How a subnet's slice of emission is decided
This lesson follows a payout past the point the previous lesson stopped: once an overall share has been decided, what actually happens to it on the way into someone's balance — and how that share can end up funding more than one contest at once.
Every block, a live subnet mints two pools of its own token, alpha, at once: one destined for whoever is building, doing, and judging its work, the other injected straight into its own trading pool, so that paying people out doesn't quietly drain the market that prices it in the first place. What's destined for participants only actually reaches anyone at the end of a judging period, and splits the way an earlier lesson described: a slice to the subnet owner, and the rest split evenly between whoever did the work and whoever judged it. [source] [source]
A slice of what's set aside for validators never reaches them at all: it is reserved for whoever staked TAO into the root network instead of into this one contest's own token, and it only actually accrues as claimable dividends in blocks where the network's contests are, combined, priced highly enough — otherwise, that reserved slice is recycled rather than paid to anyone. [source]
Even within one validator's own payout, backers aren't all paid the same way. Someone who backs with plain TAO is paid through a different mix than someone who backs the same judge with this contest's own alpha — the split follows the same network-wide weighting that already decided how much a TAO-backer's stake counted for judging in the first place — and the judge keeps its own agreed cut before either group is paid. [source]
A subnet's own slice doesn't have to fund one contest, either. Its owner can split it across up to two separate scoring competitions running side by side inside the same one, each judged completely independently of the other — a miner or validator keeps a single registration across both, and simply earns the combined total from whichever of the two it takes part in. [source] [source]
How it actually works
The participant-destined pool is capped at up to 1 alpha per block at a subnet's current halving rate, with an equal amount minted into the trading pool as alpha_in — normally tao_in / price, so the injection doesn't move the price on its own. That injection is itself capped at a root_proportion share of the subnet's alpha issuance; as a subnet matures and its own alpha issuance grows, the cap shrinks, and the TAO that can no longer be injected as liquidity is instead used to buy alpha on the subnet's own market — converting a young subnet's liquidity injections into an older subnet's chain-driven buybacks. [source]
That per-tempo split has two edge cases worth knowing. If a judging period ends with no miner earning any incentive at all, the miner-half of that period's pending alpha goes to validators instead of being withheld from anyone. And a subnet that's been switched off from receiving emission gets a zero share of the network total, redistributed proportionally to whichever subnets are still enabled — its trading-pool injection stops immediately, though alpha already destined for its participants keeps accruing. [source] [source]
The mechanism split itself is stored on-chain as a list of numbers that must add up to a fixed total, and leaving it unset just means an even split. The chain enforces real limits around it too: the number of mechanisms a subnet runs can only change roughly once a day, and the total number of registered participants stays capped network-wide, so running more mechanisms means fewer slots in each one. All of this runs through one function in the chain's own source, split_emissions, whose doc comment states its job plainly: "Split alpha emission in sub-subnet proportions stored in MechanismEmissionSplit." [source] [source]