Generally speaking, it's tough to game the test. Sure, the absolute best strategy would be to get most of the earlier questions right without spending extra time (or much extra time)... but if you can do that, you don't need to plan anything b/c you're already a 99th percentile test-taker!

How do you balance spending, say 2.5 min on average for some number of questions and then switching to 1.5 min per question for the remaining questions when you don't have a stopwatch and you're in the middle of a very stressful situation? Particularly when the questions have been designed to be done in 2 minutes, so you're really shooting yourself in the foot on all those questions you're trying to do in 1.5 min instead. Some small percentage of people might make that work - but most people won't.

From a practical standpoint, once people decide that it's okay to spend more time on early questions, most people then get sucked into spending too much time (both on those and later questions) and then running out of time before the end. I've been doing this for more than a decade and I've only run into a handful of people who can do that successfully (and, most of the time, I suspect it's luck). The vast majority of people I've encountered end up hurting themselves - sometimes quite a bit - by the end.

Teachers / 99th percentile test takers can "play the game"... but we're not even really playing, because we actually can answer even the hardest questions within the expected timeframe. So trying to "game" the timing just isn't a practical strategy for anyone - top test takers don't need to in the first place and everybody else is much less likely to make it work successfully.

By the way - that quote above that said if your initial correction rate is high, your final score could be high. Well, sure, it could be high. It could also be low or anywhere in between. It depends entirely on how much your earlier effort costs you at the end - and most people end up losing this game when they play it too aggressively.

To add some data here:

-the test will make larger adjustments to your score early in the test if you complete the test, and have time to attempt every question. If you proceed through the test and attempt each question to the best of your ability, the test will gradually zero in on your ability level; the adjustments will be large early, and small late. It is for this reason that many say you should spend more time on earlier questions. This ignores the fact that adjustments can be very large at the end of the test if you do not proceed according to your ability- that is, if you need to guess, and have a long string of bad answers. When some advise you to spend more time early, the assumption is that spending more time early will help more than losing time later will hurt.

It's a questionable assumption, to be sure, and there is published research that demonstrates just how questionable it is. The GRE is quite similar to the GMAT- it's a computer adaptive test, with similar mathematical content- and ETS conducted a large-scale study to determine just how much test-takers would benefit from having additional time to complete the test. The results are here:

www.ets.org/Media/Research/pdf/RR-03-13-Bridgeman.pdf

GRE test-takers were given 50% more time to complete each section of the test. That would be like having 3 minutes per math question instead of 2 on the GMAT- a lot more time than normal. You might think people would see a large score improvement. Well, the average score, for those with additional time, increased from 664 to 671 on the Quant- 7 points on the /800 scale, a minuscule improvement (and the improvement was also only 7 points on the Verbal). Additional time is nowhere near as valuable as most people think; when questions get too hard for a test-taker an extra minute will not often make a difference. Having less time, on the other hand, is likely to be severely damaging- with too little time, you will clearly have no option besides guessing. Spending three minutes on early questions isn't likely to help much on those questions, and having only one minute per question on late questions is likely to hurt a lot.

The only group that saw a significant score improvement with added time were those at the absolute lowest end of the ability scale- those with scores below 350 on the /800 scale- so it may be that the weakest test-takers should adopt a different time strategy than others. But for the average or above level test-taker, I'd recommend the following:

-spend at least one minute trying to understand (and solve) each question- some questions are much easier than they seem at first;

-if after between 60-90 seconds, you still can't see any path to a solution, try to eliminate bad answers (many strategies can be learned for this purpose), and move on;

-if you can see a path to a solution, but the question might take 2.5 to 3 minutes in total to finish, finish it; correct answers are very valuable. Don't limit yourself to 2 minutes per question, but only spend extra time if you can see it will get you an answer;

-whatever you do, do not get stubborn about any one question. Often questions look easy, but the solution doesn't come to you under the time pressure of the test. Don't be afraid to cut your losses and move on- the time you save will help you later;

-don't worry about whether a question is early or late in the test- early questions are not nearly as important as some claim;

-be sure to finish the test no matter what, even if you need to guess. There is a substantial penalty for not completing the test.

The main point to conserving time is to avoid having less than two minutes per question late in the test- you don't want to run down the clock so far that you need to do a lot of random guessing. Worst case, you should be guessing at no more than two questions at the end, and preferably none; more than two, and you're risking a substantial penalty.

Perception does, indeed, have a huge impact in what you might expect to score. We're used to our perception being pretty good in the real world, but your perception is actually pretty faulty on this test.

First, as amitdgr points out, you don't know which ones count and which ones don't count.

Second, there are plenty of times you think you got something right when you actually got it wrong because you fell into a trap - the whole point of falling into a trap is that you don't know you fell into it.

Third, it's incredibly difficult to gauge difficulty levels. Difficulty levels are based on the abilities of the population as a whole. Your perception of difficulty is based on your individual strengths and weaknesses. The way you "feel" about a question is not necessarily correlated with the way the population as a whole actually performs on that question. Sometimes it is, sure. But, plenty of the time, you think something's easier or harder than it's actually rated simply because of your bias (your personal strengths and weaknesses).

Also, with verbal, there's a bit of an additional problem (relative to quant): because verbal is somewhat subjective (again, relative to quant), verbal is rather dependent on the word choice and rhythm of the language used by the testwriters. That's why it's so important to study OG and GMATPrep for verbal - because you want to learn the rhythm of questions written by the people who are going to be writing the official test. A lot of people study from so many different sources that they don't learn what I'll call the "official language" for verbal questions well enough.

I also suspect that, because ACT writes the test now instead of ETS, we're starting to see new wording and even types of questions (eg, the "argument" type questions popping up in RC, asking us to support some part of an contention presented in the passage). But I also suspect that most of those questions, at this point anyway, are experimental.