Tech Quality
Stack modernity, structure, tests, docs, and commit activity. Is there real code, or a template clone.
Source: Claude reads the code, manifests, and CI config.
Methodology
How scoring works
RepoRank does not rate your code in isolation. It rates your project against the entire Solana hackathon corpus. Paste a repo, and we score it 0 to 100 across six dimensions, then place the overall number in one of four bands. Every number carries evidence you can check.
The six dimensions
The overall score is a weighted blend of these. We always show the breakdown. A single number invites arguments, the breakdown invites scrutiny.
Originality
Whether the idea fills a gap the corpus has not seen. Full gap, partial gap, or false gap with names attached.
Source: Copilot gap classification against the project corpus.
Competition
How crowded the lane is. The count and strength of similar projects, and how many are still active.
Source: Copilot similarity search over 5,400+ projects.
Winner Similarity
Overlap with the traits, stack, and tags of projects that actually won their tracks.
Source: Copilot analysis run on the winners-only cohort.
Ecosystem Fit
Track and cluster match, target-user clarity, and primitive usage measured against corpus norms.
Source: Copilot cluster match plus Claude reading the README.
Momentum
Commit cadence, contributor count, and recency. No model needed, this is metadata.
Source: GitHub repository metadata.
The four bands
The overall score maps to a band. Colors are the only place we use the band palette. They mark scores, never buttons.
0 to 39Outmatched
The lane is saturated or the repo shows too little original work to grade. Pick a real gap before spending more time here.
40 to 69Contender
A viable entry with soft spots. The idea has room, but positioning or execution trails the projects that place in tracks.
70 to 89Arena Ready
Above cohort median on most dimensions. A credible submission. The gap to a win is usually positioning, not engineering.
90 to 100Laurel
Matches the profile that judges reward. Genuine gap, winner-grade engineering, clear ecosystem fit. Ship before the primitive becomes table stakes.
Where the data comes from
The corpus is the Colosseum Copilot dataset: more than 5,400 Solana hackathon projects across Renaissance, Radar, Breakout, Cypherpunk, and Frontier, plus over 84,000 archive documents covering pitches, submissions, and results. When we say a lane is crowded or a mechanic is a gap, we mean it against that record, not against a hunch.
Live repository data comes straight from GitHub: README, manifests, languages, commits, contributors, and age. Claude reads the code. Copilot compares it to the corpus. GitHub supplies the raw activity.
An honest note on scores
Scores are estimates. They are our best read of how a project sits in the corpus at the moment it was scanned, not a verdict from the judges you will actually face. That is why every dimension ships with the evidence that moved it: real project names, hackathon results, and links. A low score on a crowded lane is a signal to differentiate, not a sentence. Rescan on new commits and the number moves with your work.