Subnet Hierarchy: Methodology FAQ
Why validator quality transfers across subnets, what the capture ratio measures, and why the subnet APYs in this report are not returns you would earn.
Why does this contradict the advice to pick a specialist validator per subnet?
Because the data does not support that advice. If validator quality were subnet-specific, the best-validator wins would be spread across many operators. Instead, four validators win 91 percent of all subnets and two win three-quarters. The same names that lead on root lead almost everywhere. Picking a specialist per subnet optimizes for a pattern that is not in the data.
Are these subnet APYs real returns I would earn?
No. The subnet APY figures are alpha-denominated: they measure how much more of a subnet's alpha token you earn, not how much value you gain in TAO. Our first report, The Yield Illusion, showed the median subnet alpha staker lost money in TAO last month despite headline APYs above 50 percent, because the token price fell faster than the yield. Treat every APY here as a relative measure between validators, never as a return.
Then what is the report actually measuring?
Validator quality within each subnet. Every validator on a subnet earns the same alpha token and faces the same token price, so the price decay cancels when you compare two validators on the same subnet. The ranking of validators within a subnet, and the capture ratio (best validator yield divided by median), are price-neutral and reliable even though the absolute APY is illusory.
What is the capture ratio?
The best validator's average yield on a subnet divided by the median validator's average yield on that same subnet. A ratio of 1.12 means the best validator earned 12 percent more than the typical one. Because both numbers come from the same subnet and token, the ratio strips out the alpha-price distortion and isolates validator performance. The median across all subnets is 1.12x.
Why are eight subnets flagged as outliers?
On eight subnets the best validator's annualized APY exceeded 200 percent, with one reading above 1,400 percent. These reflect short-lived emission spikes or low-liquidity dislocations that produce nonsense when annualized. We show the validator ranking on those subnets but the magnitude should be ignored, and they are excluded from the summary statistics so they do not distort the averages.
How is this different from the Validator Hierarchy report?
The Validator Hierarchy ranked validators on root staking only, where your stake stays in TAO. This report extends the analysis to all 128 subnets, where staking means holding a subnet's alpha token. The key finding is that the root ranking transfers: the validators that are best on root are best across subnets too. The two reports use the same underlying dataset.
Does the validator choice matter much if the subnet token is losing value?
It matters at the margin, not at the core. The dominant driver of your outcome when you stake alpha is whether the subnet's token holds value, which mostly it does not. Validator choice is a distant third behind that and your timing. The 12 percent capture edge is real and free to capture, but it will not rescue a losing alpha position. Choose your validator well, but spend your real attention on whether to be in the subnet at all.
Why is tao.bot prominent in this report?
tao.bot is the best validator on 30 subnets, second only to 1T1B.AI. FlowSniper delegates to tao.bot by default, so we disclose it directly. We selected them before any of this analysis, on publicly observable characteristics, and the data has repeatedly confirmed that choice rather than been shaped by it. Had the data ranked them poorly, we would have published it and switched.
Can I reproduce these numbers?
Yes. All figures derive from the Taostats validator yield history endpoint across 25 validators, 129 subnets including root, and 365 days, the same 860,642-snapshot dataset as the Validator Hierarchy report. The per-subnet table, capture ratios, and transfer statistics are all computed from that public data.
Will this be updated?
Yes, quarterly, alongside the root Validator Hierarchy. A recurring measurement shows whether the concentration persists and whether any validators break into the top tier over time.
Still have a question?
If something here is unclear or you believe a figure is wrong, the full dataset and analysis code are retained for verification. Reach us in the FlowSniper Discord.