carding made somewhat tenable: a guide

[etotheipi]

I wrote up some thoughts on carding for a couple of curious people in the UKQB Discord the other day, and thought I might post it publicly. As of now it's tailored to those who are looking to get into seriously carding, but even if you already seriously card it might have something for you. I might add some more advanced tips next time I'm too tired to do anything actually productive.

-

When I talk about carding I am very specifically talking about using Anki to retain clues I've learned. The basic idea is this: there are a lot of things that I need to know for quizbowl that I wouldn't normally be too troubled about knowing. There are also a lot of things I know things about but the names of which I simply cannot seem to remember (perhaps because I have a bad memory for that sort of thing). Carding lets me: (a) memorize for quizbowl purposes facts I don't particularly care about with a minimum of effort, (b) be able to recall the names of things I know a lot of "context" about very easily, even when under a lot of pressure, and—interpolating the two—(c) build up a web of factual, quizbowl-ified knowledge around things I've read about.

Examples:
(a) I don't care about architecture, I will probably never care about architecture—I don't think it's "art" in any real sense. But quizbowl has decided that it is, so I can easily use carding to binary-associate some projects to their architects, memorize stock clues, etc. This way, I will still be able to get 10s (and sometimes 20s and 30s) on architecture bonuses, and will often be able to snag a tossup or few on the subject just from this purely "fake" knowledge.
(b) On the other hand, I do care a lot about European literature, and read a decent amount of it, but that doesn't mean I'm particularly good at remembering it. for example, I read about ten Thomas Bernhard books in the span of one or two months in 2022, and if you ask me about a detail of one of the books I am liable to forget which particular book it is in. More embarrassingly, I also might struggle to recall Thomas Bernhard's name in a game setting, just because that's a mental operation which really has no equivalent anywhere else in my life, and so it doesn't come naturally to me. Another example: I read a volumes of E.H. Carr's History of Soviet Russia once, which had a couple of chapters detailing the exploits of Vinnichenko, Petlyura, and Hrushevsky. But I'm liable to get these names mixed up—and ultimately in quizbowl it doesn't matter whether I answer a Hrushevsky bonus part with "Vinnichenko" or "Vladimir Putin": I'm getting the same amount of points. Carding helps me actually get questions on things I know rather than just "being in the right place."
(c) These days I'm pretty interested in Byzantine intellectual history—but not in Byzantine political or military history. This is perfectly fine in the context of what I'm trying to do with my knowledge "in the real world," but will not get me quizbowl points (since it is very hard to write a pure intellectual-history tossup, but easier to toss in some intellectual-history clues into a political-history tossup). Carding Byzantine political history, then, gives me a framework onto which I can map the intellectual history I know, but which question-writers will also often draw from. (For example, the actual ecclesiastical texts and doctrine of Saint Photios are a good deal more important than the details of his politickings at court, but one of the easiest ways to ask about that doctrine and those texts is to clue them early in a tossup about the then-emperor [Basil I]—you might not remember this, or might not be sure whether a certain event happened during Basil's reign or that of his son Leo the Wise, just from reading about Photios, but carding will make sure this doesn't happen.)

