Platforms to Practice

Online judges that auto-grade your code against hidden tests, so you build pattern recognition and verified correctness instead of guessing whether a solution works.

Why it matters

Reading about two-pointers or dynamic-programming does not transfer to recall under a 45-minute interview clock; deliberate, graded reps do. A judge removes the two excuses that kill self-study — “I think this is right” and “I’ll handle edge cases later” — by running your code against adversarial inputs (empty, max-size, overflow, duplicates) and timing it. The goal is not puzzle count but a small, well-chosen rotation drilled until the underlying pattern is automatic.

How it works

Pick a platform by what you are training for, then drill a curated list rather than browsing randomly:

PlatformBest forGradingFree tier
leetcodeInterview DS&A patternsHidden tests + runtime/mem percentileLarge; some “premium” locked
edabitSyntax fluency, beginnersVisible unit testsFully free
HackerRankRole/skill certs, screensHidden testsFree
CodeforcesCompetitive, contests + ratingFull system tests post-contestFree
Project EulerMath-heavy number-theory-basicsSingle numeric answerFree
  • Match difficulty to a ladder, not ego. Spend most reps one notch below your ceiling so you finish and reinforce; reserve hard problems for stretch sessions.
  • Use editorials deliberately. Time-box (e.g. 30 min); if stuck, read only the idea, close it, and re-implement from a blank file.
  • Track the pattern, not the problem. Tag each solved item by technique (sliding-window, merge-intervals, backtracking) so review surfaces gaps.

Example

A 6-week interview ramp using a judge’s topic filter, ~5 problems/day:

Wk1-2  arrays/strings: two-pointers, sliding-window, prefix sums
Wk3    hashing + stacks/queues; start trees (BFS/DFS)
Wk4    graphs: topo-sort, union-find, shortest-path
Wk5    DP 1D->2D; backtracking
Wk6    mixed mock sets, timed, no editorial

Re-solve every problem you needed a hint on after 7 days; if you still stall, it is a true gap, not a memory lapse.

Pitfalls

  • Grinding count over patterns. 500 random solves teach less than 150 grouped by technique with spaced review.
  • Memorizing solutions passes the judge but fails a follow-up question; always re-implement cold.
  • Ignoring the complexity readout. A green check at the wrong asymptotic-notation still fails real interviews — read the runtime/percentile.
  • Platform mismatch. Codeforces speed-rating habits do not map cleanly to the design-heavy questions some companies ask.

See also