National Merit Cutoff Predictions Class of 2017

@PAMom21 I tried this
http://talk.collegeconfidential.com/discussion/comment/19166404/#Comment_19166404
and cannot reconcile the concordance table with past Understanding Scores data.

Yeah, I didnā€™t really factor in subtest percentiles at all in what I was doing, though my sonā€™s ā€œcorrectedā€ scores do land close to the current percentiles (comparing his actual percentiles to the ones tabulated somewhere here on CC for his converted scores).

If the publish similar reports in February, that will be interesting, as that will highlight numbers of top math kids, versus top verbal kids (but sadly not all top kids).

What. A. Mess.

Question for @GMTplus7 ~ When you compared your own kidsā€™ data, were you using the percentiles for the year AFTER they tested? Or the year they tested? Because those percentiles given to us in 2015 were from the previous year. Gotta wonder whose data they actually based their decisions on!?!

http://i67.ā– ā– ā– ā– ā– ā– ā– .com/2qk144x.png

@paynforward, thanks for making the chart.

Interesting that so few people seem to have taken the 10/28 test. I wonder if the population who did this was substantially different, and whether that or test difficulty made the scoring more lenient.

Iā€™m a bit confused also. Wouldnā€™t they have tested out the questions for both of these tests in prior years and constructed both tests to be of equal difficulty based upon how the test groups scored? If so, why would they need to adjust the scoring such that missing 5 on writing was a 33 on the first test date and a 35 on the second test date?

Still wrestling with all these different percentile reports also.

@PAMom21 just this morning I kinda did the same thing for MN that you did for PA. My results are Mean = 215.6, Median = 215.5 and Mode = 215. Testmasters predicts 215. The previous cut-off was 214.

My range was 213 - 218 with over 70% of the data in that 215-216 range. Definitely resembles a bell curve, though there are only 18 data points involved :slight_smile:

For anyone interested in the methodology I tried to factor in individual test percentiles as well but the whole thing got too complicated so I didnā€™t worry as much about that other than make sure that the subscores ā€œmade senseā€ based on what I knew about MN. Specifically I tried to use all combinations ranging from 690 to 740 that summed to 214. I believe that going outside that range would have to involve some pretty low scores which Iā€™m fairly certain werenā€™t representative of what made up that 214 cut-off.

This is very different from my earlier prediction of 218 for MN but that turned out to be too high because I mistakingly thought the previous cut-off was 215, not 214. In fact, the cut-off had decreased by one point. Oops. Plus, this time I looked at a range and figured out the ā€œmaximum likelihood estimateā€ from my 18 data points :slight_smile: While 218 is still in the realm of possibilities, it seems more likely that the cut-off will be lower than that.

@MrAustere hopefully youā€™ll like this result better, although itā€™s of course just as preliminary as the prior one had been (and not any more scientific - perhaps a bit more accurate now that Iā€™m starting with the correct cut-off #! ). Iā€™m sure there are a ton of errors and that Iā€™m not using the appropriate distribution etc. etc. All I have really been doing is eye-balling the available data.

@Mamelot So you believe that a 215 will be the lowest possible cutoff for MN? Last years cutoff was 214. That was out of 240 though. Because itā€™s now out of 228, donā€™t you think it will be lower?

http://i65.ā– ā– ā– ā– ā– ā– ā– .com/2rom4y0.png

Scores posted previously were for Oct. 14 test

Anybody have an idea on the Washington cutoff?

@MrAustere not necessarily. While the concordance data is preliminary itā€™s still worth considering. What I can see by eye-balling the tables is that the new scores donā€™t track the old ones very well once you get past the very highest scores. Itā€™s quite possible that the distributions are very different. Certainly with two very similar tests, a re-scaling from 240 to 228 would produce lower cut-offs across the board but thatā€™s not the case here - the redesigned test is a very different test from the old one.

Everyone has had an opinion on those concordance tables by now, of course, and itā€™s quite possible that they have been completely debunked on CC LOL. Iā€™m slow on the uptake so still using the tables, on the theory that they are the best set of data we have (even if they turn out to be completely wrong).

FWIW Testmasters also predicted a small increase for MN - in fact, their analysis concludes that the lowest cut-off scores will come up, and the highest come down, with not a lot of movement in the middle (although there is still a bit of movement for some states). I noticed that too the other day just by eye-balling the tables and looking at some simple combinations of scores. It seems incorrect - and it may well still prove to be! - but then again the distributions might be very different . . . I think weā€™ll just have to wait for more data. The final tables are available in May, I believe. I will also be checking to see what the state of MN publishes for PSAT results, if anything. Do you happen to know what state reports may be available regarding these results? I havenā€™t looked yet.

Thanks and salutations to Payn, GMT, Mam and all ccā€™ers spending the weekend grinding the numbers (and pasting tables!) After comparing 2014 scores to 2015, I agree there appears to be a lot of noise due to weighting esp in the 720-740 range where there will be a LOT of kids in some states. DC improved from 217/240 to 222/228 with all of the improvement in the Writing section. Given the concentration of mathy BS kids assigned to Massachusetts, I suspect the threshold will be quite high. Feeling Good, Louis!

@Mamelot But it all depends on how the kids actually did on THIS test, since the actual cutoffs are based on the highest scorers in each state, vs all the math weā€™ve done which is based on a conversion of this year to last year, and using these fake sample national and user % the CB has given us. Thatā€™s where the cutoffs could drop, if they donā€™t have enough high testers. Its hard to see how the ā€˜normalā€™ kids did, since CC has only high testers who are vying for NM.

Just curious, what makes you guys think there is a remote correlation between 2014 and 2015 PSATs as found in the College Boardā€™s table? That table is PRELIMINARYā€¦ itā€™s based on a tested sample. You have no idea how the real numbers will play outā€¦ Those who signed themselves up for the Saturday testing date (and therefore part of that testing sample) are probably motivated, intelligent students wanting extra test prep. Not your typical American HS student.

@Mamelot Iā€™m not aware about what resources there are. At this point it looks like Iā€™ll have to wait and hope for the best!

any speculation as to whether 222 is good enough for California?

For CA, 220 and higher is highly likely to make it.

@suzyQ7 youā€™ve touched on some of the limitations of the preliminary analysis. Once the final tables are released there will of course be more number crunching. Iā€™m done with it for now, I think! :slight_smile:

@MrAustere I think thatā€™s about as good an action as any at this point. I really hope you make it but to me at least it looks like you have a great chance!

@appgodxoxo I agree that itā€™s preliminary and therefore not as accurate . . . Iā€™m guessing that CB isnā€™t going to release something that they expect to be wildly different from the final. But weā€™ll see one way or the other come May.

Also, when are the cutoffs usually released?

I think we just have to wait until may, and assume that if you scored over 220, youā€™re safe pretty much everywhere!