What you would earn here
EstimateHow it ranks against the network
-Live pool data
Full scoring breakdown
All metricsWho else is in this pool
Fee & pledge history
-Governance
-Relay infrastructure
-Charts
-Cumulative blocks - actual vs statistically expected
Rolling 10-epoch luck
Blocks per epoch
Net return per epoch
Active stake per epoch
Delegators per epoch
Compare with another pool
Side by sideEpoch history
-| Epoch | Blocks | Expected | Luck | Active Stake | Delegators | Rewards | Fees | Net ROS | Tax |
|---|
Lifetime return is a meaningless number. Here is why.
Cardano pays staking rewards out of a reserve that releases a fixed fraction of what remains in it each epoch. The reserve was far fuller in 2020 and 2021 than it is now, so the same fraction was a much larger amount of ADA. Pools operating then paid their delegators 4 to 5 per cent a year. That money has been paid out. It cannot be earned again by anybody.
A pool's lifetime return averages that era together with today's. The result is that a pool's lifetime figure mostly measures how old the pool is. It is not a measure of how well the pool is run, and it is not an offer. This pool's own numbers make the point: GNP1 shows a lifetime return of 2.10 per cent and a recent return of 1.62 per cent. The gap is not skill. It is the calendar.
Do not compare lifetime figures between pools of different ages. Doing so ranks pools by launch date. Any pool advertising a lifetime return much above 2 per cent is showing you its history rather than telling you what you would earn, whether or not it intends to mislead you. The network average today is around 1.5 per cent and every honest pool is close to it, because the reward formula leaves operators very little room to differ.
The reserve continues to deplete. Returns across the entire network will keep drifting downward regardless of which pool you choose. Nobody can tell you what staking will pay in five years, and this tool does not pretend to. Any figure you see anywhere projecting staking returns over multiple years is a forecast about monetary policy dressed up as a statistic about a pool.
Lifetime figures are shown on this page only as historical context, deliberately de-emphasised. Every forward-looking number here uses the pool's return over the last 40 epochs instead.
Before you use these scores
These results may contain mistakes. This tool is maintained by one person. If a figure here looks wrong, or disagrees with an independent explorer, treat the explorer as authoritative and tell us via the chat widget on this page. We will look into it and correct it.
This tool is for shortlisting, not for deciding. Its job is to reduce a list of three thousand pools to a handful worth investigating properly. It is not a verdict on any operator and it is not financial advice. Every number here describes the past. None of it predicts the future, and none of it is a promise.
This tool is not neutral, and here is the conflict
This tool is built and hosted by the operator of GNP1, a Cardano stake pool that is ranked by it alongside every other pool. That is a real conflict of interest and you should treat it as one. The scoring rules were written before any pool was scored against them, and all of them are described below so you can judge whether they happen to favour the author's pool. You should assume this tool is biased in GNP1's favour until you have checked otherwise, and you should verify any pool you are seriously considering against an independent explorer: cexplorer.io, cardanoscan.io, adastat.net and Koios - the last an API rather than a browsable explorer, for anyone who wants to check the underlying per-epoch figures directly. If you find a rule that looks self-serving, please say so.
What a high score does not mean
The score measures what a pool has done on chain: block production, fee and pledge changes, delegator movement, relay setup, governance participation. It cannot see anything else. It cannot measure whether an operator is competent, honest, responsive, solvent, or still interested. It cannot see the quality of their infrastructure, whether their keys are secure, whether they will still be running the pool next year, or whether they intend to raise fees next month.
A high score means a pool has behaved well so far. It is not protection against an operator behaving differently tomorrow. Fees can be raised at any time with no notice. Pledge can be withdrawn. A pool can retire. None of that is predictable from history, and a pool with a spotless record is exactly what a pool looks like immediately before it changes.
How much to trust the number
Treat the headline score as accurate to roughly plus or minus five points, and wider than that for small pools. A pool scoring 78 is not meaningfully better than one scoring 74; those two are the same result. Differences of ten points or more are worth attention. Anything smaller is noise.
The score is a weighted opinion, not a measurement. The metrics, the weightings between them and the bands that turn a measurement into a score are all GNP1's own judgements about what makes a pool good for a delegator - not a Cardano standard, and not the only reasonable way to measure any of it. Another tool applying different choices to the same on-chain data would rank pools differently, and would not be wrong to.
Where the numbers are weak
- Lifetime return figures are historical only. See the explanation at the top of this section. They are shown for context and should not be used to compare pools.
