A validator charging 10% with perfect performance pays you more than one charging 2% that misses attestations. Commission is a headline, not a return.
Dario FennDeFi & Markets Lead · DeFi protocols, yield, market structure and on-chain data
Delegator returns depend on the network's issuance rate, the validator's effectiveness at capturing rewards, whether MEV or tips are shared, and the commission taken. Compare validators on performance over at least ninety days, commission stability, slashing history, client and infrastructure diversity, and how large a share of stake the operator already controls. A zero-commission validator with poor uptime or an unannounced future fee is not the cheapest option.
Delegation interfaces sort validators by commission, which teaches everyone to optimise the smallest variable. The spread between a good operator and a mediocre one on performance is routinely larger than the entire commission difference, and the spread on tail risk — slashing, an operator going dark — is larger still.
Four things stack up. The protocol's issuance rate, which is set by the network and identical for everyone. The validator's effectiveness — the proportion of the duties it was assigned that it performed correctly and on time. Additional revenue such as priority tips and MEV, and whether the operator passes it through. Then commission, subtracted last.
Effectiveness is where the money is. On Ethereum, missed attestations and late inclusion cost a small amount continuously; over a year, the difference between a 99.9% effective operator and a 97% one exceeds several percentage points of commission. On networks where a missed block proposal forfeits a large one-off reward, the variance is worse.
MEV policy is the second underrated factor. Two operators charging the same commission can pay materially different amounts depending on whether they run MEV-boost or equivalent and how they split the proceeds. Ask, or read the operator's published policy; the good ones state it plainly.
Use a window of at least ninety days. A thirty-day view flatters operators who have recently fixed problems and punishes ones who had a single bad day, and most explorers default to a short window.
Look for the pattern rather than the number. Consistent 99.5%+ with no gaps is an operator with monitoring and redundancy. A flat line with occasional multi-hour holes suggests a single machine and no failover. Sudden improvement six weeks ago usually means new infrastructure, which is good news that has not yet been tested by an incident.
Check slashing history explicitly. Slashing is rare and almost always caused by an operator running the same keys on two machines — a redundancy mistake, not bad luck. An operator who has been slashed once and published an honest post-mortem is arguably safer afterwards than one who has never been tested; an operator who has been slashed and said nothing is not.
Zero commission is a customer acquisition cost. It is legitimate when an operator is building a stake base and says so, and a trap when the fee rises quietly once you are delegated. Check whether the network requires notice of a commission change and whether the operator has a history of raising it.
The more useful signal is stability. An operator who has charged the same rate for two years is telling you their business works at that rate. Very high commission needs a justification — usually institutional service, reporting, or a coverage arrangement — and the professional operators covered in our staking ratings generally do provide one. Figment and its peers price above the retail median and back it with SOC 2 audits, published performance data and contractual terms, which is a different product from an anonymous validator with a low rate.
Two selection criteria serve the network and, indirectly, your own holdings. Do not delegate to the largest operator on the list: concentration of stake is a liveness and censorship risk for the chain, and a correlated-penalty risk for you on networks that punish mass failures more severely than isolated ones.
And prefer operators who publish which client software they run and deliberately use a minority client. A consensus bug in a client running two-thirds of the network is the scenario that turns an outage into an inactivity leak, and diversity is the only defence.
Before delegating, know the unbonding period, whether rewards compound automatically or need claiming, and whether redelegation is restricted. Exiting is not instant on most proof-of-stake networks — Ethereum's queue is covered in withdrawal queues — and an operator you cannot leave quickly is a longer commitment than the interface implies.
If you want the ability to exit immediately, the answer is not a better validator; it is a liquid staking token, which converts an exit queue into a market sale at whatever price the market offers.
Ninety-day effectiveness above 99%. No unexplained slashing. Commission unchanged for at least a year, with a published MEV policy. Named operator with real infrastructure disclosure. Not in the top handful by stake share. Minority client, or at least a stated client. If an operator satisfies those, the remaining difference between candidates is small enough that splitting across two or three of them is a better use of attention than optimising further.
A plain-English walkthrough of proof-of-stake staking — what you're actually doing, how rewards are calculated, and the risks worth knowing before you lock up a single coin.
The fee comparison everyone quotes is the wrong one. What separates these two is what you give up, and the answer changes with how much you stake.
How delegators can lose principal to slashing and downtime penalties, well before any price volatility comes into the picture.