Changelog

Ban history

"Has it ever been banned?" now knows every ban and unban since 1994, separately from "is it banned right now?". Hybrid mana symbols count as both of their colours.

Printings, sets and years

Questions like "was it printed in the last year?", "has it been banned in a format?", "has it been printed at rare?" and set names now answer correctly, and persist creatures like Kitchen Finks know they come back.

Colour identity and protection

"Does it have a colorless color identity?" now gets an answer, and "does it give protection?" counts protection an Equipment or Aura gives a creature. Thanks to Umaru.

Sharper readings

"Does it care about X?" now means the card reacts to X, not just mentions it. "Can it counter any spell?" means every spell. "Pay life" covers life a card makes you lose for its effect.

Clearer words, clearer turns

"Does the word 'creature' appear?" now counts "noncreature", "lifelink" counts for "life", and "does it trigger each turn?" means every turn, opponents' turns included. Thanks to Umaru and Larcondos.

More fixes from the sweep

Double-faced cards now answer "Sometimes" for abilities on their other face, delayed triggers count as triggers, and modal abilities are no longer mistaken for activated ones.

Mystic Remora and friends

"Does it trigger on creature spells?" now checks which spells actually trigger it, so Mystic Remora says No and Beast Whisperer says Yes. Cards with upkeep costs now "sometimes" make you sacrifice them, and "does it gain flying?" counts flying the card has or can get.

Fixes from our first full sweep

Lands that enter tapped or with a choice no longer claim to have a triggered ability. Lightning Bolt and Chaos Warp now answer "only creatures?" correctly, and dozens of other famous cards were fixed.

Lightning Bolt, Mystic Snake and "your hand" fixes

Lightning Bolt no longer claims it only targets creatures or hits multiple targets. Mystic Snake counters spells. Tutors, draw spells and regrowth all put cards into your hand, and "does it give you protection?" now means protection for you, the player.

The One Ring, Jitte and flashback fixes

The One Ring now admits it gives you protection. Umezawa's Jitte counts as removing counters for effects. Faithless Looting can be cast from your graveyard, and Snapcaster Mage gives flashback. Smothering Tithe no longer claims to trigger on your own draws, and Cyclonic Rift can affect multiple permanents.

Better answers for manlands, double-faced cards and creature-stealing cards

Manlands like Mutavault and Celestial Colonnade are now "Sometimes" a creature, and "Sometimes" have the abilities they gain when they animate. A double-faced card is whatever its front face is, and a color only its back face has is "Sometimes". Stealing a creature for good, like Control Magic, now counts as removal. "Does it buff power and toughness?" now means both go up.

Fixes for questions with a catch

Royal Assassin no longer claims it can destroy an untapped creature. Solitude now admits it can be cast for free, and that it can't exile itself. Chaos Warp counts as getting rid of things. Vampiric Tutor's card goes on top of your library, not to your hand. "Can it deal damage?" is now Yes for creatures, and cards that draw or deal damage repeatedly over a game (Beast Whisperer, Phyrexian Arena) answer "Sometimes" to "more than one?".

A reminder about land colors

By Magic's rules, lands are colorless — basics included. The game now reminds you the first time you ask about color, and the Sphinx says so in its color questions.

Hints now come straight from the card

Hints used to be written by an AI, and now and then one got a detail wrong, like a rarity or how many sets a card was in. Every hint is now a plain fact read from the card's own data (its cost, type, power and toughness, keywords, first printing, artist or flavor text), and you won't get one your questions already answered.

Cards that only sometimes have a keyword now say so

Dragon's Rage Channeler said no to "does it have flying?" even though it flies with delirium. It and 364 other cards that give themselves a keyword only under a condition (Bloodghast's haste, Dragonlord Ojutai's hexproof while untapped, threshold and hellbent cards) now answer "Sometimes".

See how your friend started

Open a friend's Daily link and you now see how many questions they needed, plus their first few questions and answers, before you ask your own. Never their guesses, and never the card.

The Sphinx keeps the card on screen after she finds it

When you tell her "Yes, that's it!", the card now stays up next to her result instead of shrinking to a name.

Galactus and six other exilers now admit it

Galactus, Devourer of Worlds said no to "does it exile a permanent?" even though that is what it does on entering. It and six other cards whose exile effect had been missed (Parallax Wave, Parallax Tide, Ugin, the Spirit Dragon, Cruel Alliance, Inevitable Defeat, Vault 75: Middle School) now say yes.

