
If you're figuring out how to prepare for the IMLC Qualification Round, start with a short skills check, not a long list of topics to memorize. Then give yourself separate days to attempt problems, revise the reasoning, and proofread the finished solutions. That separation matters: it's hard to spot a missing assumption when you're still trying to finish the calculation.
Set aside about 20 minutes. Without opening your notes, try three tasks: find the mean of 2, 4, and 9; calculate the fraction of predictions a model gets right if 8 out of 10 are correct; and explain in one sentence why you might test a model on examples it didn't train on. Afterward, check your answers. The mean is 5, and the fraction is 8/10, or 80%. Testing on unseen examples helps you judge how the model performs on data it hasn't already used to learn. Mark each task as confident, shaky, or unfamiliar.
Pick one shaky or unfamiliar skill and use the IMLC learning resources to look up the idea. Then close the explanation and work through a fresh example. If you struggled with means, try 3, 6, 6, and 9. Write down what goes in the numerator and what goes in the denominator before dividing. Getting 6 matters less than being able to explain why you divided by 4.
Pick a short exercise from the training and practice problems. Before solving it, write down what you're given, what you need to find, and what rule or definition might connect the two. If the problem has multiple parts, do this for each part. A setup you can explain gives you something to check even if you can't finish the problem yet.
Choose a practice problem you can start without looking up a method. Work it through to an answer, then go back to the prompt and check that you answered the right question. If it asks which model made fewer mistakes, for example, don't stop after reporting how many predictions each model got right. Make the comparison explicit.
Work independently before consulting a reference. If you stall, shrink the problem: try two data points instead of a whole table, or trace the stated rule for one example. A small case can expose a mistaken interpretation, but it doesn't prove a general claim. If you still can't continue, write down the exact step you couldn't justify.
Return to yesterday's sticking point. Look up the definition or method you were missing, then put the reference away and redo that part on a blank page. If you can only reproduce it with the reference open, make a smaller example and try again. Keep a short note about what changed, such as “I counted all rows instead of only the rows in the group.”
Choose two exercises that ask for different kinds of work, such as a calculation and a written explanation. Set a time limit that fits your schedule. When it ends, mark where your time went: reading, choosing a method, calculating, or checking. Use that information to adjust your next session rather than treating the clock as a score.
The Qualification Round has five problems, labeled A, B, C, D, and E. If you're preparing before the problems are available, use a set of practice problems for this part of the plan. Read the whole set and mark which problems you can start now, which need more thought, and which depend on a concept you need to review. Begin where you can make progress rather than assuming the labels tell you how long each problem will take.
Continue the problems you started and move on when you've captured your reasoning. Leave a specific note at each gap, such as “check whether ties count here,” rather than a vague “fix later.” Keep calculations beside the claims they support. Don't spend the session making one answer look finished while the rest of the set is untouched.
Give each remaining problem or subpart a deliberate first try. Write down the relevant information and test at least one reasonable approach before deciding to set it aside. If two interpretations seem possible, reread the exact wording before doing more arithmetic. End the day by choosing which gaps to tackle first during revision.
Recompute key results from the original data, not from the numbers in your draft. For instance, if 8 out of 10 predictions are correct, the error rate is 2/10, not 8/10. When you find an error, trace it forward: one incorrect count may affect a percentage, a comparison, and the final sentence.
Read each solution as if you hadn't written it. Mark any step where a reader would have to guess why it follows. Add the missing reason, especially when you choose a formula or draw a conclusion from a calculation. If you couldn't finish a problem, distinguish what you've established from what you only suspect. A confident tone is not a substitute for a justified step.
Check the problem labels and subparts one by one. Look for copied numbers, inconsistent notation, and final answers that don't match the calculations above them. Read any sentence containing “always” or “never” especially carefully. Your work may support a claim about the given example, not about every possible example.
Check the Qualification Round submission guidance when preparing your solution file. If you're practicing, export or photograph your work anyway and open the result. Make sure the pages are in order, nothing is cut off, and small symbols are legible. For an actual submission, also check your name, school, city, and date of birth, and leave time to handle an upload problem.
What if I miss a study day?
Combine two short review tasks or drop an extra practice exercise. Keep the time you set aside for revision and proofreading.
Do I need to know machine learning before I start?
No prior experience is needed to start. Use the first day's skills check to pick one idea to practice rather than trying to learn every topic at once.
Can I use this plan before the problems are available?
Yes. Use practice problems to rehearse the full process, including revising solutions and checking the finished file.
About the International Machine Learning Competition: an online machine learning competition for high school students from around the world. Qualification Round 2026: submissions are open until Sunday, 13 December 2026. How the rounds work