Grinding 500 LeetCode problems is the slow way to prep. The fast way is pattern recognition — because most interview questions are a remix of ~15 underlying patterns. Once you can name the pattern in the first 30 seconds, you stop solving from scratch and start adapting a template you already know.

Here are the 15 that, in my experience, cover the vast majority of what actually shows up. For each: the signal that tells you to reach for it, and a representative problem.

The 15 patterns

1. Sliding Window — Signal: a contiguous subarray/substring, "longest/shortest/at most K." Ex: Longest Substring Without Repeating Characters.

2. Two Pointers — Signal: a sorted array, or pairs/triplets that sum to a target. Ex: 3Sum, Container With Most Water.

3. Fast & Slow Pointers — Signal: cycles, or finding the middle of a linked list in one pass. Ex: Linked List Cycle, Happy Number.

4. Merge Intervals — Signal: overlapping intervals, scheduling, "merge/insert." Ex: Merge Intervals, Meeting Rooms II.

5. Cyclic Sort — Signal: an array of numbers in a known range [1..n], find missing/duplicate. Ex: Find All Numbers Disappeared in an Array.

6. In-place Linked List Reversal — Signal: reverse a list (or a sublist) with O(1) space. Ex: Reverse Linked List II.

7. BFS (Tree/Graph) — Signal: shortest path on an unweighted graph, or level-order traversal. Ex: Binary Tree Level Order Traversal, Word Ladder.

8. DFS (Tree/Graph) — Signal: explore all paths, connected components, backtracking-ish traversal. Ex: Number of Islands, Path Sum.

9. Two Heaps — Signal: you need the median of a stream, or to balance "smaller half / larger half." Ex: Find Median from Data Stream.

10. Subsets / Backtracking — Signal: generate all combinations/permutations/subsets. Ex: Subsets, Permutations, Combination Sum.

11. Modified Binary Search — Signal: a sorted (or rotated-sorted) input, or "minimize the maximum." Ex: Search in Rotated Sorted Array, Koko Eating Bananas.

12. Top K Elements — Signal: "K largest/smallest/most frequent." Reach for a heap. Ex: Kth Largest Element, Top K Frequent Elements.

13. K-way Merge — Signal: merge K sorted lists/arrays. Ex: Merge k Sorted Lists.

14. Dynamic Programming — Signal: "count the ways," "min/max cost," or overlapping subproblems. Start with the recurrence. Ex: Coin Change, Longest Common Subsequence, House Robber.

15. Topological Sort — Signal: ordering with dependencies, a DAG, "can you finish?" Ex: Course Schedule.

How to actually study these (so they stick)

1. Learn the signal, not just the solution. For each pattern, write down the words in the prompt that trigger it. That recognition is the whole game.
2. Do 3–5 problems per pattern, not 50 random ones. Depth on one pattern beats breadth across many.
3. Re-derive the template from memory. If you can rebuild the sliding-window skeleton on a blank page, you own it. If you're copying, you don't yet.
4. Mix patterns once each is solid. Real interviews don't tell you the pattern — practice cold recognition.

The goal isn't to memorize 500 solutions. It's to walk in able to say, "this is a top-K problem," and reach for the heap before you've even finished reading.

I wrote the full breakdown — with template code for every pattern, Big-O, and the exact signal-words that map a prompt to a pattern — here: The 15 LeetCode Patterns That Cover 90% of Coding Interviews.

What pattern trips you up the most? For me it was two-heaps until it suddenly clicked. Curious what yours is.

---

Disclosure: I work on CoPilot Interview, an interview-prep tool — but this pattern list stands entirely on its own, and the linked guide is free to read.