Contest Strategy
Scan everything, solve in profit order, bank partial credit, never gamble on an unverified guess.
The first 10 minutes
Read EVERY problem before solving any. Rank them by (expected solution time, confidence). Start with the highest confidence-per-minute — usually the easiest A-problem, but not always: sometimes your strongest topic appears in problem C.
During the contest
- One problem at a time; set a mental timer. Stuck 20+ minutes with no new idea? Move on. Coming back fresh often cracks it in 5.
- Submit brute force when the subtask allows it. Points on the board change your strategy for the better.
- Never submit untested code — 30 seconds of local checking on the samples beats a 20-minute penalty queue (in judges without penalty, it still saves your confidence).
The last 20 minutes
Protect what you have. A risky rewrite that breaks a passing problem is a net loss. Fix compile warnings, re-read output formats, check the largest inputs you can construct by hand.
After
Every contest ends with a post-mortem: which problems did others solve that you missed? Which of your misses was a knowledge gap vs a time-management gap vs an implementation slip? The three have different fixes: study, drills, and checklists respectively.