I really think that @thshadow was trying NOT to offend. What he said can be taken more than one way.
@thshadow, a similar question for you: if x and x-1 have identical P values under definition A, how does it make sense using your heuristic to assign P to x-1 under definition B? Thanks a lot. I really appreciate the clarity of your explanations.
@likestowrite I think I understand what @thshadow and @DoyleB mean when they say âmove one SI Unitâ rather than move one percentile. If you look at Understanding Scores 2014 you see that 202 is assigned the 97th percentile. What that meant then was that the 97th percentile included all scores UP TO BUT NOT INCLUDING 202. That is actually not the same thing as saying that 202 was in the 98th percentile - itâs just in a percentile thatâs higher than the 97th. This is part of the problem with discretizing continuous functions. The SIâs are in discrete intervals but they are fit to a continuous percentile distribution which is then chopped up. If everything was continuous the value of 202 would be assigned something that was barely above the 97th. Hope that makes sense.
Anyway, we now look to the page 11 SI table where 202 is assigned the 98th percentile. We know that this year that means all scores up to AND including 202 are in the 98th (cumulative). Therefore, 202 is actually assigned a higher cumulative percentile now (98th) than it was last year (something barely above 97th) by an amount that approaches (or is barely smaller than) 1.0%.
So what happens if we try to put last yearâs table to this yearâs terminology? We want to find the percentile that INCLUDES 202. We know that itâs barely above 202 but we need an approximation. Well, the closest approximation we have from the table is the percentile thatâs assigned to 203. The SIâs are whole numbers in units of 1. Thatâs why they are saying âshift one SI unitâ.
@WGSK88 from post 2079 this is extremely useful information so thank you! It looks like 11 kids between 215 and 217. Do you have any idea how many are, say 217? Iâm thinking the SI may be able to get down about that far without blowing away the historical NM numbers. Odd that so many are at 218.
So the strong evidence (just one school of course) is that the cut-off would be a minimum of 217 with a higher likelihood of 218. I canât believe it would jump more than three points because I donât. want. to. go. there. So Iâm saying 217-218.
That would put MN in a 216-217 range if it continues to track IL as it has been doing.
This anecdotal evidence is very helpful - I hope we get more of it.
@liketowrite, you probably know my profession - junior in HS, and I guess you are not a teacher either. But I would appreciate very much if you can answer me a few questions
From your post
2014 PSAT SI % chart
99+ = 224-240
99 = 213-223
2015 PSAT % chart as I converted it
99+ =222-228 (most subjective part)
99 =214-221
Q1: What is your basic or fundamental concept to have 17 slots (99+ range in 2014 PSAT) compress to 7 slots ( (99+ range in 2015 PSAT) ?
Since, every thing has ripple effect then
Q2: What are your basic assumptions to have 11 slots (99 range in 2014 PSAT) compress to 8 slots ( (99 range in 2015 PSAT) ?
Q3: What is your necessary and sufficient condition to answer Q1 and Q2 ?
I came up estimates with a different angle from other CC posters, so I welcome your new way very much
By the way, I believe @thshadow did not mean rude to any body in this thread, especially to you, Sir?
Edit on last para. of 2081: We know that itâs barely above 97%. NOT barely above 202. Sorry for the confusion.
I agree that post 2019 is very helpful. Data from a large school consistently producing a good number of NMSF is especially helpful, especially for determining the cut off in your state. For what it is worth, I think you can be pretty confident that you daughter will qualify for NMSF. It sounds like your daughter attends a very rigorous school and, if anything, that school will be over represented this year as compared to previous years. Congratulations to your daughter. She will be well prepared for college no matter where she decides to attend.
@Mamelot Sorry, I have already bombarded the GC with so many email that I canât ask how many scored 217?
What happens for example if the cutoff for a state like IL say 218 and there are 699 NMSF from IL. If there are 715 children meets the cutoff, since 218 or so many they canât exactly cut off at 699 children, then what criteria they will use to eliminate 16 children (Letâs say 250 children scored 218 - How they will eliminate the 16 children with 218 cut-off?)
Will they use the scaled score of 1520? Since 218 index can be for children who scored from 1450 to 1470?
And if I am incorrect and she just misses it. I know I am not incorrect about the last thing I said. She will be well prepared for college no matter where she decides to attend.
@WGSK88 the NMSC uses different criteria than all the predictors on this thread! Fortunately! The way they choose NMSFâs is that they decide how many to assign to a state based on itâs allocation of total HS graduates in a given year. Obviously that number wonât change from year to year so the number will be close to what it was last year. Then, they rank the students from highest to lowest SI (SI only - Total Score does not play into this whatsoever). And they stop at the cut off that comes closest to the allocated number. It might over- or under-shoot. Itâll just be the SI that results in the number of NMSFâs closest to the allocated number. Itâs that simple. They wonât split down the middle of a 218 or any other SI cut-off so no need to worry there.
Edit: and you are correct that the same TSâs may result in different SIâs and that as a result one person gets NM and another with the same TS doesnât. However, NM has always been determined 1/3 math and 2/3 verbal. This year is no different. We really donât look much at D3âs total score except to predict her SAT. Right now SI is the number du jour.
