.png)
The 7sage predictor returns an acceptance, waitlist and rejection probability for every law school from three inputs: LSAT, GPA and underrepresented minority status. 7Sage publishes its model and one internal accuracy figure, and concedes the model cannot see essays, work experience or timing.
The 7Sage predictor returns three probabilities for every law school, being the chance of acceptance, waitlist and rejection, and sorts each school into one of four bands. 7Sage's own product page defines those bands numerically, and the four 7Sage bands are defined by probability rather than by feel. Verified July 2026.
7Sage band7Sage's published definitionSourceSafetyAbove a 70 percent chance of acceptancehttps://7sage.com/admissions/predictorTargetA 35 to 70 percent chance of acceptancehttps://7sage.com/admissions/predictorReachA 5 to 35 percent chance of acceptancehttps://7sage.com/admissions/predictorSuper-reachUnder a 5 percent chance of acceptancehttps://7sage.com/admissions/predictor
7Sage's product page states the instruction plainly: enter your LSAT score and GPA to gauge your chances of acceptance, rejection, or waitlisting at every law school (https://7sage.com/admissions/predictor). Verified July 2026.
The four 7Sage bands are defined by probability rather than by feel. 7Sage publishes safety as above a 70 percent chance of acceptance, target as 35 to 70 percent, reach as 5 to 35 percent, and super-reach as under 5 percent, all on that same page. Verified July 2026.
One inconsistency in 7Sage's own material is worth knowing before you use the output. 7Sage's teaching lessons define reach, target and safety by where your numbers sit against a school's medians rather than by probability, so the same word means two different things in two places on 7Sage's site (https://7sage.com/admissions/predictor). Verified July 2026.
The 7Sage predictor runs on self-reported admissions outcomes rather than on official school records. 7Sage says so directly but does not currently name the database those outcomes come from, and 7Sage's published wording is that the predictor is trained on admissions outcomes reported by real applicants. Verified July 2026.
7Sage's published wording is that the predictor is trained on admissions outcomes reported by real applicants, and that its probabilities reflect how applicants with numbers like yours actually fared (https://7sage.com/admissions/predictor). Verified July 2026.
7Sage also states that the model compares your LSAT score and GPA to each school's most recent medians, which is the layer where published school data enters (https://7sage.com/admissions/predictor). Verified July 2026.
What 7Sage does not publish is the size or provenance of the training set. Its July 2026 rebuild post describes the corpus only as reported outcomes across the three most recent cycles and gives no sample count (https://7sage.com/blog/admissions-predictor-update-same-look-new-engine). Verified July 2026.
7Sage names its method, which is more than most admissions predictors do. The 7Sage predictor is a gradient-boosted decision tree model blended with an ordered logistic model, and 7Sage's rebuild post adds that your numbers are treated as a package rather than a checklist, meaning the model learns how the inputs interact instead of scoring them separately. Verified July 2026.
The exact published wording on 7Sage's product page is that the predictor uses a machine-learning model, gradient-boosted decision trees blended with an ordered logistic model, trained on admissions outcomes reported by real applicants (https://7sage.com/admissions/predictor). Verified July 2026.
Naming the model is a real disclosure and 7Sage deserves credit for it. What 7Sage does not publish is the feature list beyond its three inputs, the blend weight between the two models, the training and holdout split, or any per-school sample size, so the disclosure stops short of anything a reader could audit. Verified July 2026.
The 7Sage predictor sees three things and nothing else: your LSAT score, your LSAC GPA, and a checkbox for underrepresented minority status. 7Sage states this limit itself, and 7Sage's rebuild post says it directly: the model still only sees GPA, LSAT, and URM status. Verified July 2026.
The other controls on the 7Sage page, including rank tier and acceptance-percentage range, filter which schools are displayed. Those controls are not model inputs and changing them does not change any probability (https://7sage.com/admissions/predictor). Verified July 2026.
That matters most for the applicant whose case is not in the numbers. Work experience, application timing, early decision, essays, letters and school-specific priorities are not inputs to the 7Sage predictor, so an applicant with five years of relevant work sees the same output as an identical applicant straight out of college. Verified July 2026.
7Sage publishes exactly one validation figure for the predictor and it is internal, unaudited and relative. No independent measurement of 7Sage predictor accuracy exists anywhere we could find, and read that carefully, because it is a comparison against 7Sage's own previous model rather than an absolute accuracy rate. Verified July 2026.
