Return to Blog Homepage

Guessing on the LSAT: Logical Reasoning Edition

BPPcolin-chances3

One thing that you should absolutely positively never ever do is leave questions unbubbled on the LSAT. There’s no penalty for wrong answers, so wild guessing can do nothing but improve your score. (I actually had someone who would always leave question 13 blank in all sections because she thought it was bad luck. I pointed out that having a 13 in your actual score isn’t really a good thing, either. She later decided against law school.)

But for the ones you don’t get to, which answer choice should you pick? Even LSAT prep companies don’t really seem to agree. According to Blueprint Competitor A: “It just doesn’t matter; you have the same chance of guessing right with each question.” But Blueprint Competitor B says: “Do not guess randomly!” With an exclamation mark! Most people assume that the distribution of correct ACs is more or less random, so the guesses you make on the ones you don’t get to can be random, too. And this, for all intents and purposes, is mostly true. Nevertheless, I decided to “run some numbers,” using Excel for the first time since high school. Today: logical reasoning.

Pretty-Much Acceptable Method: Historical Score Distributions

If answer choices were indeed totally random, one would expect that if you were to total all the times A, B, C, D, and E were correct since the inception of the LSAT, the sums would all be nearly identical. After all, over the course of 3,176 LR questions, things would probably have equaled out, more or less.

(Drum roll)

It actually looks like they’re not totally random. While not hugely different, each answer choice doesn’t by any means get a completely fair share.

Picture 1

It’s not a huge difference, but it looks like A and E are less common, while D shows up the most as the correct answer choice. So if you were to wildly guess on the entire LR section, (D) would probably be your best bet. But that’s assuming that answer choice distributions have remained relatively constant, something I’m not willing to blindly accept. If you look at LSATs from the 90s, they’re rather different, so acting on the hypothesis that maybe the AC distros were also at variance, I ran the numbers. Here’s what it looks like when you cut out the 90s:

Picture 2

In this century, (A) has become even more of a loser, but (B) and (D) are now tying for first. So now it looks like you could pick either (B) or (D) if you were doing wild guessing. But then I thought I would look at even more recent data – just the last five years.

Picture 3

(D) is once again the clear winner, but not by much. And now (E) is the loser, but again not by much. So it looks like the LSAT might be moving to more of a truly random model.

But none of this information is really that helpful unless you’re guessing on the entire section, something you would never do. If this was your strategy, you might want to explore other avenues. What actually happens for most people is that they guess only on the last few questions, if on anything. Questions do tend to get harder as you go, so you should definitely start at the beginning and work your way back, without jumping around. But even so, often you just can’t get to the last few, so you guess on them. So I looked at the distributions for just the latter parts LR. Below is the frequency of answer choices again, but just for the last five questions per section. To keep things representative, I left out the 90s again, looking at roughly 5 and 10 year periods:

Picture 8

Picture 5

Our sample sizes are getting smaller now, so these percentages are perhaps less telling. But if you’re having to guess on the last 5, (A) is historically by far the lowest-chance guess. It looks like that trend might be lessening, but it’s still there even today. This is probably due to the fact that LSAC knows that you’ll be rushed on the last few, and doesn’t want to give it to you right away, because LSAC’s classy like that. Looking at either the short or long time period, (C) is the best option. Here’s the same data, but just with the last three questions:

Picture 6

Picture 7

Here, for the ten-year span, B and C are tied, but C pulls ahead when just looking at the last five years.

So what does all this mean? Honestly, not a whole hell of a lot. If you run out of time in LR and have to blindly guess on some or all of the last five, your best bet is probably C, and probably not A. But it’s far from any sort of guarantee. All of this is based on the assumption that what has happened in the past will continue to happen in the future. That’s a classic category mistake (something that you should know for flaw questions, by the way). On the September test, none of the last three on either section were C. But if you had guessed A, you would have got 3 of the 6.

Pretty Much Horsecrap Method, At Least on LR: Score Distributions Per Section

Another common strategy, having nothing to do with the above distributions, is to guess based on your previous answers. As the thinking goes, if you’ve answered fewer of one the answer choices so far, you should guess that one for the last few that you don’t get to. For instance, if you complete three games and notice that you haven’t chosen (B) that often, you would guess (B) for the remaining game. This strategy would make a great LSAT question, because it entails a number of necessary assumptions.

First of all, you’re assuming that you correctly answered the vast majority of the ones you completed. Let’s say Rodion is taking the LSAT, and he just finished question 22 in LR when he realizes that he has 45 seconds left. He looks back, sees that he answered only three questions with (A), so he guesses (A) on 23, 24, and 25, and with a twinkle in his eye believes that he probably got one of them. Well, for this to really help, it would have to be the case that everything he did so far was almost all correct. But it could very well be the case that he got a number of questions wrong whose actual correct answer choices were (A). So while it looks to Rodion like there have been 3 correct (A)s so far, there may have been four, or five, or six (or more).

Rodion would also have to be also assuming that the distributions of A-E are roughly equal per section. In his defense, this is generally true. On most LR sections, each answer choice is correct 4-6 times. But certainly not always. On the second LR section of December 2008, (B) and (C) were correct only three times each, while (E) was correct a whopping 7 times. So maybe Rodion was correct with all his answer choices, and it was simply the case that there were just three correct (A)s on his particular test.

So this method is far from reliable. But the biggest reason that you shouldn’t try doing this, at least not on LR, is that it’s a waste of valuable time. With those 45 seconds, Rodion very well may have been able to answer one more question. Sometimes those last monster questions just click, so it’s better to go for something that you might be able to figure out, rather than for something that’s little better than a crapshoot.

Conclusion: Guessing (C) is an OK idea, but you’re a bad person for not studying more.

So based on the historical distros, guessing (C) on the ones you don’t get to is probably a decent idea, but it’s not going to make or break you. The best use of your time is answering the actual questions, and the surest (and really only) way to a high score is through lots and lots of studying. If guessing is going to be making a huge difference in your score, your score probably ain’t gonna be that high. Keep up the work and you’ll get faster, hopefully making the whole guessing thing moot.

Submit a Comment

Submit a Comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.