: a deck with no Game Changers but with high card quality and tight synergies scores PL 5.9 and gets assigned Bracket 3. Under a marker-only reading of WotC’s guidance, that deck would fit Bracket 2. ScryCheck notes this explicitly in the bracket card on your analysis page:
“This deck’s power level pushes it into Bracket 3, though under a marker-only reading of WotC’s bracket guidance it may fit Bracket 2.”
This matters for the Rule Zero conversation. If your pod goes strictly by WotC markers, you’re a B2 deck. If your pod factors in overall deck strength, you’re a B3. Both readings are valid — ScryCheck shows you both so you can have an honest conversation.
cEDH shell vs. cEDH strategy
A deck can run fast mana, free interaction, and tutors — a genuine cEDH shell — while its win condition relies on coin flips, tribal combat, or chaos effects. These decks have high card quality but low strategic reliability. ScryCheck distinguishes them using a strategy ceiling:
- • cEDH shell + metagame-first payoff (redundant deterministic wins, commander is a combo piece, wins without the commander) → Bracket 5
- • cEDH shell + theme-first payoff (coin flips, tribal, chaos, voltron) → Bracket 4 ceiling
The ceiling is shown on your deck analysis as the Strategy Ceiling indicator.
Why a powerful deck can still be Bracket 4
This is the most common surprise in the whole system, and it comes straight from WotC’s own chart. Bracket 4 is already defined as “lethal, consistent, and fast, designed to take people down as fast as possible” — it just does not adhere to the cEDH metagame. Bracket 5 is described as “meticulously designed to battle in the cEDH metagame… often built using existing cEDH knowledge, tools, and/or decklists.”
So the line between 4 and 5 is not speed and not power— a Bracket 4 deck is allowed to be maximally fast. It’s which card pool the deck draws from. That means a deck can win on turn three, run redundant deterministic combos, and still correctly sit at Bracket 4.
ScryCheck measures this directly. We derive the competitive card pool each week from edhtop16 tournament results, then check how much of it your deck runs — measured only against the staples your color identity actually allows, so a mono-green deck is never penalized for not playing Force of Will.
Cards only enter that pool if they’re genuinely distinctive of competitive play, not merely common in it. Sol Ring appears in roughly 45% of tournament decks but 84% of casual ones, so it tells us nothing and is excluded. Force of Will, Mystic Remora, and Tainted Pact are dramatically more likely in a competitive list, so they count.
When this is what’s holding a deck at Bracket 4, the bracket card on your analysis says so and shows your score against the tournament baseline. The Strategy Ceiling tooltip breaks it down further, including how many competitive staples you run out of those available in your colors.
Fully optimized competitive decks with redundant, deterministic win lines. Power level 9 or higher.
Many Game Changers, mass land denial, or deterministic 2-card combos. Strong, focused, and often faster than casual tables expect.
A few Game Changers, or any 2-card combo. Tuned and efficient. Most power-7 decks live here.
Upgraded precons and synergy-focused builds. No Game Changers. Clearly more powerful than a stock precon.
Precons and very casual decks. Theme over optimization.
Game Changers
Game Changers are cards identified by WotC as having outsized impact on Commander games. ScryCheck detects all of them and uses their presence to set your bracket floor — the more you have, the higher the minimum.
They fall into four categories: fast mana that dramatically accelerates your development, efficient tutors that find any card with minimal investment, resource engines that generate overwhelming card or mana advantage, and stax and denial pieces that lock opponents out.
The full official Game Changers list is published by WotC and available on the Commander Brackets page.
Win speed estimate
Every analysis surfaces a three-number win speed estimate next to the power header — best, typical, and contested. It collapses card quality, synergy, mana efficiency, and combo potential into one practical question: how fast does this deck actually close out games?
The fastest realistic win turn with good draws and no opposing disruption — the upper bound for what the deck is capable of.
The expected win turn in a normal multiplayer game where opponents are playing their own plans. This is the number that maps most naturally to WotC’s bracket turn expectations (B1 9+, B2 8+, B3 6+, B4 4+, B5 any).