The figure 7Sage publishes is a log loss, which it describes as a standard measure that punishes confident wrong answers, falling from 1.35 to 0.85 when tested against actual 2025 outcomes (https://7sage.com/blog/admissions-predictor-update-same-look-new-engine). Verified July 2026.
Read that carefully, because it is a comparison against 7Sage's own previous model rather than an absolute accuracy rate. 7Sage publishes no calibration plot, no Brier score, no area under the curve, no per-school breakdown and no description of the holdout set (https://7sage.com/blog/admissions-predictor-update-same-look-new-engine). Verified July 2026.
The most useful number 7Sage published is an admission about the old model. 7Sage states that applicants the old model gave a roughly 90 percent admit chance were actually admitted roughly half the time (https://7sage.com/blog/admissions-predictor-update-same-look-new-engine), which is the clearest published evidence that any admissions predictor can be badly miscalibrated while looking authoritative. Verified July 2026.
No third party has published a measured accuracy or calibration figure for the 7Sage predictor. Two review pages that appear high in search results give no figures at all, and one of them describes inputs 7Sage says the model does not use. Verified July 2026.
7Sage replaced the predictor's engine in July 2026 and published its reasoning. The stated problem was that the old 7Sage model had been trained on a less competitive era, and that warning is the honest part and it is also the structural lesson. Verified July 2026.
7Sage's wording is that the model is now trained on the three most recent cycles including 2025, and that the old version was learning from an era that was meaningfully less competitive (https://7sage.com/blog/admissions-predictor-update-same-look-new-engine). Verified July 2026.
7Sage warned users what to expect from the change: most people will see lower numbers, and the drop will be largest at reach schools and for profiles that were previously getting very high admit odds (https://7sage.com/blog/admissions-predictor-update-same-look-new-engine). Verified July 2026.
That warning is the honest part and it is also the structural lesson. Any predictor trained on completed cycles is backward-looking by construction, so it cannot anticipate an applicant-volume shift and will drift again the next time the market moves. Verified July 2026.
The 7Sage predictor is free and requires no account, which is a genuine point in its favour and worth stating plainly on a page that also names its limits. 7Sage's own answer on the product page is that the predictor is free to use and does not require an account. Verified July 2026.
A free 7Sage account adds the ability to star schools, track applications and follow decisions reported by other applicants, still at no charge (https://7sage.com/admissions/predictor). Verified July 2026.
7Sage's paid products are separate from the predictor. Its self-study tiers are Core at $69 a month, Live at $129 a month and Coach at $299 a month, with a $1 a month rate for LSAC fee-waiver holders, from 7Sage's own pricing page (https://7sage.com/self-study/pricing). Verified July 2026.
The 7Sage predictor breaks in four specific places, and three of the four follow from its design rather than from any error. Naming them is not a criticism of 7Sage, which concedes most of them, and applicants who report outcomes to 7Sage are volunteers, and a volunteer population skews toward the engaged and the successful. Verified July 2026.
First, self-selection. Applicants who report outcomes to 7Sage are volunteers, and a volunteer population skews toward the engaged and the successful (https://7sage.com/admissions/predictor). Verified July 2026.
7Sage confirms the data is self-reported but publishes no weighting for that skew and no bias analysis, so the direction of the error is knowable and its size is not. Verified July 2026.
Second, thin cells. With roughly 200 accredited law schools, three cycles and a corpus 7Sage describes only as reported outcomes (https://7sage.com/blog/admissions-predictor-update-same-look-new-engine), the sample behind any one school at any one score band can be very small, and 7Sage publishes no per-school count. Verified July 2026.
Third, invisible softs. 7Sage concedes that essays, recommendations, work experience and application timing are invisible to the model, which means the predictor is wrong in the same direction every time for anyone whose file is unusual off the numbers. Verified July 2026.
Fourth, a single binary for underrepresented minority status collapses distinct groups and distinct school practices into one flag, and 7Sage staff have previously said in public that they suspected this factor was overweighted. Verified July 2026.
A calibrated population probability is not a statement about you. The 7Sage predictor tells you how a group of applicants with your numbers fared, which is a different question from what will happen to your file, and it also cannot tell you what an offer will cost you. Verified July 2026.
7Sage says this itself, and says it well: treat the output as a well-calibrated starting point for building a school list, not a verdict (https://7sage.com/blog/admissions-predictor-update-same-look-new-engine). Verified July 2026.
