If you know that each option (ABCD) is roughly 25% likely to occur, is it a better bet to guess all Cs (guaranteeing around 25%, probably a fail in most tests), guess 25% of each option randomly (so if there were 20 questions, 5 for each option), or do something in between?
Or what if you know that some option are more likely than others, let’s say C has a higher chance of occurring than ABD. Should you go for all Cs or should you mix in some ABD?
And if you have already completed 50% (or any other amount) of the test already with presumably correct answers, is it better to guess all Cs for the ones you missed or to distribute them so that there are 25% ABCD each (or a different ratio if you know some are more likely on average)
edit: I’m not planning on random guessing every option on any future tests, and certainly not counting every option to ensure a spread out distribution (that would take up lots of time!). It’s just a silly probability question
It depends on the answer key.
Guessing ‘C’ all the way down and expecting 25% assumes that the distribution of answers is roughly equivalent across the whole test. This is an extra variable you can’t account for without knowing the answers.
Whereas, if you pick a random answer for each question, you will answer 25% correctly, assuming an infinite number of questions. However this strategy has error bars that grow as the number of questions gets smaller.(ie: you are far more likely to end up with more or fewer than 25% correct in practice)
The best strategy for any given multiple choice question that you don’t know the answer, is to eliminate options that are definitely incorrect, leading to a much higher chance of success with a random guess.
Or just study the content more, as breadsmasher suggested.
Back when I was in school, we had an honors class focused on test taking. One of the exercises we did was taking multiple choice tests where we were unlikely to know or understand the content being tested.
The idea was to read through the test and quickly scan for information that could be used to answer other questions on the test.
After doing that and filling in the answers that were provided by the material itself, you looked for predictable patterns in the answers; this differed depending on the test. For an answer key created by hand where the test writer doesn’t know about “pick c” C is the best answer. But these days, test writing software usually “randomizes” the answer order, attempting to make each answer a truly random letter.
But “random” is rarely truly random, and the human eye can often spot distributions and patterns once enough answers have been filled in.
And this brings us to the key bit: usually with multiple choice there will be one answer that’s obviously wrong, one correct answer, one answer that illustrates you understand the technique but not the general concept, and one that reflects a common error.
If you can identify which answer is likely to be each of those, they’ll also have a distribution and a pattern, and these can help you eliminate incorrect answers and highlight the correct solution.
Using these analytic techniques, it’s usually possible to score at least 80%, although the success rate generally improves with the length of the test and the amount of time provided to do the analysis.
We had it drilled into us that, after we’d filled out the answers based on our knowledge and skill, to ALWAYS leave time to go back and run these techniques against the test to ensure that we hadn’t made mistakes AND to gain a deeper understanding of the particular test writer to apply against their future tests.
I find it’s best to know the answers beforehand.
Big, if true.
Treating this a s a pure statistics problem where the “correct” answers are selected randomly:
If all the options are equally likely, then there is no guessing strategy that will change the expected result.
If one option is more likely than the others, then you should choose that one every time. Any deviation will reduce your score, on average.
or you could study
That’s the best suggestion! Probably not a good idea to use the power of randomness for any actual tests, otherwise it’s random :D
As someone building quizzes as part of my job, my strategy is usually to assign the correct answer to C or D by default, but then i would randomly change to the others options in an attempt to homogenize the answers. I definitely like using A more than B (not sure why)
Having said all that… be careful if you’re taking a quiz on Canvas or Blackboard, both offer the option to randomize the multiple choice answer list so the order is different for each person.
its really about eliminating definately wrong answers. if you eliminate one now the guess is 33%. eliminate 2 and its a straight up 50/50 and of course if you eliminate 3 then you have the answer.
Now I’m imagining a test with the instruction to mark the correct answer if present. So it’s possible that none of the choices are correct and you’re expected to leave it blank in that case.
Thinking about it, it’s equivalent to just adding a “none of the above” option to every question, but somehow it feels much worse.
yeah but it now brings it to five options. questions that always have all and none of the above are kinda easier with 4 but 5 or more it slides worse.
We had negative marking for incorrect answers to account for the assumption that you start w From a baseline of 25%. So, optimal strategy was to not answer questions if you didn’t know the answer with a reasonable level of confidence.
I actually went to a competition where you got 0 marks for an incorrect answer (so no negative marking), BUT you got 0.25 marks for leaving the question blank, so you had to be strategic on whether you guess (e.g. if you’re confident it’s 50:50) or not answer. I imagine the boundaries for the awards are adjusted to account the 25% being the “baseline”
If there’s all of the above, choose that.
Ah, the tests we used to have to take in military trainings sometimes…
There were a lot of them, and a lot of ‘all of the above’ answers. And in my entire career, there were only two instances when ‘all of the above’ wasn’t the correct answer.
Oh, and another big tip: the longest answer is usually the correct one. If in doubt, pick the longer one.
When I was in a class specifically for competing in Academic Decathlons, we were told if we didn’t know the answer to just pick C because not answering deducted more points than being wrong, and statistically, C was going to be the correct answer more often than not on a 4 choice question.
Just don’t do this on a true/false question.
No, no. It still works if you always choose “si.”
The one pattern that shows up frequently is that repeat answers are less likely than statistics would suggest. For example, if the answer to question #2 is C, then the answer to question #3 is much less likely to be C.
This is because people designing tests want the answer patterns to look more ‘random’, so they overcompensate trying to avoid streaks of the same answer over and over again. But in a truly random set, streaks would happen from time to time.
Less of an issue in software-generated tests, though, because the software often handles the randomization and doesn’t have such biases. But for hand-made tests, it’s a fairly reliable trend.
