ManaCurve beta
A plain-language log of the player-facing changes going into ManaCurve. This is the stuff you might actually notice while building decks, reading reports, or importing a collection.
Plan scores and coherence badges no longer mistake healthy sacrifice, token, graveyard, artifact, or other routed decks for generic value or politics shells. Construct-aware fallback scoring now stays with generic routes, while decks with a dedicated plan template are judged by that plan.
Generated modified-combat and graveyard decks now reserve room for the pieces those strategies need before flexible cards. Counter decks look for counter enablers, scaling, payoffs, combat threats, and finishers, while graveyard decks look for fuel, discard outlets, recursion, reanimation targets, and loop support.
Plan scores now recognize more slow Commander win patterns. Grindy value engines, pillowfort and politics shells, and pure control decks get credit for the roles they actually need, such as incremental advantage, protection, table redirection, answer density, and compact late wins.
Generated equipment, Voltron, and typal decks now reserve room for the pieces those strategies need before flexible cards. Suit-up decks look for equipment or aura density, attach support, protection, combat payoffs, and a real close, while kindred decks look for tribe bodies, lords, value engines, protection, and finishers.
Several commanders whose deck plans were leaning on stale theme guesses now route from the actual clauses on their cards. Gishath builds as Dinosaur typal, Zhulodok stays in colorless cascade, and Atraxa, Grand Unifier now follows her enters-the-battlefield card-selection plan instead of a card-type reminder route.
Deck detail and generator audits now read plan scores from the same stabilized metric everywhere. Strategies without a dedicated archetype-aware scorer no longer get the old double role-soup penalty, while control, combo, stax, and prison decks still use their construct-aware paths when those paths apply.
Deck generation no longer marks the build direction as fallback-only just because the precon synergy corpus has no exact commander match. When EDHREC commander data is available and mapped to the card catalog, the result stays focused on the real deck constraints instead of showing a misleading enrichment fallback warning.
Derevi, Empyrial Tactician now builds through the stax shell instead of a generic combo route. Stax and prison plan scores also give credit for lock pieces, parity breakers, protection, grind engines, and compact wins, so a real stax deck no longer looks incoherent just because its best cards are restrictive permanents and table-control tools.
Ellyn Harbreeze, Busybody now follows her actual token-fueled card-selection text instead of being forced into an artifact shell or a generic goodstuff pile. Draw-matters and tempo commanders can reserve room for repeatable card selection, draw payoffs, cheap board presence, tempo interaction, and a real close.
Commanders that clearly copy or double tokens, triggers, spells, or matching permanents now give those payload cards a small explainable valuation lift. Adrix-style token doublers, Isshin and Teysa trigger doublers, Veyran spell-trigger doublers, and Miirym-style permanent copy commanders can now explain when a card is being valued for doubled output.
Generated artifact and lifegain decks now reserve room for the pieces those strategies need before filling with flexible cards. Artifact shells look for density, mana, payoffs, recursion, and finishers, while lifegain shells look for repeatable life gain, life-to-card engines, payoffs, drain finishes, and protection.
When a collection or budget cap cannot honestly reach the bracket you asked for, ManaCurve now names the deck-wide budget as a blocker and keeps B5 guidance focused on power shape and stronger support instead of asking for an impossible pile of game changers.
You said: Collection generation gave me a bracket below the one I requested after using the budget constraint
The beta feedback popup now keeps its send controls reachable on small screens after attaching a screenshot, and the deck wizard gives lower-confidence build directions a larger mobile tap target so you can choose them when the recommendation is not the plan you want.
You said: On mobile, attaching a screenshot pushed the send button away, and non-recommended deck directions could not be selected.
Commanders that clearly make a category of spells cheaper now score those matching cards by the cost the commander enables, not just printed mana value. Lorehold-style miracle decks value expensive instants and sorceries as cheap miracle payoffs, while creature, spell, and typed-spell cost reducers can lift the big cards their discounts are meant to unlock.
Miracle and topdeck commanders now add their signature engine on top of the normal spellslinger shell. Lorehold-style decks look for topdeck control like Sensei's Divining Top and Scroll Rack plus native miracle payoffs when those cards are available, while collection-only decks explain when the stronger miracle package needs cards you do not own yet.
