Wycord Rankings · Methodology
How a resource is scored
Every ranking on Wycord is computed, not curated. A resource's score is a pure function of objective signals the registry already holds — the same formula, the same weights, every time. If you don't trust a rank, you can recompute it by hand from the numbers on the page.
The formula
score = status_multiplier × base_score × 100
base_score = 30% × ecosystem
+ 25% × uptime
+ 25% × freshness
+ 20% × verified
1 · Documentation status — a hard gate
Documentation status is not a point — it's a multiplier. A "best" resource that is currently broken or shut down is a contradiction, so those score zero no matter how many SDKs or how much freshness they carry.
2 · Ecosystem — how many SDKs wrap it
Counted from SDK_FOR edges in the relationship graph: how many client libraries target this resource. Writing an SDK is real engineering effort, so it's a hard-to-fake proxy for adoption. Normalised as min(sdk_count / 3, 1) so 3+ SDKs earns full marks and a single outlier with ten SDKs can't swamp the field.
Weight: 30%
3 · Uptime — measured, not claimed
From the 30-day lookback of the 6-hourly probes stored in BenchmarkResult. The signal is the fraction of decisive probes (verdict other than INCONCLUSIVE — "couldn't tell" is not "down") that returned LIVE.
A resource with no decisive data yet scores a neutral 1.0 for this term — monitoring is still young, and absence of data must not punish a resource that simply hasn't been probed. Weight: 25%
4 · Freshness — trust decays
A resource verified yesterday is more trustworthy than one last checked two years ago. Freshness is an exponential decay with a 90-day half-life: 1.0 the day it's verified, 0.5 after 90 days, 0.25 after 180 days, and so on. Never verified scores 0.
Weight: 25%
5 · Verified badge
A human actually opened and checked this resource — its docs, its status, its licensing. A meaningful trust signal, but kept below the objective terms so editorial judgement can never dominate the arithmetic.
Weight: 20%
What is deliberately NOT in the score
- Featured status — a human-curated placement, never a rank boost. Paid providers get a richer profile, never a better rank.
- GitHub stars / npm downloads — popularity signals are not stored in the registry, and popularity is the easiest signal to fake.
- Pricing — free-text and unreliable. "Free tier + paid" is honest, but it can't be ranked.
The edge case, guaranteed
Because documentation status is a multiplier rather than a point, a broken API with ten SDKs always scores 0 — below a live API with no SDKs at all. A dead API is never a "best" API, and no amount of ecosystem weight can rescue it.