The estimated turn when opponents are actively interacting and disrupting your plan.
How it’s calculated
The estimate starts from deck-derived signals — velocity, finisher density, combo compactness, average mana value, interaction load, and the calculated power level. Every contributor is shown in the expandable detail under the win-turn card on your analysis page.
For commanders with enough tracked games, ScryCheck additionally anchors the typical estimate against real gameplay data from Playgroup.gg — a public dataset of roughly 933,000 tracked Commander games across about 3,800 commanders. When the engine prediction differs from the empirical average, the typical turn is pulled toward the observed value. The tooltip on the win-turn card shows the exact number of games the anchor is based on, plus the empirical average.
An honest note about the empirical anchor
Playgroup’s data is collected from casual Commander games. The average win turn for a popular commander reflects how a typical casual build actually plays at a real table — including mulligans, real disruption, and pilot variance — but it skews toward casual pacing rather than highly tuned builds.
To prevent this from misleading the estimate for tuned decks, the empirical anchor is weight-reduced for high-power and very low-power builds. A B5-tier Krenko deck shouldn’t be pulled toward the casual-table average for Krenko, and a janky brew shouldn’t either. Mid-power decks — where the empirical anchor is most informative — receive the full blend.
Coverage caveat: about half of the commanders Playgroup tracks meet the minimum trust threshold (40 games and 10 unique pilots). New, fringe, or recently-printed commanders fall back to engine-only prediction until the dataset catches up. The tooltip says so plainly when that’s the case.
Scoring vectors
Every card in your deck contributes to one or more of five scoring vectors. Together they form the radar chart on your analysis page. Each vector is normalized to 0–100, where 100 represents a ceiling that only the most optimized decks approach. Your power level is a weighted combination of all five, modulated by combo weight, archetype, and the strategy ceiling.
Speed
How quickly the deck can develop mana and get ahead of curve.
Fast mana (0–1 CMC) is weighted most heavily; mana dorks and land ramp are weighted lower because they’re removed more easily or operate at sorcery speed.
Consistency
How reliably the deck can find what it needs, when it needs it.
Repeatable draw engines are weighted near-equally with tutors; one-shot draw spells count but carry less weight. A deck with 0 tutors and 0 draw engines will have a low Consistency score regardless of card quality.
Interaction
The deck’s ability to disrupt opponents and protect its own gameplan.
Counterspells are weighted near the top; instant-speed spot removal follows. A stax-heavy deck can score high on Interaction even with few counterspells.
Mana base
Land quality and color fixing — not mana acceleration (that’s Speed). A strong mana base means your lands come in untapped and produce the colors you need.
This is the only vector that can pull your score down: a tapland-heavy mana base lowers the Mana base vector, which can cap your overall power level ceiling. It does not produce a raw negative; it reduces headroom.
Threats
Win potential: how capable the deck is of actually closing games.
Detected combos add a bonus directly to the Threats vector, weighted by compactness and outcome type. A deck with no explicit finishers or combos will have a low Threats score even if its other vectors are high.
What the numbers mean
Vectors are normalized to 0–100 within each category. They’re relative measurements, not counts — a Speed of 85 doesn’t mean you have 85 fast mana sources; it means your acceleration density puts you well into the upper range for Commander.
Not every vector needs to be high — a stax or control deck intentionally lets Threats sit low while maxing Interaction. What matters is whether the profile matches your intended strategy.
Combo detection
ScryCheck uses the Commander Spellbook database to detect combos — 100,000+ documented variants, from 2-card lines to complex setups. Detected combos contribute to your Threats score based on how compact they are and what they actually do.
Card count
Compact combos are weighted more heavily than long ones. A 2-card combo is harder to disrupt and faster to assemble than a 5-card setup — the scoring reflects that.
Outcome type
Combos are classified by what they do: instant wins, deterministic kills, infinite resource generation, defensive locks, or repeatable value. Game-ending outcomes score higher; lock and value loops still count, but carry less weight because they do not immediately win the game.