It also cannot tell you what an offer will cost you. The 7Sage predictor outputs no scholarship estimate, so it answers the admissions question and leaves the money question untouched, which is the question that decides most real choices. Verified July 2026.
And it cannot price the new federal lending environment. Federal borrowing for professional students is capped at $50,000 a year and $200,000 in aggregate from July 1, 2026 under the Reimagining and Improving Student Education final rule (https://www.federalregister.gov/documents/2026/05/01/2026-08556/reimagining-and-improving-student-education-federal-student-loan-program-final-regulations), a constraint no admissions probability accounts for. Verified July 2026.
Lovare Institut builds the Lovare Diagnostic, an admissions and cost tool, which competes directly with 7Sage. We are telling you that before the comparison rather than after it, because a review that hides its own interest is worth nothing. Verified July 2026.
The practical effect on this page is that we have described 7Sage's methodology from 7Sage's own published material and quoted it, rather than characterising it in our words.
Where we say 7Sage does not publish something, that is a statement about what we could find on its own site in July 2026, and it is checkable. Where we prefer our own approach we say why, in terms you can test against both tools.
The Lovare Diagnostic answers a different question from the 7Sage predictor and the two are better used together than against each other. The 7Sage predictor prices admission odds; the Lovare Diagnostic prices the decision, and the Lovare Diagnostic does not output a probability, deliberately. Verified July 2026.
Question7Sage predictorLovare DiagnosticChance of admissionYes, a published probability and bandNo, by designNamed statistical modelYes, gradient-boosted trees plus ordered logitNo model, published quartiles onlyInputsLSAT, GPA, URM statusLSAT, GPA, cost tolerance, target marketScholarship or cost outputNoneGrant reach and the federal cap gapPriceFree, no accountFree
The Lovare Diagnostic does not output a probability, deliberately. Where the 7Sage predictor gives a single number per school, the Lovare Diagnostic places your LSAT and GPA against each school's published quartiles and then against its published grant reach, so the output is a position rather than a forecast. Verified July 2026.
The Lovare Diagnostic also carries the federal cap arithmetic, because after July 1, 2026 a school whose cost of attendance exceeds $50,000 a year leaves a gap federal lending no longer covers (https://sfs.harvard.edu/changes-federal-student-loans). Verified July 2026.
If you want odds, use the 7Sage predictor and read its own caveats, which are honest. If you want to know which offer you can actually carry, a probability will not answer that and the Lovare Diagnostic is built for it. Verified July 2026.
Every figure on this 7Sage predictor page carries a Verified July 2026 stamp because a dated stamp is the only honest way to publish a number that moves. Each source below was retrieved in July 2026 and is linked in full rather than named vaguely, so nothing here is modelled, averaged or inferred.
What it supportsSource7Sage predictor output, bands, inputs, data source, pricehttps://7sage.com/admissions/predictor7Sage model description, 2026 rebuild, log loss figurehttps://7sage.com/blog/admissions-predictor-update-same-look-new-engine7Sage self-study pricinghttps://7sage.com/self-study/pricingFederal loan caps effective July 1, 2026https://www.federalregister.gov/documents/2026/05/01/2026-08556/reimagining-and-improving-student-education-federal-student-loan-program-final-regulationsCap figures and Grad PLUS phase-outhttps://sfs.harvard.edu/changes-federal-student-loans
7Sage publishes one internal figure, a log loss falling from 1.35 to 0.85 against 2025 outcomes, which compares its new model to its old one rather than giving an absolute accuracy rate. No independent measurement exists. Verified July 2026.
Self-reported admissions outcomes from real applicants across the three most recent cycles, compared against each school's most recent published medians. 7Sage does not name the database or publish the sample size. Verified July 2026.
7Sage states the predictor is free to use and requires no account. Its separate self-study plans are Core $69 a month, Live $129 a month and Coach $299 a month, with $1 a month for LSAC fee-waiver holders. Verified July 2026.
Essays, recommendations, work experience, application timing and early decision. 7Sage states the model sees only LSAT, GPA and underrepresented minority status. Verified July 2026.
Treat it as a population statistic rather than a promise. 7Sage itself reported that applicants its old model gave roughly a 90 percent chance were admitted roughly half the time. Verified July 2026.
A probability is a starting point for a school list. What it will cost you is a different question and it is the one that decides most real choices. The Lovare Diagnostic takes about fifteen minutes and returns your own numbers against published school data rather than a single probability, and it is free.
Written by Ali, Georgetown Law, founder of Lovare Institut.
August 5, 2026
August 5, 2026