- Recent luck is mostly noise for small pools. A pool expecting sixteen blocks over the scoring window swings about twenty-five per cent either side through pure chance. For small pools this component adds or removes a point at random, for reasons that have nothing to do with the operator. Good luck is currently penalised the same as bad luck, which is wrong and is on the list to fix.
- Luck figures read slightly low. Expected block counts are derived from a theoretical slot figure that assumes pools produced every block since Shelley. They did not: federated nodes produced a share of blocks before March 2021, and one to three per cent of blocks are always lost to propagation battles. Every pool's lifetime luck therefore reads a little under its true value, and pools that were large before March 2021 are understated most. Expect these figures to differ from other explorers.
- Growth metrics are correlated. Stake trend, block trend and delegator trend all respond to the same events. One large delegator leaving moves all three at once and can cost most of a ten-point category for a single event that another category has already priced. This category behaves closer to one signal than to three.
- "Wallets" means stake addresses. Concentration, retention and delegator breadth count stake addresses, not people. One delegator split across twenty addresses looks like twenty delegators and hides their concentration. Every on-chain tool has this limitation, including this one, and there is no way around it.
- Concentration counts the operator's own pledge as a whale. The top-wallet figure includes the pool's own pledge address, which is the stake least likely to ever leave. Pools with a high pledge are rewarded for it in one category and penalised for it here. This is a known inconsistency and is on the list to fix.
- Pledge is a statement, not a lock. A declared pledge can be withdrawn by the operator at any time. Pools declaring a trivial or zero pledge get no credit for meeting it, because the check is arithmetically meaningless in that case.
- Fee history is judged against the margin the pool held while it accumulated its stake, not the margin on its first registration certificate. Registering with a high placeholder, cutting it to recruit delegators, then raising it later is a known pattern that a first-certificate comparison fails to catch. This does not stop a pool raising its fee the day after you read this page.
- Multi-pool detection is best effort and beatable. Three independent signals are used and a confidence level is given. An operator running several pools under entirely separate keys, metadata and relays cannot be detected by any on-chain method. The absence of a multi-pool flag is not evidence of a single-pool operator.
- The "Delegators" figure shown for a pool comes from Blockfrost and runs a few per cent higher than other sources count it, so a small discrepancy against an explorer is expected rather than a sign of a fault. The delegator counts used for peer ranking and network context come from our own db-sync snapshot instead, so those two numbers are not directly comparable.
- Missing data scores neutrally. Where a metric cannot be computed it neither adds nor removes points and the detail text says so. A pool with several unscored metrics has a score resting on less evidence than one with none, and the headline number does not show you that.
What the reward estimate assumes
The estimate is the pool's net return over its last forty epochs, after margin and fixed fee, stated as an annual rate. It assumes nothing changes: not the pool's size, not its fee, not the network's parameters, not the reserve. All of those will change. Rewards are paid in ADA and this tool says nothing whatsoever about what ADA will be worth. A positive return in ADA can be a substantial loss in any currency you actually spend.
Pools with fewer than twenty epochs of reward history are estimated at the network average instead, because their own history is too short to carry information. Rewards also do not begin immediately: expect roughly fifteen to twenty days between delegating and the first payment.
No projection beyond one year is offered anywhere on this page, and you should be suspicious of anywhere that does offer one.
Where the data comes from
Chain data comes from a locally maintained cardano-db-sync instance. Live pool metadata and current stake come from Blockfrost. Peer-ranking, percentile and network-context figures come from our own db-sync, refreshed hourly, and are as of the last completed epoch snapshot rather than live. If that refresh stops, the peer ranking suspends itself and says so rather than showing stale ranks.
If Blockfrost is unavailable, a banner appears and the analysis proceeds on chain data alone: live figures come from the local database, and every metric that needs the live source is set to its midpoint and labelled as scored neutrally. A pool that genuinely does not exist still returns an error rather than a degraded analysis, and an error during an outage names the outage rather than claiming the pool was not found.
Cases this tool handles badly
Retired pools, delisted pools, pools with IPv6-only relays, and pools in unusual registration states have not been tested. In those cases this tool is more likely to produce a confident wrong number than an error message. If a result looks wrong to you, it may well be wrong. Please report it rather than assuming the tool knows something you do not.
What this tool has not had
No independent third-party audit. No formal testing beyond the author's own review and a code audit. No warranty of any kind. It is a hobby project by one person, published because it seemed more useful shared than not.