High School Grades, College Admissions, and Scholarships

10 May 2017 Version

Roughly Granby Memorial HS


Roughly UConn - Storrs





Text Box: •

Candidate’s Predicted Percentile Class Rank at specified College



Candidate’s HS Percentile Class Rank, with or without grade renormalization

Text Box: Equivalent mean SAT (M+CR) at High SchoolText Box: Linear extrapolation of 25th and 75th percentile college values.                                                                                


Text Box: Calculation using Candidate’s equivalent mean HS SAT (M+CR) and linear extrapolation of the College’s published 25th and 75th percentile college values.


                                                                               Text Box: Calculation using Candidate’s equivalent mean HS SAT (M+CR) and linear extrapolation of the College’s published 25th and 75th percentile values (•).


Text Box: Calculation using Candidate’s equivalent mean HS SAT (M+CR) and linear extrapolation of the College’s published 25th and 75th percentile values (•).


     Text Box: Calculation using Candidate’s equivalent mean HS SAT (M+CR) and linear extrapolation of the College’s published 25th and 75th percentile values (•).






Summary:  Existing high school transcripts do not contain enough information to quantitatively predict a candidate’s performance at a specific college.  The easiest improvement is to use the existing grades, but convert the class rank to percentile rank.  A further


improvement would be to retrieve all the students in all the classes taken by the candidate and recalculate the percentile class rank based on the candidate’s actual competitors.  In either event, the basis of the prediction is the use of the mean SAT (M + CR) for the candidate’s high school class rank rather than the candidate’s individual scores.  This change allows evaluation of the candidate in a meaningful context.

The problem:  At first glance, it seems that a student’s high school grades should be a strong and easily evaluated predictor of that student’s college grades. But there are over 25,000 high schools in the U.S., and grades actually vary by school, by year, by teacher, and by course (a group I will call here the grade “cohort”).  There are over 150,000 new cohorts every year, and


while local cohort data is well known to each high school, I cannot find a single instance of cohort data being made public, let alone being provided to colleges in admission applications

It is therefore impossible for a college to quantitatively predict a student’s likely performance, and even a good guess (a process some ironically call “gut-ology”) is practical only with well-established high school/college pairs.  For the same reason, colleges cannot put all their candidates on a single, quantitatively-sound scale.


Standardized tests like the SAT and the ACT provide nationally consistent results, but are relatively poor predictors of college grades.  As is frequently noted, there is a considerable difference between doing well on a multiple choice test Saturday morning and doing well in a nine month long course with many and varied requirements.

In short, the entire college application/admission process relies on information – the high school transcript – which is hard to interpret at best, and frequently misleading.

Some useful tools and patterns:  Academic records are now almost universally digitally stored and readily retrieved for analysis.  “PowerSchool” – or its competitors – can easily retrieve all the data for all the cohorts where a student has received a grade, allowing calculation of a student’s class rank whenever desired, and with an algorithm which excludes arbitrary premiums for “honors” courses.

Patterns of class rank vs. SAT are very stable at both high schools and colleges:  while individual SAT (Math + Critical Reading) scores show considerable scatter, the best-fit SAT score line barely changes from class to class.  The mean SAT (M+CR) of the fiftieth percentile student in a class, for example, is typically within four SAT points of that value for both the preceding and the following class.  The shape of the curve is similarly stable, especially at colleges:  nearly linear over all but the highest and lowest class ranks.

(This averaging technique is precisely analogous to the method almost universally used in quantum mechanics to find easily observed values by averaging over many quantized values:  each student being analogous to a quantum particle like a photon.  This made the calculation trivial in my mind, but totally unfamiliar to anyone outside quantum physics:  about  99.9%  of the population, I would guess.  For 16 years, I used the technique unnoted and unexplained,until I finally realized that real people never use or encounter this method.). 

Therefore the mean equivalent SAT values for a given percentile class rank can be taken from those of the most recently graduated class at that high school.

Further, most American colleges publish the mean SAT scores for their 25th and 75th percentile first year students, and tables of these values are updated every year.  Like the earlier values, these values are typically stable from year to year:  about 200 SAT points apart, with only slowly changing median values.  

By most reports, a typical college application gets about eight minutes of review.  Within those precious 480 seconds, the reviewer must assess both the quantitative information like GPA, and the subjective information like essays and recommendations.  Making the quantitative task sound and simple is surely beneficial to all in a process which literally shapes lives, costs the nation many billions of dollars a year, and has so much good and bad potential.  This is particularly true as both students and colleges look internationally or nationwide rather than just statewide or regionally. 

A proposed solution:  The patterns above allow direct prediction of a candidate’s first year class rank at a college by finding the class rank at that college with the same mean SAT as the candidate’s high school class rank:  see the graph above.  Note that this automatically adjusts for:

1.     The academic aptitude of the candidate’s actual high school competitors and potential college competitors.

2.     The grading patterns of the specific high school cohorts where the candidate earned grades, and,

3.     The grading pattern at the college under consideration.

Voila!  Local murk becomes specific college clarity.

Note that this method avoids the need to change either the overall grading pattern at the school or the many cohort patterns.

It is interesting to compare the tracking of potential new students against the (non-) tracking of graduates.  Districts understandably work hard to determine exactly who will appear the first day of school, and what each student’s program for that year will be.  Parents happily cooperate, since they also want things to go smoothly.

By contrast, after graduation, most Districts have only casual contact with students, with little or no quantitative reporting.  Their college performance is known merely as accumulated anecdotes, ignoring the opportunity to measure their high school preparation by their first year performance in college.  This makes it harder for students at low grading schools like Simsbury High School to gain admission to colleges attended by their performance peers.  And it significantly reduces their chances for scholarships.

DRAFT — For Illustration Only





Valid only if embossed here with







Official Simsbury HS School Seal.



        Susan A. Example

Equivalent Nationally Normed GPA


Projected First Year Class Rank at

        Somewhere College

     66th percentile from bottom


Prepared 22 June 2015.

Please see following chart for the basis of these results.


Hope R. Eternal

Guidance Counselor




Grading Analysis Flowchart

Applicant’s Projected First Year College Class Rank

Flowchart: Process: Local Class Rank of Applicant3 May 2015 VersionFlowchart: Process: Weighting by course credits

Local Course Grades of Applicant

Flowchart: Process: Grades and Class Ranks of all students in those school-year-course-teacher cohortsFlowchart: Process: National SAT v. GPA Curve from College Board data 

Flowchart: Process: SAT v. Local Class Rank Curve – most recently graduated class

SAT Value equivalent  to Local Class Rank

Flowchart: Process: Mean SAT Scores for College’s 25th and 75th percentiles

Applicant’s Equivalent Nationally Normed GPA