Cards that bounce their own creature now say so

Whitemane Lion, Familiar's Ruse, Urban Retreat and about a dozen others that return a creature you control to its owner's hand said no to "does it return a creature to hand?". They now say yes. Cards that return a creature from the graveyard or library still say no, which is right.

Send the Daily to a friend

The share button on your Daily result now reads "Send to a friend", and the text ends with a dare: "Same card for everyone today. Can you beat my 9?" with your own number. Lose, and it asks whether they can get it.

Bounce spells count as bouncing a creature

Boomerang, Cyclonic Rift, Upheaval, Kor Skyfisher and about 190 other cards that return a permanent to its owner's hand said no to "does it bounce a creature?". They now say yes.

Your Daily streak, where you can see it

The Daily now shows your streak before you start, the longest streak anyone is on right now, and the fewest questions anyone has needed today. Shared results spell the streak out ("5-day streak") instead of a flame and a number. Also fixed: your streak read zero while a Daily was still in progress.

The rest of Reality Fracture is in

93 more cards from Reality Fracture and its Commander decks, including Jace, Multiverse Architect and The Ur-Sphinx. The Sphinx can guess them now, and they are in the card pool for Random.

Hobbit, Marvel and Star Trek cards now count as Universes Beyond

About 1,000 cards from The Hobbit, Marvel Super Heroes and Star Trek were not marked as Universes Beyond, so the Sphinx could talk herself out of the right card when you answered Yes to that question. They are marked now, and they follow the Universes Beyond setting in Random.

Tell us when an answer was wrong, in one tap

After a Daily or Random game, your questions are now listed on the result screen with a "Wrong?" button beside each answer. Tap it if the game told you something false about the card. That list used to be hidden behind a small link, so almost nobody found the button. Every tap gets looked at.

Paying life counts as losing life

Sylvan Library, Toxic Deluge, Necropotence, the shocklands and about 280 other cards that ask you to pay life said no to "does it cost life?". They now say yes.

"When this enters" abilities the game was missing

Whitemane Lion, Mystic Snake, Meteorite and 35 other cards with a "when this enters" ability said no to "does it have an enters-the-battlefield ability?". They now say yes.

"Enters or attacks" is an attack trigger

Sun Titan, Inferno Titan, Primeval Titan and about 180 other cards with an ability that triggers when they enter or attack said no to "does it trigger when it attacks?". They now say yes.

"Does it cost 2 blue?" now means two blue mana

Asking whether a card costs 2 blue was being read as two generic mana plus one blue, so Counterspell said no on its own Daily. The same went for any number and colour, like 3 black. It now means exactly that: two blue mana and nothing else.

Reported by Larcondos

Searching your library is not reanimation

The game answered yes to "does it reanimate creatures?" for Wild Pair, which only searches your library. The same mistake was on 74 cards that put something onto the battlefield from the library, including Eldritch Evolution, Defense of the Heart and Birthing Ritual. All of them now answer no.

Reported by madsam55

A dropped connection no longer ends your game

Right as the Sphinx was recording her guess, our database dropped the connection for a moment and the page showed a raw error instead of the result. The game now retries once when that happens, and if something still goes wrong you get a plain sentence, never a wall of technical text.

Reported by Animu36

Solitude does not make a token

The game answered yes to "does it create tokens?" for Solitude. It does not; a bad label on its Evoke line said otherwise. Fixed before it could run as a Daily.

"Was this fun?" is gone

The question after Random games asking whether it was fun has been removed. Very few people answered it and nearly every answer was yes, so it was not telling us anything. That spot now asks whether you knew the card, which does.

Did you know this card?

After a game you may now be asked whether you knew the card before today. It is one tap, and it is the most useful thing you can tell us: picking a Daily means guessing which cards people actually know, and until now we have been guessing from deckbuilding sites that only see Commander. On the Daily it sits under the card. In Random practice it shows up after you answer whether the game was fun. The Sphinx does not ask, since you picked the card.

Rarity questions now count every printing

Ask whether Karakas is uncommon and we used to say no, because the copy we store is its mythic reprint from Ultimate Masters, even though the original from Legends in 1994 is uncommon. Rarity belongs to a printing, not to a card, and plenty of famous cards have been printed at several. Now the answer looks at every paper printing. If a card has always been one rarity, you get a plain yes or no. If it has been printed at more than one, you get sometimes. Lightning Bolt has been common, uncommon, rare and mythic, so asking about any of the four gets sometimes. Time Walk is simply rare, because its only odd printing was online-only.

