Petrozavodsk / OpenCup Hard Mixed 01¶
This pack is a repo-native stand-in for camp-style hard mixed practice: advanced switching, hard implementation, and no guarantee that the first visible route is the cleanest one.
Who This Is For¶
Use this pack if:
- several advanced topics already feel like review
- your main misses now come from switching between very different hard models
- you want a camp-style sitting without pretending the repo is a full Petrozavodsk archive mirror
Entry Gate¶
Try this pack only if these already feel stable:
- Centroid Decomposition
- Persistent Data Structures
- DSU Rollback / Offline Dynamic Connectivity
- Convex Hull Trick / Li Chao Tree
At minimum, these should no longer feel like first exposure:
Pack Shape¶
- Type:
camp-style advanced hard mixed drill - Topic mix:
tree decomposition + versioning + offline dynamic connectivity + affine DP optimization - Problems:
4 - Suggested time:
4.5-6 hours
Topics Under Test¶
- whether you can keep advanced models compact under fatigue
- whether you know when to park one investment slot instead of contaminating the whole session
- whether your retrieval layer is good enough for advanced implementation, not only core CP
Suggested Order¶
| Slot | Problem | Topic | Why it is here |
|---|---|---|---|
| 1 | Ciel the Commander | centroid decomposition |
Structural first slot: the idea is advanced, but the implementation surface is still contained enough to start the pack cleanly. |
| 2 | Range Queries and Copies | persistent segment tree |
Versioned data structure slot: harder implementation discipline without the full timeline complexity of rollback. |
| 3 | Dynamic Connectivity | DSU rollback + offline segmentation |
Camp-style mid-to-late slot where modeling drift and implementation drift interact. |
| 4 | Monster Game II | Li Chao tree |
Optional late investment: sharp signal, but the best play is often to open it only after the earlier structure-heavy slots are stable. |
How To Run It¶
- spend
12-15 minutesreading all4slots and tag each as: shippable with workinvestment but realpark for now- solve in the suggested order unless slot
2feels dramatically cleaner than slot1 - do not open slot
4if slot3still lacks a compact model - after each finished or parked slot, log:
- first model leak
- first implementation leak
- whether the slot should have been parked earlier
- allow at most one short hot-sheet refresh between slots
Allowed Refreshers¶
- Centroid Decomposition hot sheet
- Persistent Data Structures hot sheet
- DSU Rollback hot sheet
- CHT / Li Chao hot sheet
- Build Kit
Debrief¶
Write down:
- the first slot whose model was wrong, not just incomplete
- the first slot whose implementation burden was underestimated
- the first slot that should have been parked earlier
- one exact recovery route for the next session
Go Back Next¶
- centroid structure still fuzzy -> Centroid Decomposition ladder
- versioned updates still noisy -> Persistent Data Structures ladder
- timeline segmentation / rollback still fuzzy -> DSU Rollback ladder
- affine DP signal still weak -> CHT / Li Chao ladder
- the whole pack felt strategically wrong -> Contest Engineering