by StaceyKoprince Sat Dec 12, 2009 4:39 pm
Technically, the algorithm does not base its next question choice on just what you did with the last question. Instead, the algorithm calculates your total performance to that point on the test (including the last question) and gives you a question based on that. So it's not just possible but somewhat common, actually, to get a 700+ question wrong but still have an overall score to that point in the section that merits another 700+ question.
Also, difficulty level is not the only metric that the algorithm uses. The algorithm also has to factor in specific question types, specific content areas within those question types, etc. The test might determine, "okay, the next question needs to be a DS geometry triangle question and it needs to be in the 700 range" and the closest question might be a 670 or it might be a 730. Remember that the standard deviation of standardized tests is fairly high (30 points on the real test, 50 points on our test), so that doesn't actually represent that much of a swing.
Finally, you may have gotten more wrong, but they were at higher difficulty levels, so that's why your score went up. As long as your wrong answers are fairly spread out, you can get quite a number of questions wrong and still get a high score because you are constantly "recovering" and lifting your score. If you have a string of wrong answers in a row, though, then that can start to bring you much further down.
Stacey Koprince
Instructor
Director, Content & Curriculum
ManhattanPrep