Features
Two products in one overlay.
DeckLogic treats constructed play and Limited as co-equal, first-class tracks — with accuracy and low resource usage gating everything.
The overlay
An overlay that stays out of the way.
Press F8 to switch between a fully transparent overlay snapped
over Arena and a standard desktop window. Hover-aware click-through means HUD panels catch
your mouse only while you're on them — everywhere else, clicks fall straight through to the
game. Battlefield-anchored highlights are painted onto the real board, calibrated once
with F11. Compact mode fits panels to their content at any window width.
During a match — descriptive tracking only
Know where the game actually stands.
DeckLogic shows you the game as it is, not what to do about it. All prescriptive analysis moves to post-game review.
Win expectancy, honestly computed
A 24-feature XGBoost model (shrunk to 3MB) with no outcome-leaking inputs, scored live on every game state — and a calibration backtest run over your own match history keeps it honest.
Draw & land odds
Probability of seeing specific cards within a chosen horizon (next draw, 2, 3, or 5 draws), computed from your actual remaining library. The same calculation feeds every surface so they never disagree.
Lethal indicators
Prominent banners for lethal and crack-back danger — present as a scalar, without a recommended line.
Deck & hand tracker
Cards you've seen with slanted segments showing remaining copies; a cost strip of your remaining spells that lights the buckets your lands can pay for. Click a category to filter the card list.
Opponent tracker
Archetype prediction with confidence, every revealed card (across all zones, persisting through games 2–3), and untapped mana pips.
Aggregate match luck
How your land draws compare to what your list predicted, clustered by game with honest coverage labels. Below 200 exactly checkable draws across 10 games, no number is shown.
Session record
Between games, your current run's W/L, streak, duration, and net rank movement — a scoreboard, not a nag. Absent during a match.
Win expectancy
Post-game review
Coaching after the match concludes.
After the match ends, DeckLogic opens the door to analysis. Your decisions are graded on your own calibrated scale, the Film Room replays your game cinematically, and practice drills are generated from your recorded mistakes.
Film Room
Cinematic replay playback — cards travel between zones, attackers lunge, life totals tick, paced by what actually happened. Watch from your seat or the opponent's. Coach moments surface as notes in the timeline.
Graded decisions
Your actual plays compared against credible alternatives and graded — "best line," "reasonable," "costly," "clear miss" — derived from your own history, not arbitrary constants.
Match thesis
A plain-language summary: how many decisions were graded, how many cost you, which concepts the misses cluster around.
"Am I unlucky?"
Your land residual summed across recorded games — how many lands you drew against what your decks predicted — with honest coverage and a 95% range.
Mana base review
What actually happened to your lands — screw, flood, color requirements — grouped by mana base, with findings named per color.
Deck strength score
A calibrated score for your 40-card deck (count-weighted 17Lands win rate, shrunk toward format average) with what decks in that range actually did across 2.1M+ decks.
Practice & drills
Spaced-repetition drills from your own games.
Concepts recurring across your reviewed matches drive drill selection. A weakness that recurs but nothing can train is named as exactly that.
Set support
Practice follows one Arena draft set at a time. The current set is actively maintained. Earlier sets stay playable but are frozen, and preview sets may change before release.
- DSK · Frozen — This earlier set stays playable, but its cards do not receive ongoing updates.
- FDN · Frozen — This earlier set stays playable, but its cards do not receive ongoing updates.
- HOB · Active — This is the current Arena draft set, and its card support is maintained now.
- PRACTICE · Preview — This set is available early and may change before its Arena release.
Daily Puzzle
A position from a concluded match — commit to a line, then see the ranking and what you actually played. On dry days, a shipped pack fills in.
Mulligan Gym
Unlimited graded opening hands against opponents from your recorded matches. Keep or mulligan, graded on the same calibrated scale as the review. Mulligan and it tells you what to bottom.
Race Math
A generated board with two life totals — do you win this race? Committed positions are checked against exhaustive block search.
Hand-Reading Gym
Commit a read of a hidden hand, then find out what they were actually holding. Cards they never revealed are counted as "not knowable," never wrong.
Limited & draft
A draft co-pilot that earns the seat.
Drafts, sealed pools, and practice runs are first-class events with automatic lifecycle routing — drafting, deck building, recap.
Pick advisor
Every card in the pack scored 0–100 from 17Lands telemetry, with badges overlaid on the real pack in Arena. Recent picks keep their ratings, and the app explains why.
