Hey, guys
Couple of things
- You're absolutely right that some of our questions require too-extensive calculations. As we identify these, they are being re-worked (see next bullet).
- At the same time, you should expect to hit problems that you cannot do in the 2-min timeframe on the real test; it happens to everyone (even me), so you'd better practice that educated guessing technique!
- A poster mentioned scoring a 700+ despite getting half the questions wrong. That's how the official algorithm works, too. Unless you're scoring at the 750+ level, expect to get a LOT of questions wrong. The difficulty level of the questions is determining your score, not so much the percentage correct.
- Our algorithm is actually pretty close to the official test's algorithm, but we do not yet have experimental question capability built in, which ups the apparent difficulty level in the math section because you never get any "breaks" with easier experimental questions. This is a flaw, yes, but it was most important first to get the algorithm right; now, we're working on the experimental question capability. We hope to have that by next year.
- Not having the experimental capability means that we can't test our questions before we give them to you (which is true of every test prep company - nobody incorporates this). However, we've been running extensive analyses on the test questions as people take the tests and we are using those as a stop-gap right now to do two things. First, the data provides difficulty levels according to the population of test-takers (as opposed to what the question writer just thinks it should be - which is nowhere near as accurate), which allows our algorithm to select questions based upon 10-point differences in difficulty level vs. the 50 to 100 points that is common in the industry. Second, we use the data to identify test questions that are not discriminating enough or have other flaws (too hard, too easy, takes too long). Those get flagged and we re-write or discard them. I've been fixing a bunch of RC questions myself this past week. This is essentially what the experimental section of the official test is all about - and, again, test prep companies have historically not bothered with this stuff at all - once the questions are out there, that's that. Until now.
- I definitely agree that GMATPrep is the best practice test available b/c it is built on the official test's algorithm. Be aware, though, that for the higher level test takers, GMATPrep is too easy. The database of questions is limited, so if you are scoring at the high end, you are going to exhaust the pool of available questions and the test will have to give you easier ones than you would have gotten on the real thing.
- A poster mentioned that, if people had unlimited time, most could get most problems right. Only if that person had access to outside study sources or other people. Studies have actually been done to show that the more time you spend above about 3 minutes, the more likely you are to get it wrong. Essentially, if you can't figure out how to do it in about 2 minutes, you don't really know how to do it, and spending even more time on it won't improve your odds - in fact, it will decrease your odds. (That doesn't mean you
won't get it right, of course. But over many problems, your performance will be lower, on average, on the ones that you take 3+ minutes to do vs. the ones you take 2 minutes to do.)
We really appreciate your feedback, guys - keep it coming. I'm going to send a link to this thread to our algorithm gurus. Thanks!
