How Do LeetCode Contests and Ratings Work?

LeetCode runs two rated contests: a Weekly Contest every Sunday and a Biweekly Contest every other Saturday. Each contest has 4 problems and lasts 90 minutes. Your contest rating is an Elo-style number, which means a score that predicts how you will place against a known field. It starts at 1500 and moves up or down after every contest you enter.

The rest of this page gives the start times in three time zones, the scoring rules, what a good rating looks like, and how to use contests for interview preparation. If you came here to learn what the platform itself is, that question is answered in What is LeetCode for?.

Contest times in China, India, and the United States

LeetCode schedules both contests in China Standard Time. Every other zone is a conversion from that. The Weekly Contest starts at 10:30 AM on Sunday, China time. The Biweekly Contest starts at 10:30 PM on Saturday, China time.

China and India do not observe daylight saving, so those two columns never change. The US Pacific zone shifts by one hour twice a year. That is why the Pacific column carries two values. In Coordinated Universal Time, the Weekly Contest starts at 2:30 AM on Sunday and the Biweekly Contest at 2:30 PM on Saturday.

How scoring and ranking work

Each contest has 4 problems ordered by difficulty. Each problem is worth more points than the one before it. The point value is printed next to each problem, and you may attempt them in any order.

Your rank is decided by total points first. Among players with equal points, the tie is broken by finish time. Finish time is the time of your last accepted submission, counted from the start of the contest, plus penalty time. Each wrong submission on a problem you later solve adds 5 minutes of penalty. A wrong submission on a problem you never solve adds nothing.

The practical consequence is that care beats speed more often than new players expect. Suppose two players each solve 3 problems, and one of them made four wrong attempts along the way. The careful player ranks higher, because 20 minutes of penalty is a large gap on a 90-minute clock.

How the rating is calculated

Everyone starts at 1500. After each contest, LeetCode compares your actual rank with the rank its formula predicted from your rating and the ratings of everyone else who entered. Finishing above your predicted rank raises your rating. Finishing below it lowers your rating.

The size of each move shrinks as your contest count grows. Your first few contests move the number the most. A player with dozens of contests behind them usually moves by a few tens of points at a time. For most players the number settles after roughly 15 to 20 contests, and from then on it tracks real changes in skill.

Only a live rated contest counts. Virtual contests, which let you replay an old contest against its original scoreboard, never touch your rating.

What a good LeetCode rating is

LeetCode awards two badges from contest rating. Both are drawn only from players rated 1600 or higher. A Guardian badge goes to the top 5 percent of that group, and a Knight badge goes to the top 25 percent. The cutoffs move a little every week as the pool changes. Knight has generally required a rating in the mid 1800s, and Guardian a rating above roughly 2100.

Those bands are approximate. The global leaderboard rank is a different number from the rating. What counts as a good rank is covered in What is a good rank on LeetCode?. Codeforces uses its own rating scale, so a LeetCode number does not convert directly. The two sites are compared in Which is harder, CodeForces or LeetCode?.

Why a rating is not required for interviews

No company asks for your contest rating. Interviewers see the code you write and explain in a 45-minute session, not the badge on your profile. Interview problems usually sit at the level of the second or third contest problem. An interview also rewards clear explanation and clean code as much as it rewards speed.

The overlap is the skill underneath. Contest problems and interview problems draw on the same 20 or so patterns, such as two pointers, sliding window, and breadth-first search. Contest practice therefore does raise interview performance.