I just got the Compass Prep email also. Notice there’s a link to their methodology just prior to the table:
http://www.■■■■■■■■■■■■■■■/estimating-national-merit-cut-offs/
I’m just starting to read that.
@Ynotgo – Thanks so much for sharing these estimates - interesting that it includes a “most likely” and a range of scores. I see most of their students are in Cal.
Hey @thshadow I’m reading the new article and, when looking at the description of their methodology, I’m seeing a lot of things that look… familiar to me. Like I typed them already. I’m going to have to change my name from “Bill” to “Art”.
Yeah – some of the Compass piece looks pretty similar to @DoyleB’s and others’ contributions here!! Hi"Art!" 
The one part that gives them a little bit of extra credence (to me) is where they say:
[quote]
Any method for estimating National Merit figures based on College Board data should contain a double-check or triple-check, because the released College Board tables are preliminary and may, in some situations, be inaccurate. One important double-check is to look at actual student results. For example, if School X always sees between 27 and 33 National Merit Semifinalists, then School X is a good predictor of its state’s cut-off. The changes made to the New PSAT are unlikely to significantly change the number of recognized students at School X. At what New PSAT score would School X have 33 National Merit Semifinalists? 27 Semifinalists? For California results, Compass has been able to perform this double-check, and this also gives us insight into adjacent scores. The same assumptions should not be made about a school that typically sees few Semifinalists or has fluctuating numbers.
/quote
They claim that they’ve done some checking with actual scores from California schools which support their claim.
Who knows, maybe they’ve just read the “anecdotes” in this thread…
Compass predicts the Commended cutoff at 207, within a possible range of 204-210. Would it be fair to say that if the actual Commended score comes in at, say 204, that the range of the state cutoff scores correspondingly would be less, and if the Commended score comes in at 210, that the range of the state cutoffs correspondingly would be at the higher end? Or is it that you can’t link the Commended score with state cutoff scores?
Now that Compass has verified (or copied) the work here, is the science now settled?
Can we join our regularly scheduled programming already in progress?
Most importantly, is it safe to ping @GMTplus7?
“One important double-check is to look at actual student results.”
And maybe actual results are from my tables 
So, has anyone gone through these predictions and other data to see whether the actual number of questions the students can get wrong has changed?
@EarlVanDorn - they’re not taking the percentiles of that school district and assuming it will be the same nationwide. (They’re not predicting anything related to percentiles.) Rather, they’re saying “this district typically has 60 NMSF”, and so they see what cutoff is needed to get 60. They don’t need to factor out anything or renormalize - they just need HISD to be about the same as it has been in past years.
For predicting commended - that’s likely a lot harder from the data they have. (Unless they know how many commended scholars HISD typically has.)
@Ynotgo : I called the college board.
FWIW
A couple of years ago I looked at the demographics of large state sample sets. HISD was one dataset I managed to get my hands on. Here is what I found:
http://talk.collegeconfidential.com/discussion/comment/17607836/#Comment_17607836
Our school (CA) had 9 last year and so far I haven’t heard more than 9 above 220 this year.
@srk2017…I think that is a good sign that CA cutoff will be 220.
I would think race (other than Asian propensity towards testing) explains little by way of performance at the top end of the scale, perhaps it explains the mean and mode. In OH there are appx 1450 high schools and only ~10% (150 or so) schools report national merit scholars. School districts, income, educational backgrounds of parents probably explains the lopsided distribution of high performing kids more than anything else. Thus, I wouldn’t be surprised if ~60 SF from HISD come from families that are at higher end of income distributions, with professional occupation families. The relative high population of latino students in HISD should have very little influence on SF (or the commended level), when these schools also have students from highly educated families right within their mix, Latinos included. I mean this would be true of any major urban center.
Long time lurker. If the Compass best guess predictions are correct, the following states will have higher cutoffs for SF than at any time in the past 6 years (assuming I typed in my data correctly)
[1] “Alabama” “Arizona” “Arkansas” “Colorado”
[5] “Florida” “Idaho” “Iowa” “Kentucky”
[9] “Louisiana” “Michigan” “Mississippi” “Nebraska”
[13] “Nevada” “NewMexico” “ND” “Ohio”
[17] “Oklahoma” “SC” “SD” “Utah”
[21] “WV” “Wisconsin” “Wyoming” “Commended”
Considering that CB has a lot of leeway in how they convert “number of correct answers” into “SI”, it seems a curious decision by CB to do this.
@VryCnfsd…didn’t I read something about them trying to get the scores closer together? Maybe that is how they plan on accomplishing that goal. I maybe totally wrong but I could have swore I read that in either a forum or article.
From a marketing perspective, compression of cutoffs would be good for the College Board. Every year there are complaints that kids in high cutoff states are getting robbed by kids in lower cutoff states. If there is cutoff compression, those complaints will be lessened.
This doesn’t say anything new, but I’m surprised that the College Board needed to release it, and I’m fascinated that they released it on Forbes.com. Seems like rather targeted marketing to high-end students/parents. Anyone know how to contact Cyndie Schmeiser?
thanks @candjsdad for the link. I read this as CB’s first run at reputation salvage. The percentiles they gave seem to be highly misleading and CB knows it and is now prepping for the backlash. From Testmasters, Compass, and the NHRP Southern cut-off of 204, we now have a good sense of where the national commended will fall: around an SI of 207. A 207 is published as a low 99% by CB. Even assuming CB was basing percentiles on a test population of 3.5 mill and not 1.7, this cut-off at 207 still does not account for about 15,000 to 20,000 testers nor does it explain the misleading user percentiles many testers would have received based on SIs of 207 plus. I expect CB will have to admit at some point their test samples gave them bad data.