That's an interesting theory but I'm afraid the reason is far more mundane.
I'd love to hear the reason... I know this thread is extremely old, but I'm really curious why the effective cap for section scores appears to be 51.
My theory was the same as above - I thought maybe the CAT was restricting the difficulty of questions for top test-takers based on their expected finish time. I.e., if a "top test-taker" (able to score 51+) is answering questions correctly but moving too slowly on more difficult questions (implied level 52+), the CAT chooses not to give the harder questions, thus settling at a lower level due to the implied time constraints.
It seems unfair for the test to cause someone to run out of time while answering 52+ level questions if the tester is able to finish on time while answering 51-level questions. Everyone talks about how the CAT knows the actual % of correct/incorrect answer for each question and uses that to determine score (e.g., rhyme's informative post at gmatclub[dot]com/forum/gmat-scoring-algorithm-my-observations-28493-60.html#p204280), but I have not seen discussion about how the CAT accounts for average response times for each question - and it seems likely that the CAT takes this information into account (or, at the least, that it could and should take this information into account). I guess I am reaching the simplified hypothesis that no 52+ level questions exist with average response times <3mins, and the CAT avoids these questions if a test-taker's time remaining is less than 3mins/question (where "3mins" is an approximation and I am sure the coding of this into the CAT's logic is actually quite a bit more complicated than my simplification).
In my case, I remember taking quite a bit of time on a particularly difficult question towards the end of the quant section on my test (I believe it was question #32). I spent ~5mins and believe I got the question right, but I definitely ended up behind the clock, I rushed through the last five questions (where were relatively easy and took ~1.5mins each), and I finished with 3 or 4mins to spare and a quant score of 51. But I also felt like I could have scored higher than a 51 (given a little extra time or one more 52-level question), and I'm curious if anyone has evidence (or suspicions) that the time remaining affects the CAT's decisions about which questions to give you.
Anyways, just curious. I'm no expert in standardized test creation (or item response theory) - just trying to understand why so many bright people appear to score 51 instead of 52+.
Would love to hear from you all - particularly any 99%ers (looking at you NonYankee, kevincan, and everyone else at gmatclub[dot]com/forum/99-ile-clubbers-add-your-name-to-the-list-77041.html) or test experts (rhyme, Hjort, any all others).