What This Episode Covers
The GMAT® scoring algorithm has been debated for decades, and Isaac thinks much of that debate is warranted. The algorithm is widely misunderstood, and the misunderstanding causes pain it shouldn't, partly because a lot of prep gets sold on the back of the confusion. A subscription business has odd incentives here: if you understood the algorithm well, your prep might get a lot shorter. Isaac has spent years working to close that information gap, and this episode is his audio-format attempt at a framework for what the algorithm is designed to measure, built on business analogies you can apply under time pressure. For the visual version, he recommends the free video on the TGS website, and a second watch if the first one didn't fully connect.
The mechanics come first, as a refresher. Scoring works the same on the online GMAT® as it does in person. The exam is adaptive: answer correctly and the next question usually gets harder, miss and it usually gets easier. The algorithm's only inputs are whether you answered correctly, the difficulty level of that question, and a standardized mix of topics that keeps each exam comparable. How long you spent on a question is not part of the calculation, and the next question's topic is not chosen based on the one before it. Your score estimate gets recalculated after every single question, which means your result tracks the difficulty level where you finish, not your raw accuracy.
From there, the episode asks the better question: what would a test like this be designed to measure? The answer is a set of executive skills. Resource management, behaving well under pressure, and the ability to zoom between fine detail and big-picture strategy. All three are learnable, and the exam is a useful place to practice them.
Key Points
The algorithm's inputs are simple. Right or wrong, difficulty level, and the general topic mix. That's the whole data set. There's no AI watching how long you spent on an algebra question and serving you another one to hammer the weakness, and there's no penalty for the clock on any individual question. The topic blend is standardized for comparability, which is why you'll see roughly ten word problems in a 21-question quant section and only one or two probability questions.
Your score tracks difficulty level, not accuracy. The exam calculates a score estimate after every question and keeps updating it until the section ends. A useful simplification: your score is the difficulty level of the final questions you see. Isaac scored in the 99th percentile with 65% of his questions correct, because he was operating at the highest difficulty level of the exam for the entirety of every section. He missed tons of questions. They were just the hardest ones.
Recalibrate what "good" accuracy feels like. Twenty years of schooling train us that 50% is failing. On an adaptive exam, ping-ponging between mediums you never miss and hards you usually do works out to roughly average. Seventy percent accuracy is a strong result, and if you're the achiever type who used to sweat the difference between 95% and 98%, this might be the single biggest mental shift in your prep. Some test-takers never make it.
Chasing early accuracy is a bet on luck. Trying to bank the first eight questions no matter what, then surviving to the end, can work. It's also comically risky, and it's the strategy behind those six, seven, eight-minute questions early in a section that wreck the rest of the exam. The goal is to end at a high difficulty level, not to peak in the first ten questions.
A blank answer costs about twice as much as a wrong one. Leaving a question blank at the end of a section is roughly twice as damaging as answering it incorrectly. Which means the exam rewards a decisive wrong answer over paralysis. If that sounds like a strange thing to reward, think about the executive version: a leader who spends the whole department's budget in the first week of the year made a decision. A leader who can't decide at all is the expensive one.
The first skill: resource management. Isaac's framing treats each section as a portfolio. You're an investor getting pitched on 21 investments in quant, and only about 70% will be winners. Your job is to invest in the ones that can succeed and stop investing in the losers as fast as possible. Spending six minutes on a question you probably weren't going to get right is going into debt: you borrowed time from a future question, and most of the time the bet doesn't justify the loan. Sometimes it does. The skill is telling the difference in the moment.
The second skill: behaving well under pressure. You probably won't feel calm on the exam, and Isaac doesn't think feeling calm is the right target. Performance follows actions, not feelings. If you keep taking the right actions under stress (letting go of the three-minute question, resetting on the next one), you win. In practice, that means finding your breaking point during timed sets: how much time pressure can you apply before execution breaks? Then work at that edge, back off slightly when needed, and condition the right action until it holds.
The third skill: zooming in and out. Sometimes a question needs you down at the word level, rewriting the prompt and double-checking scratch work, because one tiny word in a Data Insights question decides the whole thing. Sometimes it needs the strategic zoom: is this question worth investing in at all? The best executives Isaac has seen do both, deep technical expertise plus the judgment to know when each mode matters.
A practical sequence for your prep. Early on, ignore the algorithm and build skills and habits. Once timed practice starts, apply the debt question to every extra minute, practice letting go (the right action often feels bad, which is exactly why it needs practice), and log bad habits so they can be remade. Expect a mistake to repeat three to six times before it changes. Then keep asking one question all the way through: what's the constraint on my score? If it's speed, most of your study time goes to getting faster. If it's content, fix that first. Working on the right thing at the right time in the right way is the whole game.
Key Takeaways
- Your score follows the difficulty level where you finish, not your accuracy alone. Missing hard questions while protecting mediums is roughly what average looks like on an adaptive exam.
- 70 to 80% accuracy is a healthy target. Isaac's 99th percentile score came with 65% of questions correct.
- Be bulletproof on easy and medium questions. If you never miss them, you keep getting shown the hard questions where high scores are decided.
- A blank at the end of a section hurts about twice as much as a wrong answer. The exam is built to reward a decisive wrong answer over paralysis.
- Run each section like a portfolio. Invest in the gettable questions, cut the losers quickly, and only borrow time when the bet justifies the debt.
- Right action beats feeling calm. You can't always control the feeling on test day; you can condition the behavior.
- Find the constraint. Speed, content, or execution: put most of your study time where the constraint is, and reassess as it moves.