Deck generation now checks the finished list for complete known combos that were picked as separate good cards. Non-combo decks remove unrelated combo pieces and refill those slots when the deck still reaches your requested bracket, but keep the combo with an explanation when it is the only way your pool can reach that bracket.
K'rrik, Son of Yawgmoth now gets the full Bolas's Citadel, Sensei's Divining Top, and Aetherflux Reservoir line with protection support instead of stopping at generic tutors, mana, and draw. Combo deck plan scores rise only when the actual combo core is present.
The Decks list no longer shows the internal 'assigned' and 'slots' counts, which mixed up distinct card rows with physical copies and could make a complete deck look short. The list now just shows your decks without that confusing detail.
Deck generation now reads commander abilities more like a rules card: costs, triggers, payoffs, and who they affect are separated before choosing a deck plan. Discarding as a cost no longer turns a commander into a discard deck, blitz and reanimation text point toward graveyard plans, cheat-into-play commanders lean combo instead of generic typal, and Vren-style removal commanders can build as control.
Commanders with card-selection text no longer get treated as mill or discard commanders just because they discard then draw. Miracle and instant-or-sorcery commanders stay on their spells plan, and higher-power non-combo decks are less likely to have an unrelated generic combo package forced in as the win plan.
You said: Lorehold is a Boros miracle-spellslinger deck, not mill/discard plus an off-commander combo
Deck detail now uses the same archetype-aware plan score as the generator audit for control and combo decks. Answer-heavy control shells and assembled combo decks no longer look incoherent just because their key cards are interaction, tutors, mana, and protection, while combo shells without a real core still stay marked as weak.
When you ask for a higher bracket from your collection and ManaCurve has to generate a lower one, the result now shows the requested bracket next to the generated bracket, explains whether collection depth, game-changer availability, role gaps, lands, or a max-card-price cap held it back, and gives direct options like switching to Hybrid or adding the missing higher-power support.
You said: Wanted a bracket 4 and only got a bracket 3 due to my card price restriction
Deck detail now counts every required copy when it shows total cards and section totals, including collapsed basic lands like 12x Mountain or 6x Plains. A complete Commander deck should read as complete instead of looking short just because basics share one row.
You said: I built me a deck with 83 cards! fix this!
Commanders that truly build around flickering creatures now generate with repeatable blink engines, high-value enters-the-battlefield creatures, trigger doublers, instant-speed protection, and ETB loops or payloads that can actually close. Control, typal, and token-copy commanders that only look nearby stay on their better-fitting plan.
Commanders that truly win by emptying opposing libraries now generate with mill doublers, repeat mill engines, burst-mill finishers, payoff bridges, and protection for the closing turn. Self-mill commanders that use the graveyard for value stay on their graveyard or reanimator plan instead.
Tax and prison commanders now build with the lock pieces, parity breakers, protected card flow, and compact inevitability they need to close a game. Commanders that only look adjacent to stax stay on their better-fitting plan instead of being pulled into the prison shell.
Commanders that are truly built around lifegain now generate with repeatable life engines, payoffs that convert life into cards or counters, drain finishes, and protection for the payoff turn. Commanders that only gain life incidentally or play sacrifice-drain aristocrats stay on their normal plan.
Commanders that want a planeswalker-heavy 'superfriends' deck — like Atraxa — now generate with a coherent plan: planeswalker density, proliferate engines, pillowfort protection, and clear ways to close, instead of a generic good-stuff pile. Decks that only look like superfriends but don't actually want it are left on their normal plan.
Deck generation now builds around a recognizable plan for each archetype — filling the roles that strategy actually needs (for example tokens get generators plus payoffs; control gets answers plus a draw engine plus a compact way to win) before adding flexible value — so a generated deck reads like it has a gameplan rather than a pile of individually-good cards. Commanders without a strong archetype match keep the previous balanced build.
Deck detail can now show a small advisory panel for cards you already own that may be easy to underrate for the commander. The panel only appears when the feature flag is enabled, keeps the suggestions occasional, and explains each pick in plain language instead of claiming you need to rebuild the deck.
ManaCurve now has a calibration harness for the fast pod simulator that compares its deck-strength ranking against independent Forge ground truth, keeps held-out decks separate from tuning, and fails closed when coverage or agreement is not strong enough. The interaction and combo models remain off by default until that trust report is robust enough for human review.
