Loading page…
Loading page…
Guide
Explore, pool pages and one ranking list show figures from the Geographic Decentralisation Index, built and maintained by Tielman Esterhuizen. GDI publishes them; Radar matches them to its pools and shows them with credit. This page explains what they mean and where they stop.
GDI scores how much of a pool's stake sits in countries, cities and network providers that hold little of Solana's stake. Higher means less crowded.
The validator allocation on a pool page shows how its stake is split between validators. It does not show where those validators are. Fifty validators in one city, on one cloud provider, share the same power cut, the same outage and the same regulator. GDI adds that missing view.
It looks at three things for each validator: its country, its city and its network provider, the company whose network connects it to the internet (formally its ASN, or autonomous system number). A place or provider holding a small share of the network’s stake is rare and scores high; a crowded one scores low:
rarity = −ln(share of network stake in that country, city or provider)
The network is Solana’s actively voting validators. A validator’s composite rarity is the geometric mean of its three rarities. A pool’s score on each dimension is the stake-weighted average rarity of its validators, and its GDI is the geometric mean of those three, so a pool spread across countries but concentrated on one provider is pulled down. GDI’s methodology (opens in a new tab) has the full method.
A pool that stakes with several validators gets a pool GDI when GDI scores it. One that stakes with a single validator gets that validator's rarity instead. The two are different measures and are never compared or ranked together.
| Pool | What Radar shows |
|---|---|
| Several validators | Its GDI against the network baseline in the Validator allocation section. Each listed validator also shows its GDI rarity, and its location in the full validator list. |
| One validator | That validator's GDI rarity against the network median. Under GDI details: its location, and a reminder that the pool depends on that one validator. |
| Not covered | Nothing. Index funds and other products Radar does not treat as a stake pool get no GDI. A stake pool GDI publishes no pool score for gets no pool figure rather than a zero, though its listed validators can still show their location and GDI rarity. |
A high rarity for one validator says that its one location is uncommon. It does not say the pool is spread out: all of its stake still depends on a single operator, a single machine room and a single provider. That is why single-validator pools are kept off the GDI list, which only compares pool scores.
Radar decides which kind applies from its own record of the pool’s newest complete validator list. If that record and GDI disagree about whether the pool has one validator or several, Radar shows neither headline figure.
Read a pool's GDI against the network baseline and a validator's rarity against the network median. Both are published by GDI alongside the scores.
| Figure | Meaning |
|---|---|
| Network baseline | The same GDI formula applied to all of Solana's actively voting stake. A pool above it holds its stake in less crowded places than the network does overall. It is stake-weighted. |
| Network median | The middle validator's composite rarity, counted by validators rather than by stake. A validator above it sits somewhere rarer than half of the actively voting validators. |
| Country, city and network provider | The pool's stake-weighted rarity on each dimension. The GDI combines them, so a low one shows where the pool is concentrated. |
| Location known | The share of the pool's stake GDI could place. Stake it could not place is left out of the score, so it neither helps nor hurts. |
| Validators counted by GDI | Every validator GDI found in the pool. Radar's list hides allocations below 0.01% of the pool, and below 5 SOL in pools of 1,000 SOL or more, so it can name fewer. |
The baseline and the median are different reference points. One weights each validator by its stake; the other counts each validator once. Compare a pool with the baseline and a single validator with the median, not the other way round.
The scores are not percentages, grades out of ten or APYs, and the gap between two of them is not a multiplier. They are shown to two decimals, and every comparison Radar makes is judged on those two decimals.
The colours are Radar's judgement, not GDI's. GDI publishes the scores, the network baseline and the median, but no good or bad bands.
| Colour | Pool GDI | Single-validator rarity |
|---|---|---|
| Mint | More than 0.25 above the network baseline: more spread out than the network. | More than 0.25 above the network median: a rarer location than the middle validator’s. |
| Normal text | Within 0.25 of the baseline, either way: close to the network. | Within 0.25 of the median, either way: about as common a location as the middle validator’s. |
| Soft coral | More than 0.25 below the baseline: less spread out than the network. | More than 0.25 below the median: a more common location than the middle validator’s. |
Both use the same band, each against its own reference: a pool’s GDI against the network baseline, a single validator’s rarity against the network median. The same band decides the wording beside a figure and what the Explore filter calls “Higher GDI than the network” or “Lower GDI than the network”. Screen readers hear the same words, so the colour is never the only cue.
Why 0.25: GDI’s methodology (opens in a new tab) notes that per-epoch figures move by several percent as stake and validators shift, so a small gap can flip from one epoch to the next. Radar uses 0.25 as a display band so small differences don’t get too much emphasis. It is Radar’s presentation choice, not a threshold GDI defines, and not a percentage difference in how crowded the places are.
Coral is not a failing grade. It says only that the stake sits in more crowded countries, cities or network providers than the network’s does. It covers geography alone, not yield, performance, operator independence or safety.
Every GDI figure belongs to a Solana epoch. Radar hides it once it is more than two epochs older than the current one.
GDI measures where stake sits. It says nothing about yield, validator performance or the safety of a pool.
Two readings from GDI's epoch 1057 publication, to show the shape of each. They are fixed examples, not current figures.
Example: epoch 1057, a pool
GDI counts 38 validators in definSOL. Its GDI is 4.37 against a network baseline of 2.56: more spread out than the network, so its stake sits in less crowded places than Solana’s overall. Its country, city and network provider scores are 3.93, 4.72 and 4.51, and GDI could place 100% of its stake.
Example: epoch 1057, a single validator
gS stakes with one validator, the Cirro validator, in Tsuen Wan, Hong Kong, on Decentra. Its GDI rarity is 5.30 against a network median of 2.70: a less common location than the middle validator’s. That describes one location. gS still depends entirely on that validator.
Source: Geographic Decentralisation Index (gdindex.app) (opens in a new tab) by Tielman Esterhuizen (opens in a new tab).
Thank you to Tielman Esterhuizen for building and maintaining GDI, for publishing its data openly and its code and methodology under the Apache-2.0 licence. Radar republishes GDI’s figures with credit and does not recalculate them.