Redundancy
Combos that share pieces with other combos in your deck receive a bonus. Overlapping lines are more consistent — drawing one piece gets you closer to multiple wins.
Outcome hints
Each combo row in your results shows an icon indicating what the combo does — mana, turns, damage, locks, wins, and more. Hover the icon for details sourced directly from Commander Spellbook.
To prevent permutation inflation, overlapping combo lines are grouped into families with a cap on their combined contribution to your score.
AI insights
Signed-in users can generate an on-demand AI review of their deck, layered on top of the deterministic scoring engine. Where the scoring pipeline above is rule-based and fully auditable, AI insights use an LLM to reason about your specific card choices — things a fixed rule set can’t easily capture.
Mana health
A verdict (solid / risky / weak) plus specific fixing or color-balance concerns.
Role coverage
Strengths and gaps across ramp, draw, removal, and protection.
Synergy analysis
Top synergy clusters, anti-synergies (cards fighting your own plan), and trap cards that look good but don’t fit.
Piloting tips
Mulligan priorities, an early-game plan, key sequencing, and things to watch out for.
Recommended card swaps
Specific cut → add suggestions, each tagged with a reason (anti-synergy, off-plan, inefficient, or upgrade) and a short explanation.
Good to know
- • Requires signing in — it’s free, but tied to your account rather than anonymous analyses.
- • Generation takes up to a minute or two and has a per-deck cooldown before you can regenerate, since each run is a live model call.
- • Suggestions are filtered against your actual deck list before display — fabricated card references, color-illegal swaps, and cuts of cards not in your list are rejected automatically.
- • Unlike the power level and bracket, AI insights are probabilistic, not deterministic — re-running the same deck can surface slightly different wording or emphasis.
- • MTG card interactions are genuinely complex, and LLMs can get specific card details wrong — treat swap suggestions as a starting point for your own evaluation, not a final verdict.
Input formats
Paste list
One card per line. Quantities are optional. Commander can be in the list or entered separately.
Sol Ring 1 Mana Crypt 4 Island Demonic Tutor
Import URL
Paste a link directly from:
- • Moxfield (moxfield.com/decks/...)
- • Archidekt (archidekt.com/decks/...)
How accuracy is measured
Power level is one of the hardest things to measure in Commander because there’s no objective ground truth — the same deck can feel like a 6 or an 8 depending on who you ask. Here’s how ScryCheck approaches measurement honestly.
Reference deck validation
ScryCheck validates against a set of 291 reference deckswith known expected power levels and brackets. Every scoring change is measured against this suite before shipping. Not all 291 carry equal weight — see below.
Independently verified decks (the real number)
39 of the 291reference decks have expected power levels and brackets sourced independently of ScryCheck — from expert reviewers, content creators, or established community consensus. These are the only decks that can measure real accuracy, since their labels weren’t influenced by the engine itself.
These decks also reveal a systematic -0.29 bias: the engine currently tends to under-rate real decks relative to independent expert judgment. That bias is invisible in the larger dataset below because those labels partly trace back to the engine itself — it only shows up once you isolate the independently-sourced set.
Full corpus, for context
The remaining 252 decks have expected values seeded from the engine itself or from unverifiable legacy sources. They’re useful for catching regressions between releases but are partly circular, so we don’t treat them as an accuracy claim. Across the full set of 291:
As of July 2026.
What’s next
The remaining accuracy gap is structural: the engine currently evaluates cards mostly in isolation and only partially accounts for how cards interact with each other and the commander. The active synergy-modeling workstream is building that layer — when it ships, both the independently-verified accuracy and the under-rating bias above are expected to improve. Growing the independently-verified set beyond 39 decks is also an active priority, since it’s the number that matters most.
Empirical win-turn correlation
The win speed estimate is independently checked against real gameplay data from Playgroup.gg. On every build we measure the Pearson correlation between the engine’s typical-turn prediction and the observed average win turn across all commanders with 100+ tracked games (currently 1,116 commanders).
