Jose Ferreira wrote:The advice here has been good so far—the key to a true increase in your GMAT score is a true increase in your ability. Just for some background, in computer adaptive testing, the model works to determine the test-taker's true ability level, known as theta. From that theta there is a true score that would exactly represent that individual's performance. The standard error of measurement (SEM) for a particular assessment (without getting too technical here) gives an idea of how reliably that test gauges that theta over multiple administrations when all other factors are equal. It is purely a statistical measurement of the difference between the individual's true score and the hypothetical high or low score from actually taking the test. The GMAT has a SEM of 29 points, which means your score would be within 29 points of your true ability. The GMAT is a pretty well-tuned adaptive assessment.
Just to amplify what Jose is saying about the standard error of measurement on the GMAT, and my apologies for taking a detour from the main purpose of this thread:
Because the test is finite in length, and because factors like lucky or unlucky guessing play a role in determining your score, the test will not always arrive at a perfectly accurate assessment of your 'true GMAT ability' - that is, the score you'd get on an infinitely long (if such a thing existed!) GMAT test. The Standard Error of Measurement measures how far the test can be from estimating your 'true level', and for test takers not at an extreme in ability, the standard error of measurement on the GMAT is, as Jose points out above, about 29 points.
So what does that mean? In technical language, the standard error is the standard deviation your test scores would have around your 'true level' if you took the test multiple times, and it is normally distributed: in numerical terms, 68% of the time, you'll score within 29 points of your 'true level', 95% of the time you'll score within 58 points, and essentially 100% of the time you'll score within 87 points.
For test takers, that means they should expect there may be a small difference between their real test score and their average score on reliable simulation tests. More than a 30 point swing up or down is unlikely, however, and more than a 50 point swing would be very rare, at least absent exceptional circumstances (e.g. falling asleep at the computer terminal in the test centre).
Of course, with good preparation you can improve your 'true ability level', which, as Jose points out, is the best way to increase your chances of a good score.