Cards that make tokens now say so

If you asked whether History of Benalia creates tokens, we said no, even though its first two chapters each make a Knight. Monastery Mentor, Garruk Wildspeaker and about thirty other cards that plainly make tokens had the same problem. A second group had it with players: Tormod's Crypt exiles target player's graveyard and Rise of the Eldrazi has target player draw four cards, but asked whether they target a player, we said no. Seventy-five cards in all now answer these correctly. They were found by a checker that compares every answer we give against the card's own printed text, and the fixes are now checked and reapplied automatically every night.

Time Walk is a rare, and we kept saying it wasn't

If you asked whether Time Walk was common, uncommon, rare or mythic, the answer was no all four times. There was no rarity you could name that got a yes, so the question was not just wrong, it was a dead end you could spend several turns walking down. The cause is that we store one printing of each card, and the Time Walk we store is the Vintage Masters one, which sits on a sheet Scryfall labels "bonus" instead of any normal rarity. Seventy-two cards are affected, including the whole Power Nine and the purple-bordered Time Spiral cards. Black Lotus and Ancestral Recall had both already been Dailies this way. Rarity questions about those cards now give you the turn back instead of answering, and the vetting that picks the Daily refuses to schedule one of them again. Doing this properly means storing what a card was printed at across all of its printings, which is a bigger job and is still to come.

Reported by Storm, Monocolored Enthusiast

A word printed only in reminder text is no longer a confident no

If you asked whether a card's rules text says "whenever", and the only "whenever" on the card sat inside the italic reminder text in brackets, we answered no. That is defensible, because reminder text restates a keyword rather than being the card's own rules. It is also a word you can point at on the card, so being told no read as the game lying to you. Monastery Swiftspear is the clearest case: prowess spells out "Whenever you cast a noncreature spell" right there in brackets. The game now declines the question and gives you the turn back instead of answering it either way. Just under a thousand cards are affected.

"Does it let you draw cards?"

The Sphinx used to ask "Does it cause you to draw cards?". That reads as though it is asking whether the card makes you draw right this second, which is not what it meant. Same question, clearer words.

Play Again keeps your card pool settings

If you turned Universes Beyond off, narrowed the product types, or set an earliest year, and then pressed Play Again, the next game quietly went back to the full pool. The toggles on the home screen still showed your choices, so there was no way to notice. Your settings now carry across every game in a run, not just the first one.

Exerting is not an activated ability

Glorybringer was listed as having an activated ability because of exert, and a player wrote in to say that is not what exert is. They are right. An activated ability is a cost and an effect with a colon between them, like "{T}: Draw a card". Exerting is a choice you make as the creature attacks, and the thing that happens afterwards is a trigger. Twenty-seven cards were tagged wrongly and now answer correctly, including every exert creature from Amonkhet.

"Does it trigger when creatures enter?" stopped counting the card itself

A player asked whether Vanquisher's Banner triggers when a creature enters, and we said yes. It does not. It triggers when you CAST a creature spell, which is a different moment, and the only thing entering in its text is the Banner itself. We were matching any card whose rules text mentioned entering at all, so every creature with a plain "when this enters" ability counted too. That is 644 cards that were saying yes and should have said no. Cards that genuinely watch for OTHER creatures arriving, like Soul Warden and Impact Tremors, are unaffected.

"Proliferate" does not contain the word "life"

Ask whether a word appears in a card's rules text and we were matching letters, not words. So a creature with lifelink counted as saying "life". So did anything that said "proliferate", because the letters l-i-f-e are sitting in the middle of it. That was 609 cards answering yes to a question they should have answered no to, and 892 wrong answers across thirteen questions once you count "nontoken" matching "token" and Chandra matching "hand". It went the other way too: "libraries" did not count as "library", and "sacrificing" did not count as "sacrifice". Now the match is on whole words, with the endings you would expect, so drawn, destroyed, exiling and sacrificing all still count.

Odd or even mana value

Asking whether a card's mana value is odd or even used to get you a shrug most of the time. It is a simple question and we did not have a way to hold it, so the answer depended on how you phrased it. Now it is one question with one answer, and "divisible by 3" works too. Two things worth knowing, because they surprise people: a land has a mana value of zero, and zero counts as even. And X counts as zero, so Fireball is a one, which is odd.