Observed synergy
Card-pair affinity and archetype-stratified win-lift, mined offline from real drafts — synergy advice grounded in data, not vibes.
Deck builder
A visual drag-and-drop build board with Arena-style CMC columns, a sortable stat table, one-click suggested swaps, and Monte Carlo auto-lands that minimize screw and flood. Builds persist across restarts — even mid-draft log rotations.
Consistency lab
A 10,000-game Monte Carlo of your working deck — curve-out and cast-on-curve odds, screw/flood/mulligan rates, and a 16 · 17 · 18 lands selector with one-click best fit — plus a live deck-health readout as you edit.
Post-draft recap
Pick-accuracy grade, per-pack grades, money picks and table gifts, splash analysis, synergy highlights, and the table read — what the other seats actually cut.
Boarding plan
After the draft: cards outside your 40 worth boarding, with stated reasons for each one — what it does, what it cannot do, and which matchups it speaks to.
Gaps tool (EVAL-4)
Where the format's pick order and results disagree — cards taken early that perform late (or vice versa), compared against same-rarity cards.
Practice mode
Full offline draft simulation against realistic packs collated from public 17Lands data — practice P1P1s and whole drafts without spending an Arena entry.
Limited sim engine
A playable Magic rules engine.
DeckLogic can now play Magic — a real rules engine, kept deliberately small and honest about what it models. Unmodeled cards are listed by name, never approximated.
Rules engine
Zones, stack, London mulligan, keyword-aware combat, triggers in turn order, state-based actions, auto-tapper. Games are byte-for-byte reproducible from their seed.
Practice decks
Two curated practice opponents (mono-red burn, mono-white aggro) built entirely from cards the engine understands. Full legal games verified — 200-game gate, zero rule violations.
Trophy-pool mining
Measures which cards would make real decks playable, the colour spread of mined decks, and what a named batch of cards would close.
Honest about gaps
Coverage is measured against the set's actual draft environment, not a subset. Every unmodeled card is listed by name with the reason.
Mastery & progression
Twelve badges. Each with a receipt.
Craft (a decision with evidence), Discipline (a habit held over a stretch), Saga (a story worth keeping). Every badge opens the film at the frame that earned it. Your existing archive is read in the background; absent badges state what they couldn't evaluate.
AI announcer
Your games, called live.
Context-aware esports commentary, composed on your machine. Point it at a local Ollama or LM Studio server for fully generative, unscripted lines — or let the built-in semantic compiler write them when no model is running. Kokoro neural TTS speaks the call, with a one-click voice download inside Settings. Everything local; the announcer works offline.
Accounts & economy
The whole account picture.
Games, decks, and stats are attributed to the Arena account that played them — even across in-client account switches, and attribution never guesses: a write that can't be verified is held and reconciled, not stamped onto the wrong account. Gems, gold, wildcards, vault progress, quests, and ranked progress are tracked per account, and an optional read-only 17Lands link (stored per account) harmonizes your decks and runs.
Trust & safety
Built to be trusted at a tournament table.
Fair play by design
During a live match, DeckLogic describes the game but never chooses your play. Prescriptive analysis lives exclusively in post-game review. The boundary is enumerated field by field and enforced at the Rust dispatch layer. [The Live Boundary](https://unrealities-media.github.io/DeckLogic/boundary/) states exactly what is shown and what is refused.
Tournament Safe Mode
When an event's policy is unknown, DeckLogic starts restrictive: tactical advice, hidden-information reads, win expectancy, draft help, and the announcer are blocked at the engine level — not just hidden — with a visible HUD state, a policy audit history, and explicit confirmation before live assistance turns on.
Guided setup & health center
A resumable first-run wizard verifies Arena, the log, Detailed Logs, card data, models, datasets, hotkeys, and calibration — and the same truthful diagnostics stay in Settings afterward, including honest degraded-mode explanations.
Signed automatic updates
In-app update checks install cryptographically signed releases, with the current version, channel, and recovery guidance always visible in Settings.
Sanitized diagnostics
One click builds a support report with your screen name, tokens, paths, and card names redacted — it lists every redaction rule it applied, and your log file is never attached.
Opt-in means off
The announcer's LLM and speech engines load only when you enable them, and unload when you don't. The streamer/OBS server is off by default and token-protected when on. Nothing runs — or leaves — without your say-so.
Accessible by default
Keyboard and assistive-technology support across the core journeys: focus management, semantic tabs, live announcements, visible focus, and reduced motion.