Scoreboard / FLOP devnet / alpha
Evidence first.
Then a miner.
Compare miners by the settlements we can see. Open a score. Read its limits. Route with an explanation.
Observed settlement sample
reported rows
20 observed miners · 100 usable settlements in the seven-day window.
A history score. Not uptime or latency.
01 / Read the evidence
A score you can take apart.
Observe
Dated gateway listings and a bounded settlement page. Nominal listings earn no score points.
Filter
Exclude self-trades and nonconserving escrow. Keep usable settlements from the last seven days.
Explain
Weight cooperation and recency by evidence confidence. A partial tape caps confidence at 0.5.
02 / Compare the sample
Miner rankings.
Scores span models. Confidence describes evidence sufficiency. Open a miner for every formula term and stored price observation.
| Rank / miner | Score / 100 | Confidence | Settlements | Listing |
|---|---|---|---|---|
27.93 | 28.1%low | 47 | Listeddevnet-nominal | |
13.39 | 14.1%low | 48 | Not listed | |
3.59 | 4.0%low | 5 | Not listed | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal | |
No data | 0.0%none | 0 | Listeddevnet-nominal |
Rank numbers always refer to settlement score order. Select another order to compare the same evidence.
03 / Decide with limits
Route me.
Discovery only. A price ceiling fails closed when no live per-GFLOP quote exists. Evaluation uses the current time.
A reason for every decision
Set your policy.
See who qualifies.
The router checks listing expiry, settlement evidence, confidence and exclusions. Historical prices never satisfy a live spending limit.
04 / Reproduce the observation
Keep the raw evidence.
pnpm --filter @flop-lab/scoreboard cli rank --file fixtures/snapshot-03.jsonRun from a checkout after pnpm install. This command reads a committed recording. It scores at the current time.
This snapshot used 5 public requests in 1946 ms. That is collection time, not inference latency. The server revalidates every 60 seconds. Failed refreshes retain the last good snapshot and back off. Browser polling pauses in hidden tabs.
This tool does not verify uptime, inference quality, proof validity, finality or miner endpoint reachability. Cooperation and recency can be gamed. Read the model limits.