Questions about power, asked of cards that have none

If you asked "is its power 2 or greater?" about a sorcery, we used to answer No. Ask the other half, "is its power 1 or less?", and we answered No to that too. Both cannot be true. A card with no power has no answer to a question about its power, so we now say we cannot answer it, and the question is free. We have also stopped using cards that you cannot look up: a few cards are hidden from Scryfall search, and finding out your card does not exist anywhere you can check it is a rotten way to end a game.

"+1/+1?" now means the size of the bonus, not a counter

Asking "+1/+1?" used to be read as "does it use +1/+1 counters?", so Crusade, a card whose entire text is "White creatures get +1/+1", answered No. Worse, "does it give creatures +2/+2?" ignored the amount completely and answered Yes to the same card. The engine now reads the actual size of a power/toughness bonus, so +1/+1 is Yes and +2/+2 is No, and a card that gives +1/-1 says No to both. Say the word "counter" and you still get the counters question you asked for.

The Sphinx now asks "Does it target something?"

She used to ask whether the word target appears anywhere in the text, and most people answered whether the card targets, which is a different question for about 300 cards (a creature that punishes spells that target it, for example). Now she asks what you were already answering: does the card itself pick a target.

The Daily share button works again

Since yesterday evening the Daily's share button did nothing, on every browser: the new result link was built with an encoding browsers do not have, and the click died silently. That is fixed. Sharing now also opens your phone's share sheet where there is one, copies where there is not, and if the browser blocks both it shows the full text to copy by hand.

Reported by Jubiiz

"Is it an anthem?" and "does it only pump white creatures?" now answer

Ask about Crusade whether it is an anthem, or whether it only affects white creatures, and the game used to hand the question back. It now reads who a bonus reaches: Crusade is an anthem and only pumps white creatures; Glorious Anthem pumps yours of any color; Goblin King only Goblins; a Giant Growth or a Rancor is not an anthem.

Share your Sphinx game

The Sphinx's result screen now has a share button. If she found your card, the share says how many questions it took her; if you stumped her, it says so and names the card. The Daily and the random game had this already; the Sphinx did not.

The Sphinx explains her color questions

Is a Forest green? By the rules, no: lands are colorless. Is a five-color card white? Yes. Enough players answered those the other way that the Sphinx now says so under each color question, so an honest answer and a rules answer are the same thing.

"Does it target?" no longer counts being targeted

Goldspan Dragon, Nadu, Phantasmal Image and about 300 other cards only mention "target" because something happens when they become the target of a spell. The game used to answer yes to "does it target?" for all of them. Now a card has to target something itself. Reported three times since April, including Grip of Chaos, which only scrambles other spells' targets.

Nicknames count: "is it E-Wit?" wins on Eternal Witness

The game now knows the names players actually use. Steve is Sakura-Tribe Elder, Bob is Dark Confidant, Goyf, Gary, Mom, Thoracle, and about a hundred more from the community list, plus initials like STE, JTMS and FoW. If a nickname fits several cards (Bolt, Edict, Moxes) the game says so and asks which one. A player asked about Steve yesterday and got a flat no; that was wrong.

"Can it take a land?" now understands what a discard can hit

Ask Thoughtseize "can it take a land?" and the game used to give the question back unanswered. It now reads what kind of card a discard is allowed to hit: Thoughtseize takes a nonland card, so a land is a no and a creature is a yes; Duress cannot take a creature; Mind Rot can take anything.

"Is it a ring?" now gets an honest answer

Ask about The One Ring "is it a ring?" and the game used to say no, because "Ring" is not a printed card type. Now a word that is not a card type gets a yes when the card's name says so, and an honest "I only know its printed types" otherwise, never a confident no. Same for crowns, helmets, swords and owls.

Links finally look like something

Paste a link to the game anywhere and you used to get a blank tile. Now you get the Sphinx. And a shared Daily result shows YOUR game in the preview, solved in how many, the claim, the narrowing bars, without ever showing the card, so it can't spoil anyone who hasn't played yet.

Something to argue about

Your Daily share now carries one measured claim under the header, "Only 38% solved it today", "First to solve it today", "Fewer questions than 80% of today's solvers", or, on a bad day, "62% didn't get it either". Every number is counted from real attempts, and none appears until enough people have played for it to mean something. Nobody argues with a score; people argue with a claim.

The reveal tells you something

When today's card is revealed you now see how everyone else did, how many solved it and the median number of questions, next to yours, and a line about the card itself: when it was first printed, how many times since, and who painted it. The first number is ours alone; nobody else can tell you that today's card took most people twelve questions and you eighteen. Take it up with them.