ManaCurve now recognizes additional deterministic combo packages such as Deadeye with Peregrine Drake, Mindcrank with Duskmantle Guildmage, Reveillark with Saffi and Altar, and Grave Titan with Nim Deathmantle. These are true infinite or kill lines, not loose synergy piles, so higher-power combo checks can trust them as real finishes.
ManaCurve can now run a broad set of fair, bracket-labeled Commander decks through the independent Forge rules engine and save win-rate rankings plus play traces. This gives future simulator and generator tuning a stronger outside reference instead of relying only on our own scoring signals.
ManaCurve now uses a generated combo package library mined from Moxfield co-occurrence and Commander Spellbook wins. Higher-powered generated decks and the pod simulator can recognize more compact finishes such as Godo with Helm, Malcolm with Glint-Horn, Bloodchief with Mindcrank, and other repeatedly evidenced lines, while staple piles stay muted unless they contain an actual combo.
Deck generation now learns land counts, ramp, draw, interaction, creature density, and mana curve from real Moxfield and Archidekt lists for the same commander, bracket, and budget. When the evidence is strong, a commander with a creature-heavy plan and a control commander no longer start from the same generic composition targets; when the evidence is thin, ManaCurve falls back to the old bracket defaults.
ManaCurve now looks at Moxfield, Archidekt, and EDHREC land data for the commander before choosing utility lands. When there is enough evidence, cards like Cavern of Souls, Nykthos, The World Tree, Field of the Dead, and commander-specific fixing can move up because real lists for that commander actually play them. Thin data still falls back to the previous manabase rules, and budget plus max-card-price limits still apply.
The deck generator's pool options now say plainly what they build from. All Cards (Catalog) builds from every Commander-legal card — not just the ones you own — while My Collection uses only cards you have. When you're in Collection mode, a hint points you to the all-cards option, so you're never left wondering why a build only used your own cards.
You said: unable to build me a deck from all of the magic cards in the world not just my collection
Building in Collection pool mode no longer dead-ends when the commander isn't in your collection. Your commander is the deck's identity — it lives in the command zone, not among the 99 — so ManaCurve now builds the rest of the deck from your available cards and simply notes that you don't own the commander yet, instead of blocking the build at the final step.
You said: wont generate my deck tried 6 times
Catalog generation now uses a compact multi-source model built from Moxfield, Archidekt, and EDHREC data. When a commander has enough evidence, ManaCurve steers picks toward cards that real decks agree on, with EDHREC synergy helping commander-specific cards beat generic staples. The consensus pull now works in mid-power decks without crossing bracket caps, and tight-budget lists reserve enough room for real nonland cards and cheap fixing instead of spending themselves into piles of basics.
Higher-powered generation now looks for complete combo and engine packages instead of sprinkling in isolated combo pieces. ManaCurve combines Commander Spellbook, real Moxfield co-occurrence patterns, and card-text relationships, then adds package tutors and enablers only when the needed core pieces fit the commander, bracket, colors, and budget.
Deck generation now gives land slots more context from the commander and chosen shell. Low-budget lists make more room for cheap utility lands like Command Tower, Reliquary Tower, Access Tunnel, Castle Embereth, artifact lands, and landfall support while still respecting total budget and max card price caps. Higher-bracket lists can reach for premium fixing and utility such as Ancient Tomb, Boseiju, Otawara, original duals, and other expensive lands when the budget allows.
Catalog and Hybrid generation now price cards from their cheapest paper printing instead of a mostly empty oracle field. Max card price caps catch expensive reserve-list cards and dual lands, and uncapped generation has a soft affordability check so pricey staples and lands are less likely to crowd out budget-friendly picks.
Catalog generation now uses the universal scorer weights from the latest archetype search, so default decks should feel markedly more synergistic and archetype-coherent across go-wide, spellslinger, aristocrats, recursion, superfriends, ninjutsu, artifacts, enchantress, and voltron shells. Landfall tuning is still coming separately.
When Catalog or Hybrid generation has a maximum card price, cards without a confirmed market price no longer count as free. ManaCurve now looks for prices on other printings first, then excludes still-unknown cards under the cap and says when that happened. Deck price estimates also call out any included nonbasic cards whose prices are unknown.
