Numerai Staking Returns in USD Terms, 2020–2026
Numerai stakers netted about $35M at receipt prices since 2020. In USD terms staking quintupled a flat token, but the NMR leg still sets the swings.
Numerai pays you in NMR, and almost every yield number published about the tournament, including on this site, stops there. Convert the whole ledger to dollars and the trade looks different: since 2020 the Classic pool has netted stakers about $35M at the prices prevailing when payouts landed, $100 staked at the pool average in January 2020 is worth roughly $587 today against $129 for just holding the token, and yet in most individual years the NMR price moved your dollar outcome more than the staking yield did. Staking Numerai is a two-legged trade, and the leg nobody underwrites is the bigger one.
Two legs, one trade
Split each year's dollar outcome into its two components, the staking yield earned in NMR terms and the token's move against the dollar, and the asymmetry is immediate.
The pool-level yield has been positive all seven years, fading from around +100% in 2020–21 to +11% so far in 2026. The token was down in four of the seven, as deep as −61% in 2022.
Multiply the legs and 2022 was a negative dollar year despite a +48% staking yield, and 2025 nearly repeated it. The yield leg has never been the reason a staker's USD account fell — the token leg has done it four times.
The $35M ledger is front-loaded
Valuing every round's net payout at the NMR price on its resolution date puts a dollar figure on what the tournament has actually paid out, and dates it.
Of the ~$35M total, more than half ($19.3M) landed in 2020 and 2021, when double-digit yields met a $30–60 token.
The trickle since is stark: $1.4M in 2025, and $0.6M from the rounds resolved so far in 2026. Both compressions documented elsewhere on this site multiply here: per-round yield fell 95% as the payout factor slid, and the token gave back most of its 2021 price. Neither force alone explains a 96% decline in dollars paid per year; together they do.
$100 staked versus $100 held
The full-period compounding view is the strongest argument staking has. The same NMR position with pool-average payouts layered on top ends at $587 instead of $129, a 4.6× improvement on buy-and-hold, and the gap only ever widens because the yield leg never printed a negative year.
The path tests conviction, though: the staked line touched nearly $2,000 in early 2021 and again in 2024, meaning today's $587 sits roughly 70% below its own peak — the pool-level version of the drawdowns individual models endure. And this series is the pool average. The median model earns a fraction of the pool's yield (the gap is the subject of the annualized-yield note), so the typical staker's USD curve sits between the two lines, closer to the gray one.
The forward-looking wrinkle is that the double bet is about to become optional. The v3 staking roadmap includes USDC staking at a lower payout factor: the first chance to hold the yield leg without the token leg. The pricing of that trade-off (how much payout factor you surrender to shed NMR exposure) will reveal what the token leg has been worth all along, and I expect it to become the single most important number in Numerai's staking economics. Until it ships, every NMR yield you see quoted, including mine, is the smaller half of the trade.
Method notes: Classic tournament, resolved rounds closing 2020-01 through mid-2026 as of the 2026-07-20 ingest. Receipt price is the CoinGecko daily NMR/USD price nearest each round's resolution timestamp (close + 30 days where missing); USD figures are notional at those prices, not realized sales. Annual yield is the additive sum of per-round (earned − |burned|) / at-stake on a constant-stake basis, i.e. no compounding, matching how one stake collateralizes overlapping rounds; the $100 curves apply the same additive units to the daily price. Pool-average outcomes overstate the median staker's. Signals and Crypto excluded.