Less fine print under "target"

The Sphinx's question about the word "target" came with seven lines of explanation, enough to push Yes and No off the bottom of a phone screen. It is one line now. The one rule that matters survives: reminder text in italics doesn't count.

Sixty-four cards that draw now say so

Master of the Feast makes each opponent draw a card; Words of Wisdom draws you two; Blink of an Eye draws when kicked. All three, and 61 more, answered "no" to "does it draw cards?" because the draw was never written into their data. Found by chasing one wrong answer, fixed by a rule that reads the card's own text and ignores the text of things it grants or creates.

The land it fetches, not the card

Ask Solemn Simulacrum "does it put a land onto the battlefield tapped?" and the game said no, it heard "does IT enter tapped?" and answered about the Golem. Ramp cards get this question every time they are the Daily, so it now reads the fetch itself: Solemn, Sakura-Tribe Elder and Rampant Growth say yes; Wood Elves says no.

The Sphinx stops asking about keywords

"Does it have any keywords?" was her second question in almost every game, and one answer in ten disagreed with her card data. Surveil, Mill and Landfall look like keywords but aren't; "Enchant creature" is one but doesn't look it. We tried seven different definitions against thousands of real answers and none matched players better. So she has stopped asking. In replays of 2,000 real games she now finds the card two points more often and a question sooner.

The Sphinx stops saying "pip"

Her first question was "Does its mana cost have exactly 1 colored pip?", and if you have never heard a mana symbol called a pip, that reads as "is it one colour". A mono-green card with three green symbols answered yes, and the game was lost before it started. She now asks about colored mana symbols and, if you tap for help, counts a few costs out for you.

Mill means mill, and "you" means you

Three wrong answers, three causes. "Does it mill?" said yes to Brainstorm, to every surveil card and to cycling, anything that touched the top of a library counted. Now only the verb counts, and cards that mill a variable number (Jace, the Perfected Mind) finally say yes. "Does it trigger when you cast a spell?" said no to Guttersnipe because the DAMAGE goes to opponents, the check was reading the wrong half of the ability. And "does it make an opponent lose life?" was the same question as "does anyone lose life?", so Thoughtseize, which costs YOU two life, said yes. Now the game reads who the card names. And "does it have variable power?" said no to Tarmogoyf, the game was searching the rules text for a star instead of looking at the printed power. Two of these came from player reports on Daily cards; thank you.

The Sphinx gets to the point

Her front door used to make you read six things before you could start, two of them requests. Now it is the premise, the one rule that matters, "Don't know" is allowed and costs you nothing, and the button. What she will and won't guess is a tap away for the rare card it applies to, and she only mentions waking up when she actually is. Same Sphinx, a third of the words.

The Sphinx has four new questions for the cards she used to get stuck on

When she narrows the pool to a handful of cards and nothing in her question list can tell them apart, she wastes her last questions. We found which cards those are and added questions a player can actually answer about them: an activated ability that costs mana, a bonus to your OTHER creatures, discarding, and protecting your other permanents. On 400 replayed real games she gets stuck 10% less often. If one of these questions reads wrong to you on your card, tell us, that is how the last batch got tuned.

When we can't read your question reliably, we now say so

Some questions don't map to anything we can check, "does it produce more mana than it costs?", and about one time in five we used to answer them anyway, confidently, by picking the nearest thing we *could* check. Now the engine reads a new question twice, independently; if the two readings would give different answers, you get a refund ("I'm not sure how to read that one") instead of a guess. Ordinary questions are unaffected. Every refund of this kind is logged so the missing predicate gets built.

Counterspell can counter a creature spell again (and Ponder puts cards on top)

Ask "can it counter a creature spell?" and we had been answering the wrong question, "does it target a creature?", so Lightning Bolt said yes and Counterspell said no. We now read the card's own wording: Counterspell and Essence Scatter yes, Negate and Spell Pierce no. Ask "does it put cards on top of your library?" and Brainstorm said yes but Ponder, Sensei's Divining Top and Vampiric Tutor said no, because we were searching for one exact phrase. Fixed for the wordings cards actually use.

Random games share the way the Daily does

Sharing a Random game used to paste a bare line, "Got it in 8 Qs". It now shows what the Daily share shows: the outcome in words, how the pool shrank question by question, the narrowing chain, and your hint count, plus the challenge link so a friend can play the same card. One share format for both games from here on.