Deck generation now labels the finished list from its actual speed, consistency, interaction, resilience, and synergy signals. Game changers still matter for bracket caps, but the generator no longer treats missing game changers as a reason a tuned deck cannot be B4. When a collection build lands below the requested bracket, the report says what the deck plays as and points at the power signals to improve.
Deck detail and public share pages now include a Why this card? view for visible deck cards. The explanation is computed from deck evidence like role signals, commander-plan fit, strategy graph links, combo detection, and card text, with cautious confidence when the evidence is thin. Public shares hide collection and budget-only reasoning.
Commander intent now needs clearer draw-timing evidence before treating a commander as a spellslinger plan. Landfall commanders that draw from extra land drops stay focused on lands, so generated deck reasons and upcoming card rationale copy do not describe the plan as spell chaining just because the commander draws a card.
Deck generation now shares the same commander-plan vocabulary as the deck strategy graph. Commanders whose text points at a clear plan, like Vampire typal tokens, landfall, graveyard recursion, Treasure exile-casting, or proliferate counters, can steer the generated shell and card picks locally even when external synergy lookups are unavailable. Pick reasons also show that intent directly, so the result can explain why on-plan cards made the list.
The Opening Hand percentage on deck detail was showing 0% or 100% for many healthy decks because the simulation drew the same hand every time instead of sampling. It now deals your deck repeatedly with real variation, so a normal 36-land build reads a realistic keepable rate in the high 60s to 70s instead of a contradictory 0%.
You said: Opening Hand 0% while early play is 100% — that's internally contradictory.
If the collection value history machinery hits an internal error right after an import or quick-add saves your cards, the save now completes and reports success instead of returning a server error. Value history catches up on its own; your cards were never at risk, but the old error made it look like the import failed.
When an import row has no collector number or exact ID, ManaCurve now matches the regular booster printing of that set instead of an arbitrary variant. Previously a name-plus-set row could silently match a showcase or serialized printing and inflate your collection value with a treatment you may not own.
The collection header now reads like a plain fraction — 50 of 2,372 cards shown — instead of stacking a plus sign on top of the total, and both numbers get thousands separators.
Hybrid and Catalog generation now show budget controls up front. You can set a total deck budget and a maximum price for any single card ManaCurve would need you to acquire, and upgrade plans respect the same cap while explaining when expensive staples or combo pieces were skipped.
Collection imports now keep a durable list of rows that did not become copies, group them by reason, and offer a skipped-rows CSV with your original import columns plus the explanation. Deckbox rows with a blank Count now call that out directly instead of disappearing from the result summary.
You said: Imported 4586/5258 rows • 5824 copies. Wish there was more feedback as to why rows were skipped.
The collection value compiler now handles thousands of owned cards and hundreds of thousands of market-history rows without overflowing, so the Value tab can build real multi-day history from the exact printings you own. Empty chart copy also distinguishes a collection with cards from a truly empty collection.
You said: The core promise ('graph renders immediately with available market history') is not yet true anywhere. Make it true.
Deck views now treat basic lands as unlimited basics instead of separate owned printings. If your Swamps, Islands, or snow basics come from several sets, the editor, deck detail, grid, and public share views show one row with the total quantity.
The Stats dashboard now falls back to your live collection value if the new precomputed trend table is still catching up during a production rollout, instead of showing a server error.
You said: Verify /api/health, /collection Value tab, and /stats after deploy — live-verify before declaring the phase done.
Beta feedback now has a full lifecycle. You can see your own submissions and where they stand, implemented items can point back to the changelog entry that shipped them, and admins have faster tools for keeping the loop current.
The collection value chart is no longer a Pro-only surface. ManaCurve now reads prebuilt daily trend points, appends today's live value, and describes the chart as the value of your current holdings over time, so large collections get a faster, more honest market history view.
You said: Full collection history becomes a core/free feature. Pro keeps alerts, automation, and deeper analytics.
Collection value history now falls back gracefully while the new trend table is still rolling out. Instead of failing when that table has not appeared yet, ManaCurve keeps using the available snapshot data until the prebuilt history is ready.
ManaCurve now uses Commander Spellbook more proactively when the bracket and request call for it. Higher-powered generation can look for compact on-plan win lines, while upgrade plans can surface a Win Path Unlock when your current deck is one owned card away from a confirmed combo. Lower-power requests still stay restrained unless you ask for combo directly.