The above is the high-level "what do I card and why?" (There are of course other reasons to card: a handful of great thinkers built up Zettelkästen simply to have a physical retrieval system allowing them to build connections between bits of information they absorbed and thought important; and obviously carding for language-learning is its own thing. But the processes there are different than the processes for quizbowl [less involved, really], so I'll limit myself to the latter case.) The rest, anyhow, is praxis, i.e. making cards in a format that maximizes recall and quizbowl-applicability. Some of the best practices for this (to my knowledge—though I am one of the more successful carders out there) are:
(i) Use Anki. Quizlet and other competitors are all horrible; Anki is free, has amazing functionality and a very high number of add-ons, has a better algorithm, is a desktop app… there are really no alternatives. You're not carding seriously if you're not using Anki.
(ii) Make your own cards. Decks other people make are often not intended for quizbowl, or really bad, or both. (I know there was an anecdote going around a bit ago with someone finding a med school deck, one of the cards in which was "what does the serum iron test measure?") Also, it's a lot easier to remember cards that you yourself made, and a lot easier to apply them (you'll often be trying to lateral-think off your cards to get to the actual context of the question you hear it)—this is the phenomenon called "encoding specificity." It should be added that having a 45k card deck you made yourself and reviewed gradually as it was being built feels much more rewarding and much less intimidating than downloading 45k cards off the Internet and having to just dive into them and start memorizing a bunch of things you've never encountered before.
(iii) One card, one clue. Make your clues as minimal as possible without being non-uniquely-identifying. If you put two or more clues in a card ("in this novel, Wertheimer and the narrator quit music after realizing that they will never be as good as Glenn Gould at the piano"—do you see how there are two uniquely identifying clues here?) you will only connect the easiest one to the answerline, not both. This means that more-than-two-element clozes (i.e. fill-in-the-blanks) are almost always bad and should be avoided; honestly, just avoid clozes in general. Also, this does not mean that you should avoid cards with more than one answerline (for example: "these are the two products of the mevalonate pathway"), but these can be hard to memorize and apply well, so use them with caution.
(iv) Card "rhizomatically." Don't make fifty clues about random moments from the play Equus and have the answerline on all of them be Equus. (I did this once. It got me some points, admittedly.) Try to anticipate all the fun or interesting answerlines writers might choose and tailor your clues to them. For example, you'll want to connect happenings in a book to the book and to the author of the book, but maybe you can think of a common-link which might be clued from that happening, and anticipate it in a card. ("The narrator of Bernhard's The Loser quits this job after realizing he'll never be any good at it.") Often, you'll want to make your cards about answerlines that are too hard to even be hard parts of a bonus—this helps a lot with science especially. For example, you're probably never going to see a bonus part on the Appel reaction, but having a card associating the mechanism with the name firms up that connection and makes it easier to buzz on an Appel reaction clue in e.g. a triphenylphospine or alcohols tossup. That being said, you also need cards whose answerlines are "triphenylphosphine" and "alcohols" to more directly contend with those hypothetical questions. ("This compound, along with tetrahalomethane, is added to the reactant in the Appel reaction"; "this is the reactant in the Appel reaction.") If you think of it pictorially, each of your cards being a node of information pointing to another node of information, make sure all the arrows are not pointing to one node, but instead that there are many nodes with both arrows coming in and arrows going out.
(v) Card things that are the hardest to learn organically, for maximum coverage. For example, named reactions in chemistry, formulae in physics, character names in literature…
(vi) Make sure your cards are actually quizbowl cards, i.e. containing feasible quizbowl clues and a feasible (with the exception of what I said above) quizbowl answerline. You shouldn't make your answerlines elaborate descriptions or explanations; you shouldn't make your clues non-uniquely-identifying, general-knowledge- or trivia-like, or even just incorrect; for the most part, you're just going to want to think like a question-writer and anticipate what they're going to do.
(vii) Be judicious about where you get the information to make your cards. Past questions have factual errors sometimes and sometimes also make very glaring errors of interpretation (in philosophy and music this is particularly bad, but it happens in every subject), though they're a good starting-point to figure out what to research as well as what the stock clues are. Wikipedia is even worse, as it offers you a lot of very trivial information into which the important information is blended, and the priorities of its writers are often very inexplicable. (Also, misinformation.) AI has a misinformation problem (especially with queries into what is more obscure) and also, well, ick. I know how incongruent this is given that this is a guide about carding, but please don't sacrifice your morals for quizbowl points. Carding off books, papers, lecture-notes, etc. works surprisingly well, actually, and is what I strongly recommend. Plus, you actually become educated. It's incredible.
(viii) Actually review your cards.
(ix) During the process of reviewing, make sure you're simulating quizbowl conditions! If you're taking a lot of time to remember the answer, mark yourself as incorrect, since you would be timed out in a quizbowl context. Anki offers four buttons (again, hard, good, easy, which have the keyboard shortcuts 1, 2, 3, 4—the shortcut to reveal a card's answer is space, so you can go through a deck pretty quickly with only those five keys); my heuristic is: use 1 if I get it wrong or spend too much time, 2 if I have to wait a moment or two before getting the answer, 3 by default, and 4 if and only if the card is a good bit too easy for me.
(x) Read your cards out loud when you review them! Or whisper them or audiate them—it doesn't matter. This is important for two reasons. First, when you play quizbowl, you're not going to see the clues—you're going to hear them—so you should build a connection to the sound of the clues rather than the sight of them. Second, only looking at the clues makes you start associating answerlines with the shape of the text of the clue on the page (this actually happens), which is obviously very unhelpful.
(xi) Put all your cards in one deck, no matter the quizbowl category into which they fall. Don't make category-based subdecks: these are a lot more trouble than it's worth to organize, and also are sort of artificial (do you really always immediately know the category of a question you are hearing?).