"Does it ramp?" now knows that fetching a land is ramp

Sakura-Tribe Elder, sacrifice it, go get a basic land, answered NO to "does it ramp?" on the day it was the Daily. The question was being read as mana rocks, extra land drops or cost reduction, and never as putting a land onto the battlefield, which is the most common kind of ramp there is. Rampant Growth, Cultivate and Wood Elves had the same problem. Fixed for every phrasing at once.

Hints are still free, and now they show

Hints don't count against your twenty questions and never will. But a solve with three hints is a different story from a solve with none, so the number is now part of the result: in the Daily share line (💡2 hints), on the reveal screen, and on the hint itself while you play. Random games already tried to say "(used a hint)" in the share text, it turned out to be counting the wrong list and always said nothing. Fixed.

Keyword abilities are now classified by the rulebook, not by guesswork

Ask whether Treasure Cruise has a triggered ability and we said YES, its Delve reminder text had been read as a trigger. Ask whether Quantum Riddler has an activated ability and we said YES. Warp is an alternative cost, not something you activate. Underneath, the same keyword was being sorted into static / triggered / activated up to six different ways across the card pool (Kicker alone had six). Every keyword ability is now classified by the Comprehensive Rules' own definition of that keyword: 2,283 corrections across 1,533 cards. One deliberate judgement: Morph, Disguise, Plot and Foretell count as static abilities, because the rules say turning face up or exiling from hand is a special action, not an activated ability, even though they have a cost. If that reads wrong to you, tell us.

"Is it a legendary creature?" no longer says yes to every legendary card

Jace, the Mind Sculptor and The Great Henge both answered YES to "is it a legendary creature?", the word "creature" was being dropped from the question, so any legendary card qualified. 720 cards were affected: legendary planeswalkers, artifacts, lands and enchantments. Fixed at the question-repair layer, so it holds regardless of how the question is phrased.

Cards that give +X/+X now count as making a creature bigger

Aspect of Hydra gives a creature +X/+X, where X is your devotion to green. Ask whether it makes a creature bigger and we said NO, about the only thing the card does. The problem was cards where the amount is a VARIABLE rather than a printed number: we recorded that the bonus existed and that its size depended on something, then read the missing size as though it were zero. Giant Growth, which says +3/+3, answered correctly the whole time, which is why this went unnoticed. 682 cards were affected. Cards that give a variable PENALTY (-X/-X) are unaffected, we read the sign off the card's own text, and where the text genuinely doesn't say, we now decline to answer rather than guess.

Reported by ragnarokorangar

"Does it search a library?" was saying yes to cards that do no such thing

If a card let you look at the top few cards of your library, or reveal one from your hand, we were calling that searching a library. It isn't, searching means going through the whole library for a specific card, the way Demonic Tutor or Cultivate do. Nearly 1,800 abilities were affected, which made this the least reliable question we had: players disagreed with our answer more than a quarter of the time. It now reads the card's own wording instead of guessing. We also learned to tell apart whose library you're searching. Bribery searches an OPPONENT'S library; Demonic Tutor searches your own. Before, both answered the same way to either question, so one of the two phrasings was always wrong. Ask about your library or theirs and you'll now get the right answer for each.

"Does it affect both players?" was wrong on a whole family of cards

Thalia, Guardian of Thraben makes noncreature spells cost {1} more, and that applies to HER controller too, which is the whole reason she is interesting. Ask us "does it hurt you too?" and we said no. The same was true of Wrath of God's twin Rout, of Kirtar's Wrath, of Sphere of Resistance, and of hundreds of others: whether an effect hits everyone or just your opponents was being judged card by card rather than read off the card, so two cards with the SAME printed sentence could get opposite answers. Wrath of God and Rout say exactly the same thing and we answered differently about them. That question is now answered from the card's text directly, so identical wording can no longer produce different answers. 941 abilities changed. This is the sort of wrong answer we care most about, because you can reason from it, told an effect is one-sided, you would correctly rule out every symmetric card, and be confidently wrong.

The Sphinx now tells you when she reopens her search, and shows you the cards she guessed

Two things about her side of the game were confusing, and it turns out they were the same problem. If she named a card and you told her no, that guess vanished, it never appeared in the list of questions below the board, even though it costs her one of her twenty the same way a question does. So the counter would read "question 16" while the list showed only twelve rows, and there was no way to see which cards she had already tried. Her guesses are now numbered in that list alongside the questions. The second thing looked much worse than it was: the number of cards still in play would sometimes jump UP. It was not the game cheating and nothing was being reset. When nothing in her collection fits every single answer you have given her, she does not give up, she starts allowing for one of them being wrong, and that naturally puts more cards back on the table. That is how she still finds your card after a slip. She just never said she was doing it, so it looked like the rules had changed mid-game. Now she says so, in the moment, and the line in the transcript where the number jumped tells you why it jumped. She does not blame you for it, either, when she and you disagree about a card, it is quite often her records that are wrong.

