David@VeritasPrep wrote:
Now what if you take several minutes and narrow the question down to two choices?
I think this is the conclusion we all agree on: if you're guessing, guess *quickly*. That said, if by spending 15 seconds you can eliminate 2 or 3 wrong answer choices, that is surely time well spent. If you'd need to spend 2 minutes to do that, that's a huge waste of time, since 2 minutes spent elsewhere on the test will get you a right answer a lot of the time, not 1/3 or 1/2 of a right answer.
Just to give a couple of examples of what I mean, opening OG12 to a random page, if I look at Q214 in the PS section, this question asks for the probability three different things all happen. If you know the basics of probability, you can rule out the first two answer choices almost immediately, because they are far too high (and you can be pretty certain C is wrong as well). Or in question 212, the radius of the circle is clearly smaller than k, but not by much - answer B almost has to be right, and you certainly wouldn't guess anything but B or C if you do a 10-second estimate. In each case you can eliminate 2 or 3 answer choices within 15 seconds, but you need to understand something about the question. Now, it doesn't happen all that often that you can eliminate so many answer choices so quickly, but when it does, it's helpful to take advantage of the fact.
David@VeritasPrep wrote:
The research Brian mentioned above is the result of some mathematical probabilities. Basically, if you take the experimental questions into account as well as the fact that you might guess correctly, a random guess has more than a 35% chance of either not counting, or you guessed correctly. So if you spend 15 seconds and randomly guess there is a 35% chance the question is not held against you (either you guessed right or it was experimental).
I think the rest of this post might only be of interest to David and other GMAT specialists here:
I can see where you can get a figure around 35% (1/4 of the time the question doesn't count, and (3/4)(1/5) = 3/20 of the time the question counts and you guess correctly), but I don't think, in practice, it's the right number to use for most test takers. People tend to guess at questions which are hard for them (or at least ones which appear hard at first glance). The difficulty level of experimental questions covers the full range from easy to hard - unlike the questions which count, experimental questions do not adapt to the test taker's level. If a test taker is near the top level of ability (a 49/51 test taker, say), then the only questions she will consider guessing at are those that are at the highest difficulty level. Those are almost never going to be experimental questions, because most experimental questions will be well below the 49/51 level.
I just did a quick thought experiment to estimate how much time is worth spending to eliminate wrong answers. My assumptions might be bad, but I looked at things this way: eliminating two wrong answers is worth about 1/3 - 1/5 = 0.13 correct answers. Since test takers normally get 2/3 of the questions they spend time on right, 2 minutes is worth 2/3 - 1/5 = 0.47 correct answers (notice that, according to this methodology, an answer which is 100% certain to be correct is worth a bit more than 3 minutes of time, which seems close to right, but may be too high - certainly a certain correct answer should be worth more than 2 minutes). So 0.13 correct answers is worth about 33 seconds of time. That is, if you can eliminate two wrong answers in 30 seconds, it's probably time well spent. If you're only going to eliminate one wrong answer, it isn't worth spending more than about 11 seconds to do so - in that case, guessing randomly is probably the best idea. And if you might get to a 50-50 situation, it's worth spending some time to do so.
These are things you should learn using your own experiences. That is the best way to apply an educated guess.