Current baseline (April 2026):
- • r = 0.30 — Pearson correlation, engine vs. empirical (CI floor)
- • MAE 1.95 turns — mean absolute error before the empirical anchor blends in
- • 55% within ±2 turns, 27% within ±1 turn (engine-only prediction)
The correlation is moderate by design: ScryCheck-submitted decks for popular commanders skew tuned, while Playgroup’s avgWinTurn averages over casual-table populations. The gate’s job is to catch regressions (engine drift or data drift), not to claim the engine is 100% accurate at predicting individual game outcomes — which would not be a credible claim about Commander.
LLM cross-validation
To check whether the engine’s accuracy is self-referential, we ran all 252 reference decks through Claude and GPT-4o independently — neither model saw our scores, ground truth values, or thresholds. They evaluated each deck from scratch using only the card list and WotC’s bracket framework.
What the cross-validation found:
- • The engine outperformed both LLMs significantly — 93% bracket accuracy vs. 48% (Claude) and 38% (GPT-4o). Rule-based scoring was better calibrated than LLM judgment alone for this task.
- • LLMs systematically over-rate decks — they struggle to recognize how casual true Bracket 1–2 decks really are.
- • 6 reference decks were identified as miscalibrated — cases where the engine and both LLMs agreed the expected bracket was wrong.
The headline finding holds: an independent evaluation with no access to our ground truth agreed with the engine more than it disagreed, and pointed to real ground truth errors when it didn’t.
Cross-validation conducted in early 2026 against an earlier engine version. The engine has since been recalibrated to a more conservative baseline (see above). A rerun against the current engine is planned.
Why measurement is hard
Self-assessment bias
Players systematically underestimate their decks — especially near the B3/B4 boundary. ScryCheck measures the deck as constructed, not as the player perceives it.
The ground truth problem
Any validation system needs ground truth to measure against. Ours comes from database-mined estimates, community assessments, and expert review — and the quality of the ground truth limits the quality of the accuracy claim.
Card quality vs. table impact
ScryCheck measures what’s in the deck. It can’t account for piloting skill, local meta, or political dynamics.
The bracket boundary problem
Brackets are discrete (1–5) but power level is continuous. Any deck near a boundary will feel “wrong” to someone. The confidence interval and bracket boundary note in your results communicate this uncertainty explicitly.
Continuous improvement
ScryCheck’s scoring engine is validated automatically on every update. The reference deck suite is reviewed quarterly, and user feedback is aggregated to detect systematic biases.
If you think your deck is significantly miscalibrated, use the feedback button after re-analyzing — it captures your expected bracket and routes directly to our calibration review queue.
FAQ
Why is my deck rated lower than expected?
There are two common reasons:
- • Card quality vs. strategic reliability. ScryCheck separately measures how powerful your cards are and how reliably your strategy can win. A high-power shell with a theme-first win condition (coin flips, tribal, chaos, voltron) will have its power level capped by the strategy ceiling — shown on your analysis page. Adding a second deterministic combo line, for example, raises that ceiling.
- • Piloting and meta factors. ScryCheck measures what’s in the deck. It can’t account for piloting skill, political dynamics, or local meta adjustments. The ± margin shows our confidence range.
What is the Strategy Ceiling indicator?
The Strategy Ceiling appears on decks scoring PL 8+ and shows the structural cap on your power level based on how the deck is built to win, not just how powerful the individual cards are. It starts at a neutral base and earns headroom through four signals:
- • Combo redundancy — multiple independent, deterministic paths to win
- • Commander engine — the commander is part of a compact, deterministic combo
- • Commander independence — wins exist that don’t require the commander on the battlefield
- • Infrastructure density — a high share of generically competitive cards (fast mana, tutors, interaction)
A ceiling of 10.0 means the deck has no structural ceiling — all signals fired. A ceiling of 8.5 means the deck runs a powerful shell but the strategy itself doesn’t reach Bracket 5 territory.
