The Mistake Tax

The Mistake Tax

Every submission costs. A wrong-answer verdict costs a re-read, a re-compile, a re-test — 5 to 15 minutes of contest time. Speed training is therefore mostly tax avoidance: habits that keep the first submission the right one.

The four most expensive habits

1. Slow I/O. Every language has the trap. In C++: no ios::sync_with_stdio(false) with heavy cin, or endl (which flushes) in a 200000-line loop. Fast input is not an optimization; at 2·10⁵ lines it is the difference between 0.1s and TLE.
2. Re-computing inside loops. Anything invariant across iterations (prefix sums, sorted copies, hash maps of counts) belongs before the loop. O(n²) pretending to be O(n).
3. Reading input twice. Stream positions do not rewind for free. Read once, store, process.
4. Submitting the plan, not the code. The plan handles n = 1; the code crashes on it. Dry-run the code against every sample by hand before submitting: trace two variables, not the whole program.

The before-submit checklist

• Samples traced by hand — both the given ones and one of yours
• Overflow: the largest two multiplicands multiplied, checked against 64-bit
• Boundaries: empty, single element, all-equal, maximum values
• Complexity written in a comment; it matches the constraints
• I/O format matches exactly — one trailing space has cost medals

What the tax buys

A verified wrong solution in this module's practice set fails exactly the way a rushed submission fails. Read its verdict, name the habit that would have produced it, and the tax becomes tuition instead of loss.