"Destroy target permanent" now counts as destroying artifacts, enchantments and lands

A player lost the Daily to this one and wrote in to explain exactly how: "When I asked if the card destroys artifacts or enchantments the answer was no ... the answer that was 'supposedly' right was Beast Within, which can definitely destroy artifacts and enchantments." They were right. Beast Within destroys target permanent, that is every artifact, every enchantment, every land, and the game was saying no to all three. Every card that destroys a permanent generically had the same hole. What makes this one worse than an ordinary wrong answer is that you can reason from it, and they did: "not artifacts or enchantments" pointed them confidently at the wrong card. Fixed, along with a second problem it uncovered, cards reading "destroy target NONartifact creature" were being counted as artifact removal.

"Is it 2 mana?" was answering No for most 2-mana cards

This is the most basic question in the game and it was wrong more often than it was right. Ask "is it 1 mana?" about Birds of Paradise and the game said No. Rabbit Battery, No. Mana Drain at "does it cost 2 mana?", No. Anything whose cost had a coloured symbol in it, which is most cards, got the wrong answer, while a card costing plain generic mana got the right one. A player caught it on Returned Phalanx and said so plainly: "the AI said it had a mana cost of 1, when its mana cost is 2." Every phrasing we could find in the logs is now covered, "2 mana", "mana cost 2?", "is it two mana?", "is it free to cast?", and there is a test for each one.

"Does it target a creature?" now works on cards with any adjective in front

Doom Blade destroys target nonblack creature. Asked whether it targets a creature, the game said No. So did Shizo ("target legendary creature"), Reclamation Sage ("target artifact or enchantment"), and roughly a quarter of every targeting card in the game. Plain "target creature" worked; the moment a card said nonblack, legendary, attacking, or offered a choice of two types, the answer flipped to No. Two players hit it on the same day's Daily, Archmage's Charm, "I asked if it targeted a spell, and it said no" , which is how it got found. It now reads the whole phrase. Around 1,300 wrong answers across the card pool went away with it.

"Does it interact with creatures?" is now its own question

People kept asking this and getting No about cards that obviously do interact with creatures. The problem was that we only understood the narrower question, whether a card is built AROUND creatures, the way a lord or an anthem is. Doom Blade is not a creatures-matter card, but it very much interacts with a creature. Both readings now exist separately, so "does it affect lands?", "does it do something with counters?" and "does it interact with artifacts?" get answered the way you meant them.

Cards that GIVE an ability, and cards that just have one

"Does it give something else flying?" was answering Yes for cards that simply have flying themselves. Bloodghast has haste; it doesn't hand haste to anything. Equipment and Auras, the cards that genuinely do give abilities away, were lumped in with every creature that happens to have the keyword, which made the question nearly useless for narrowing anything down. It now checks who actually receives the ability.

A new check that reads every card and flags where we contradict it

Most of the fixes on this page started with a player telling us an answer was wrong. That is a slow and slightly embarrassing way to find them. There is now a check that reads a card's printed text and flags any answer that contradicts it, if the card says "counter target spell", we are not allowed to answer No to whether it counters spells. It runs over every card in the pool, and it runs on each upcoming Daily before that puzzle is ever shown. It caught two already-scheduled Dailies with the same problem as Archmage's Charm, before anyone played them.

Beast Within is not a modal spell, fixed the same day it was the Daily

Asked "does it have modes?" about Beast Within, the game said Yes. It doesn't, "Destroy target permanent. Its controller creates a 3/3 green Beast creature token" gives you no choices at all. Two players reported it within hours of each other, both having lost the Daily to it, and both were exactly right. Fixed and live the same afternoon. The reports came through the feedback form, which doesn't ask for a name, so we can't credit you here the way we'd like to. If either of you is in the Discord, say so and we'll fix that.

