Two numbers govern the split, and the direction they point in catches people out constantly.
The cut is what the indexer keeps.
An indexer publishes two parameters on chain:
- Indexing reward cut: the share of indexing rewards the indexer retains.
- Query fee cut: the share of query fees the indexer retains.
The remainder goes to the delegation pool. So a lower cut is better for you, and the documentation states the boundary case plainly: “If an Indexer’s reward cut is set to 100%, as a Delegator, you will get 0 indexing rewards.”
If you take one thing from this lesson, take the direction. A delegator who reads a 90% cut as “I get 90%” has the arithmetic exactly inverted, and will pick the worst available option while feeling clever about it.
Working it through.
Say an indexer earns 10,000 GRT of indexing rewards over a period, and sets an indexing reward cut of 10%.
- The indexer keeps 1,000 GRT.
- 9,000 GRT goes into the delegation pool.
- If you hold 5% of the pool by share, your claim grows by 450 GRT.
Nothing about that arithmetic is subtle. What is subtle is every input to it. “10,000 GRT of rewards over a period” is a function of how much stake the indexer actually allocated, to which subgraphs, and for how long. Two indexers with identical published cuts can return very different amounts because one of them allocates its pool attentively and the other does not.
Try it with your own numbers.
The last two figures are the ones nobody quotes. Capital committed for the period you choose is locked for that period plus the thaw, and earns for only the first part of it. Shorten the holding and the gap widens: the same indexer at the same rate returns materially less to somebody who rotates every quarter than to somebody who leaves it alone. Gross pool APR varies with issuance, the size of the delegation pool and how much of the indexer's stake is actually allocated, so take it from live data rather than from this box.
Why the second pair of figures matters more than the first.
Change the holding period in that calculator and watch the last figure move. That is the 28 days thawing period expressed as a drag on return.
The logic is not complicated. If you commit capital for some number of days and then want it back, it is unavailable for that number of days plus the thaw, and it earns for only the first part. The shorter your intended holding, the larger the proportion of the total commitment that earns nothing.
This is why comparing advertised delegation rates against, say, a lending rate is not comparing like with like. One of them lets you leave on Tuesday.
The pot itself just got smaller.
Indexing rewards come from issuance, targeted at 3% a year. As of 2026-08-31, GIP-0089 redirects 20% of protocol issuance to the Foundation treasury. The Subgraph Service rewards manager now receives 96.584 GRT/block, reduced by 24.146 GRT per block.
The consequence for you is direct and mechanical. Indexing rewards are the larger part of most delegators’ return, and 20% less is arriving at the source. Any indexing rate measured from data collected before 2026-08-31 is describing a larger pot than the one that now exists.
This is not a criticism of the change, which has a stated rationale worth reading. It is a warning about the age of the figures you are looking at.
The capacity multiplier, again.
An indexer’s delegation capacity is 16x their self-stake. Past that ceiling, the excess cannot be used and rewards for everyone in the pool are diluted.
Two indexers, identical cuts, identical diligence. One has ample headroom, the other is over capacity. Your return from the second is worse and nothing in the advertised rate says so. Checking headroom takes about thirty seconds on a dashboard and it is the highest-value thirty seconds in the whole exercise.
The cut is not a promise.
The cut is a parameter the indexer sets on chain. It is not a contract with you, and it is not fixed for the life of your delegation. An operator can change it.
The official delegating page describes how to evaluate an indexer’s parameters. It does not warn you that those parameters can move after you have delegated, which is a genuine gap. Practically, this means:
- Look at an indexer’s parameter history, not a snapshot.
- Treat a cut that is dramatically better than everyone else’s as a question rather than an opportunity.
- Remember that if a cut changes for the worse, your exit still takes 28 days.
Before reading on: two indexers, both 10% cut, both plenty of headroom. What could still make one materially worse?
How much of the pool is actually allocated, and to what.
Stake sitting in the pool unallocated earns no indexing rewards. An indexer with a large pool who has allocated a third of it is generating rewards on a third of your capital, and the cut applies to the rewards that exist, not to the ones that should have.
The subgraphs they allocate to matter too. Allocating to subgraphs with no signal is allocating to subgraphs that earn no indexing rewards, which produces the same result by a different route.
None of this shows up in a cut. All of it shows up in the actual reward history, which is why the next lesson is about looking at that history properly.
An indexer sets an indexing reward cut of 85%. As a delegator you receive:
The cut is what the indexer keeps. 15% goes to the pool, and your claim is your share of that pool. The docs state the boundary case: a 100% cut means delegators get zero. Reading the cut as your share picks the worst option while feeling like the best.
An indexer is above its delegation capacity. What happens to your rewards?
The dilution falls on everyone in the pool, not just the last arrivals. That is what makes it a trap: an indexer that was a good choice when you delegated can become a poor one purely because other people delegated afterwards.
Why is an advertised delegation rate not the return on your committed capital?
The thaw is real time during which the capital is neither available nor earning. Nobody is lying by quoting the earning-period rate, but it does not describe your position, and the shorter your intended holding the bigger the gap.
0 of 3 answered