A deck that earns headroom above Bracket 4 has one further check to clear: the cEDH metagame test described in “Why isn’t my deck Bracket 5?” below. When that test holds a deck down, the tooltip shows the ceiling it earned alongside the one it was held at.
Why isn’t my deck Bracket 5?
Because Bracket 5 isn’t simply “the top of the power scale.” WotC’s bracket chart already describes Bracket 4 as “lethal, consistent, and fast, designed to take people down as fast as possible” — a Bracket 4 deck is allowed to be maximally fast. The stated difference at Bracket 5 is that the deck is “meticulously designed to battle in the cEDH metagame… often built using existing cEDH knowledge, tools, and/or decklists.”
So the question isn’t how fast or how powerful your deck is — it’s which card pool it draws from. A deck that wins on turn three with redundant deterministic combos can still be Bracket 4 if it isn’t built out of the competitive card pool.
ScryCheck measures that directly rather than guessing:
- • The competitive card pool is derived weekly from actual tournament results (edhtop16), so it moves as the metagame moves.
- • A card only counts if it’s distinctive of competitive play. Sol Ring is in ~45% of tournament decks but ~84% of casual decks, so it carries no signal and is excluded. Force of Will and Tainted Pact are far more likely in a competitive list, so they do.
- • Your deck is scored only against staples your color identity permits — a mono-green deck is never docked for not playing blue counterspells.
- • Staple lands are deliberately excluded, so an expensive manabase neither earns nor costs you Bracket 5.
When this is what’s holding your deck at Bracket 4, the bracket card on your analysis says so and shows your score against the tournament baseline; the Strategy Ceiling tooltip breaks down how many competitive staples you run out of those available in your colors. It isn’t a judgment that the deck is weak — Bracket 4 explicitly includes the fastest, most lethal decks that aren’t built for the cEDH metagame.
What does “Confidence” mean?
Confidence reflects model certainty, not just card recognition. It combines coverage with signals like archetype clarity, how many cards were categorized by baseline heuristics only, and deck completeness.
What does the ± margin mean?
The ± margin is the confidence interval around your power level. It widens when uncertainty signals are present — for example, boundary proximity, weak archetype confidence, or many unrecognized cards — and narrows when coverage and rating quality are strong.
Why is my mana base score low or negative?
The mana base score reflects land quality relative to format expectations — not mana acceleration (that's Speed). A tapland-heavy mana base scores low and can cap your maximum power level rather than producing a negative score directly.
How are decks stored?
When you analyze a deck, results are saved with a unique hash. Decks with identical card lists share the same URL, making it easy to share your analysis. You can re-analyze at any time to get updated scores.
Why doesn’t my commander show synergy bonuses?
Commander synergy detection covers 100+ hand-curated profiles plus 700+ commanders with synergy signals auto-generated from a corpus of analyzed decks. If your commander isn't covered yet, the deck is still analyzed normally — you just won't see commander-specific bonuses. Coverage grows as more decks are analyzed.
My deck has no Game Changers. Why is it assigned Bracket 3?
ScryCheck combines WotC’s card markers (Game Changers, combos, mass land denial) with your overall power level to assign a bracket. A deck with no Game Changers can still reach Bracket 3 if its card quality, synergies, and consistency score high enough on the power level scale.
When this happens, your analysis page notes the distinction explicitly:
“This deck’s power level pushes it into Bracket 3, though under a marker-only reading of WotC’s bracket guidance it may fit Bracket 2.”
Both readings are valid — it depends on how your pod uses brackets. If your group goes strictly by WotC markers, you’re a B2 deck and that’s a fair reading. ScryCheck shows you both so you can have that conversation honestly rather than having the bracket feel arbitrary.
Why do I need to sign in for AI insights?
AI insights are tied to your account rather than anonymous analyses, since each one is a live, on-demand model call with a per-deck cooldown. Signing in is free and unlocks mana health, role coverage, synergy analysis, piloting tips, and recommended card swaps for your decks.
MTG card interactions are genuinely complex — LLMs can get specific card details wrong, so treat suggestions as a starting point for your own evaluation rather than a final verdict.