[Good Goblin Housekeeping]

i think ur takes are normally insane (but fun 2 read) but pretty much everything you've said here is right (with a disclaimer of I only use flashcards for school and vocabulary because I am busy being married to multiple choice question exams)
(shoutout to the anking card "Serum {{c1::iron}} is a measure of iron in the blood")

(Kai also had pretty strong thoughts on making flashcards "atomic" as possible... let the oafs of the world keep memorizing the shapes of their flashcards instead dammit)

[Santa Claus]

link

Great guide. Only thing I’ll say is that bad cards (pre-made decks of reasonable quality, non-atomic, etc.) are way faster than making good cards and a lot of the time you can get away with it for a long time then make a new card - key is actually make the new card though.

[oriley]

Fantastic guide, this is great.

I feel like adding a proverbial two cents. Hopefully my ramblings are useful.

I think you got at this point a bit — I (as well as most people, I think?) like to format my clues in the form of a quizbowl clue/sentence. For example, I would write “This man was the target of the Conway Cabal” as a clue for “George Washington.” Doing that helps me sort of get in the mind of a question writer, and helps me remember the context in which the clue might be worded. If all I write is “conquest of Hejaz” and the back says “Saud,” although this is a useful binary, I might be slower in-game to figure out where a tossup that clues this is going.

Can we put point (viii) in larger font? Actually review your cards. Actually review your cards! I almost want to argue that a majority of quizbowlers who card via Anki rarely review. I know people who have more unreviewed/due cards than I have cards.

Point (xi): personally, I separate my cards into decks by top-line category (history, literature, fine arts…) because I think it controls the chaos a bit better. I do not agree with people who make a deck for every 1/1 subcategory, though, for the same reasons you stated.

Related notes:

To quote Jordan Brownstein, “I sometimes find that I can click through flashcards by just sorta recognizing the shape of the words, but don't know the clue ‘deep' enough to actually buzz on it in a game.” For me, I find that this can happen after many months of reviewing a card on something I studied a long time ago, and its a signal that I might have to go back and review notes or do some quick research to jog my memory. You’re probably making cards that you want to be useful to you a year from now, not just for two weeks. As such, you may want to do a deeper review of what you learned at some point.

Unrelatedly, I will sometimes make a card on something I run across that is completely useless for quizbowl points, but I want to remember it so I can write a question about it someday. I have…several pages in my Notes app just of “question ideas,” and the ones I have in Anki are easier to find, haha.

Finally, as a side-note: if you are a specialist, or not the first scorer on your team, and you develop a good system for how to learn and card things, one fun idea is to make a couple of random binaries out-of-category on clues you hear come up in power on a tossup. Be very careful about this if it’s science or sounds like a generic/not-obviously-unique history clue, but it’s a great way to swing a game, annoy other specialists, and so on.

[Stinkweed Imp]

Point (xi): personally, I separate my cards into decks by top-line category (history, literature, fine arts…) because I think it controls the chaos a bit better. I do not agree with people who make a deck for every 1/1 subcategory, though, for the same reasons you stated.

Separating cards into multiple decks is bad because it hurts retention and can mess with the spaced repetition algorithm. Sub-decks can work if you always study them together but in my experience this isn't any easier to organize than using tags.