Hi everyone, my name is Akshat Srivastava. Five years of on-and-off prep, no paid course or tutor (the
GMAT Club tests and adaptive quizzes I use are redeemed with GMAT Club points, and GMAT Ninja’s Wednesday sessions are free), and one official mock sitting at 625. I got tired of guessing why, so I exported my whole
error log and sliced it by difficulty, failure mode, answer-choice bias, and time of day.
Three things came out that I honestly can’t tell are real signal or just me fooling myself:
On Data Sufficiency, I pick "C" on 47% of my misses. C is right only 16% of the time. Almost half my DS errors are one reflex.
A full year of logging, and my Hard-level accuracy hasn’t actually climbed. Volume looks like it just plateaued me.
When I model the swing between sessions, a single mock turns out to be ~70% noise. My good days and bad days move far more than sampling alone should allow.
The whole thing is in the attached report: the difficulty matrix with confidence intervals, the DS and CR breakdowns, the time-of-day effect, and the variance math behind that last point. It opens in any browser. Here is the link :
https://claude.ai/public/artifacts/6bed ... 5748feb1adWhat I’m after isn’t applause, it’s holes. If you’ve crossed 700, are these the levers that actually moved your score, or am I chasing the wrong things? Tell me where I’m reading my own data wrong.
(Once I’m through my own exam in early July, I’m glad to build the same report from anyone’s error-log export, free. Drop a note and I’ll find you.)