You said: Spellbook's ability to find Potential Combos could be a really great thing to consider for our deckbuilder / generator and upgrade planner.
The upgrade planner now reads the generator's newer role language before suggesting swaps. It is more careful about preserving scarce ramp, draw, interaction, protection, and real win-condition slots, so an upgrade path should not trade away the one card that actually closes the game just to add another value engine.
You said: Those changes should also apply to the logic within the upgrade planner.
The Collection Leaders list now ranks by the highest individual copy value instead of letting a pile of low-value duplicates outrank truly expensive cards. Bulk stacks still count toward your total collection value, but they no longer crowd out your best singles.
You said: I just want highest value SINGLE CARDS, not the total value of all of my bulk commons.
The Stats dashboard now uses bounded daily snapshot rows for its value pulse instead of doing heavy value-history compilation while the page loads, so large collections can open the dashboard without tripping the known trend-builder crash.
You said: The /stats crash needs fixture-first reproduction, and the recursive getCollectionValueTrend must be understood.
ManaCurve now has regression coverage proving owned collection value is based on the exact printing and finish you imported, not just the card name. If an owned printing is missing a market price, it stays unpriced instead of borrowing a broad card-level estimate, so totals are more honest while price history work continues.
You said: We need to ensure this plan is able to recognize WHICH version of a card to seek value for.
ManaCurve was already finding concrete combo lines; now the combo panel proves what it checked. Detected Combos counts confirmed lines separately from loose synergy packages, empty results say how many Commander Spellbook and curated packages were checked, setup-heavy infinite lines stay labeled as infinite, and incomplete enrichment is called out instead of pretending nothing was found.
You said: Decks that plausibly contain combos show No combos detected and No complete combo packages.
The card-art collage now lives only on the sign-in page; app pages sit on a clean dark canvas so panels and text stay readable. Every panel has one small info button that explains all of its numbers at once instead of a question mark hovering next to each stat. Delete Deck moved into the Actions tab where destructive things belong, the generated-deck details collapsed into a single line, and the collection's delete-everything controls are tucked behind a disclosure.
You said: Formatting is not all standardized and clean, text is not always readable, and the chips are obvious.
Clicking the ManaCurve logo from What's new now sends signed-in players back to their collection instead of the public homepage. The homepage also recognizes active sessions, so returning there no longer looks like you have to log in again.
You said: Clicking what's new while logged in, then clicking the main ManaCurve logo takes us back to the main page and makes me log in again.
The beta feedback popup can now be moved around on desktop, and the page behind it stays visible and usable while you write. Deck Strategy card popups also keep the section title in view, so it is easier to remember which part of the strategy you opened.
You said: Allow me to move the feedback popup around the screen to where I want it. Allow the background to be undimmed and clickable so I can see it better.
Plain card names across deck overview, analysis, public shared decks, and generator reports now open the existing card view. When ManaCurve mentions a card, you can inspect it without hunting for the image somewhere else on the page.
You said: Whenever there is a card named, it should be clickable to show the card.
ManaCurve is stricter about what counts as a real closer. Generic support cards and loose theme density do less work now; a deck needs cards that can actually end the game, create a credible clock, drain the table, overrun, evade, or assemble a near-deterministic combo before the Win Plan tile calls it ready.
Vren, the Relentless and similar commanders are now understood through their actual play pattern: removal pressure, death/exile replacement effects, and token payoff. That means the generator is less likely to shrug at a commander just because the card does not use the obvious deck-building words.
Deck Strategy no longer calls a list Combo when the final analysis found no concrete combos. Strategy labels, graph paths, and win-condition copy now lean on the same evidence as the combo section, so the overview and analysis stop disagreeing with each other.
You said: Labeled Combo with 0 detected combos.
The scored and derived numbers now have small info buttons on deck detail, shared deck, and generator result surfaces. Power, bracket, foundation, opening hand, win plan, radar axes, role health, and similar metrics explain what they measure and how to read them without making the page heavier.
Collection pages now load from the cards you actually own instead of walking broad catalog data first, and decks/collections show loading states instead of briefly pretending they are empty. Large CSV imports also move in bigger, safer chunks with resumable session support, so a big paper collection is less likely to stall halfway through.