@SLparent â You could â it might help to have a focus some targeted out reach and then feed back info to this
thread â Just know there are other threads out here too & there are people reporting there we could reach out to to update the detailed list (in this thread) that was created a few days or so ago.
http://talk.collegeconfidential.com/high-school-life/1848461-class-of-2017-psat-cut-off.html#latest
and
http://talk.collegeconfidential.com/discussion/comment/19233501#Comment_19233501.
And we have good work done here that would be amazing to incorporate as a team effort:
http://talk.collegeconfidential.com/discussion/comment/19233501#Comment_19233501
and here:
http://talk.collegeconfidential.com/discussion/comment/19225499/#Comment_19225499
Yes, I really didnât want to be rude. @likestowrite had explicitly asked me to respond in I think 3 different posts. I ignored many of them because I couldnât think of anything helpful to say. I finally did respond, with âapologies in advanceâ and smiley faces. Regardless, my response was bad because I knew it was rude, but I honestly couldnât figure out a way to phrase what I wanted to say without being obnoxious. So in summary - I hate being rude - I knew I was being rude - I did it to try to communicate - I hoped he would take it in a light-hearted way. And again, Iâm sorry, I was definitely hoping that no oneâs feelings would be hurt.
Just to use different numbers from an imaginary test.
Letâs say the New Test (out of 220) has the 99% percentile for exactly the range 200 - 210.
Say the Old Test (out of 120) has 99% for exactly the range 100 - 110.
If the New Test has <= percentiles (like the new PSAT), and the Old Test has < percentiles (like the old PSAT), the easiest thing would be to say:
OldTest (with <= percentiles) would have 99% for exactly the range 99 - 109.
I guess another way to put it. If you have a percentile table where the percentiles are < percentiles, and you want to convert it to <= percentiles - just subtract 1 unit from every single score in the table.
So if your old table, with â<â percentiles, is (and Iâm just doing this from memory / making it up):
224 - 240: 99.9+
214 - 223: 99.9
200 - 213: 99.8
etc, then changing it to â<=â percentiles is very trivial (and extremely specific):
223 - 239: 99.9+
213 - 222: 99.9
199 - 212: 99.8
In short, it is trivial to update any old PSAT percentile table from the old < percentiles to the new <= percentiles. Just subtract 1 from the score everywhere. Thatâs it!
After that, deciding how to match up the percentiles between the tests is a separate question, which has nothing to do with the percentile definitions.
@ca1571 #2071 Well said. I think most of us on this thread should be proud and grateful if we have kids who scored in the 99% (whether they get SF or commended or not and whether the 99% is maybe a 98% or 97%. Itâs still an amazing achievement). For those here for whom NM is a deal breaker I really hope it comes through. While weâve been through SATs (DD took it December and did better than PSAT and DS took the infamous June SAT for CTY qualification (he qualified but was furious about the section they disregarded)), perhaps I suffer from newbie syndrome wrt the PSAT b/c Iâve found this whole thing so fascinating and will follow it until the mystery is resolved. That said, Iâve found the level of commentary to be sophisticated and educational but then if most of the posters are parents of testers in the top 1%, that shouldnât be too surprising. 
FYI, hereâs another anecdote from my area, though it doesnât really have much detail.
Note that my area is one of those super-high-achieving areas in California. I asked my daughterâs SAT tutor (who is very expensive, so I assume he tutors many high-achieving kids) what he heard about PSAT scores. I asked him if he thought the cutoff would be closer to 223, or to 214. He said heâs only seen a handful of 220+ scores - which I think implies that whatever data he has access to makes him think the cutoff will be a lot lower.
So take that for what itâs worth.
Based on what have been posted so far, it does appear that high scoring state like CA will have relatively lower cutoff score and the gap in cutoff between states indeed coming down. May be CA is in 220 or 221 and IL is in 218 or 219 for example. CA has almost 2000 NMF, we have not heard too many of super high scores here posted/discussed about. being in CA I hope so!
@likestowrite, It seems to me that the new definition of percentile would have a very small impact on scores at the upper level. Before, with the OLD definition, your percentile equaled (% of the total test takers which scored below you). For example, if you scored a 214 last year, your percentile was 99% because 99% of the total test takers scored below you. With the NEW definition, your percentile will again be the (% of total test takers which scored below you), but now we will also add in the (% of test takers who also scored the same as you at SI 214). With so many bands at the 99% level, the (% of test takers who scored the same as you) will be very small (well under 0.1%). Your percentile is not going to change much with new definition.
@WGSK88 thanks for all the data
Some new data from GC at The John Cooper School (Houston area elite school). They traditionally have 5-7 NMF; last year they had 5 for 102 juniors. This year they had 109 take the test and the mean average was 1283.
@CA1543 you are right at post #2089. Actually, this is how I started my excel fileâŠwent out there back to Jan 7 when people first started asking about scoresâŠbut Iâve been attached to this thread & kind of forgot about other threads. If anyone wants to turn in actual scores, from friends, acquaintances, classmates, itâs greatly appreciated!
@thshadow, good idea. I just sent an email to my daughterâs tutor. Also very expensive tutor ($175/hour). Tutors mostly kids from the Eanes School District. For those of you unfamiliar with Eanes, it is ranked 2nd best in the COUNTRY by Niche (https://k12.niche.com/rankings/). Extremely affluent area of Austin.
Iâll let you know what she says.
Thanks poster 2096. Wow. I am really afraid the clustering of NMSF is going to worse than ever before and it was embarrassingly discrepant already.