Personally, it underpredicted my score by ~15 points or so. My NBME scores had a nice upward trend as I got closer to the date, and I think I did NBME 4 first. I would say you should use NBME's for their predictive value only as a ballpark. Everyone'll have a different experience as far as which NBME predicted their score closest, which ones they think were not at all predictive. But if for the most part you're hitting around where you would guess based on your grades/test-taking ability/goals/UW percentage and for the most part improving as you go, be confident that you'll even outdo your best NBME. Don't get hung up on trying to see into the future, whatever happens, use the result as another tool to produce the result you want in the end. Keep at it man.
yeah it's hard to say which nbme most accurately predicts your score. each one only has 4 sections whereas the real one has 7. so therefore, some nbme's might be biased towards one subject whereas the real one has more sections and may survey more disciplines. for example i took nbme 5 2 weeks prior and that underpredicted my score by 27 points. and it's not like i was very under-prepared 2 weeks prior-i was actually pretty close to ready by that point. i took nbme 6 a month prior to the test and scored 4 points over my real score. nbme's 3 and 4 underpredicted my score about 10 points or so, whereas nbme 2 overpredicted by 2.
i averaged all my nbme scores and overall, it underpredicted my scores by 7 or so.
only under-predicted my score by 7 points - but then again, i did take it pretty early in my prep (after only 1 read of FA and Kaplan Home Study).
anyway, any of the NBMEs is pretty accurate as far as predicting the "ball park" of ur score (i.e. 240 Vs 250 etc). and that is what you should utilize them for (besides identifying weak point/subjects)
Not sure why people seem to be surprised that an NBME might underpredict their real score. People usually take these things before they take the real thing, so that they have an opportunity to assess their progress and improve upon their performance... so wouldn't you expect to have a lower score than the real thing? The purpose of the NBME is to give you a snapshot of your current knowledge base, not your knowledge in 1, 2, or 3 weeks. Sometimes it is really that second time through the material which really solidifies your knowledge in weaker subjects that can allow big jumps.
While the style of the questions can change from from the older NBMEs to the newer, you're still being graded relative to thousands of other peers who have taken the thing. That being said, people need to realize that it isn't perfect. Statistically there are far fewer questions in the NBME than there are in the real thing. You should expect a certain amount of variance. If you're desperate for a more accurate snapshot, take two consecutive nbmes and average your two scores.
I took an NBME at 0 weeks and 4 weeks. I thought it perfectly predicted the trend in my scores, correlating well with my question bank performances up until the real thing.