Cards that make an opponent sacrifice now say so

Sporogenic Infection reads "target player sacrifices a creature", but the game answered No to sacrifice questions about it. The reason: we were only counting a sacrifice when it was a COST you pay ("Sacrifice a creature: draw a card"), and never when it was something the card MAKES someone do. That hid it on 360 cards, including Diabolic Edict, Cruel Ultimatum and Rankle's Prank, some of the most recognisable sacrifice cards in the game. Cards that merely care about sacrifices ("whenever you sacrifice a creature, draw a card") correctly still answer No, because they do not cause one.

Reported by dolgarien

Questions about a specific token, land or ability no longer get a vaguer answer

Asking "does it create Angel tokens?" about a card that makes Soldiers got a Yes, because the game quietly dropped the word "Angel" and answered the easier question "does it create any token at all?". The same thing made Scalding Tarn claim it could search up a Forest. Now the detail is either used properly. Scalding Tarn correctly answers No to Forest and Yes to Mountain, or the game admits it cannot answer, instead of confidently answering a question you did not ask.

Reported by dolgarien

"Is it protection?" no longer wastes your question

That exact phrasing hit an internal error every time and cost you the question, with a refund but no answer. It now works.

Equipment that makes a creature token no longer "becomes a creature"

Nettlecyst has Living Weapon, which creates a 0/0 Germ token and attaches the Equipment to it. The Equipment itself never becomes a creature, but the game said it did, so "does it become a creature at some point?" answered Yes. The same mistake affected every Living Weapon card, and also planeswalkers and other cards that make creature tokens: 323 cards in total now answer that question correctly. Cards that genuinely do animate, manlands like Mutavault, Vehicles, and auras like Ensoul Artifact, are unchanged.

Reported by dolgarien

Protection questions were wrong on cards that protect from two things

Sword of War and Peace reads "protection from red and from white". Asked "protection from white?", the game said No, because it was searching the card text for that exact phrase, and the phrase is split up by the word "and". Any card protecting from two things could give a wrong answer on the second one, including Akroma, Baneslayer Angel and every Sword. The game now reads protection properly instead of searching for words, so it also understands protection from a creature type, protection from everything, and cards that let you choose the colour, those now answer Sometimes, because they genuinely depend on the choice. 442 answers changed across 116 cards.

Reported by dolgarien

Questions about amounts now respect the number

Asking "does it draw 4 or more cards?" used to throw the number away and answer "does it draw cards at all?", so Ancestral Recall, which draws exactly three, said Yes. Numbers are now honoured for card draw, damage, life gain, milling, scry, tokens and discard. Spelled-out numbers work too.

Reported by dolgarien

"Does it do anything besides X?" now counts abilities

Asking whether a card has an effect other than one you already know about used to quietly drop the "other than" part, so a creature whose only ability was its enters-the-battlefield trigger would claim to have another one. It now counts the card's abilities instead.

Reported by dolgarien

A month of daily cards, with far more modern ones

The daily queue was about to run out, and every card lined up for the following week was first printed in 1993. It is now filled through early October with a deliberate mix, about a third from 2018 onward, and a spread of creatures, spells, artifacts and lands rather than a run of similar cards. Cards you have already played will not come back around while anyone remembers them.

The Sphinx shows you what it was still considering

When the Sphinx narrows your card down to five or fewer, it now tells you which cards those were. Previously you just got a number.

Word questions now say exactly what they mean

Thirteen questions that asked "does it mention X?" have been reworded to "does the word X appear anywhere in its rules text?" The old wording was ambiguous, people read it as asking about the card's meaning, and answered from that. Disagreement on those questions dropped from 18% to 3%.

+1/+1 counters are no longer confused with a +1/+1 bonus

Asking about "+1/+1 counters" matched any card whose text contained "+1/+1", so Glorious Anthem and Giant Growth, which give a bonus but no counters, answered Yes. The question now means actual counters.

Reported by dolgarien

Double-faced cards answer for the right face

Type questions on double-faced cards were answered from a merged version of both faces, so a card could claim to be two types it is never simultaneously. Split cards, adventures and transforming cards each follow their own rule now, and a card that genuinely differs between faces answers "sometimes".

Reported by dolgarien

"Does it have an enters-the-battlefield ability?" stopped over-answering

Cards that merely trigger when some OTHER creature enters were counted as having an ETB ability themselves. 1,358 cards were affected.

"Is it removal?" now explains what it counts

Removal is a term players disagree about, so the question keeps the term but now spells out exactly which effects count toward a Yes.

Equipment that changes toughness answers correctly

Skullclamp gives +1/-1, but toughness changes were not registering, so questions about modifying toughness answered No.