Classic vs Signals vs Crypto: Numerai Tournaments Compared

Historical scope reviewed September 8, 2026. The figures compare the original tournament-specific samples. Payout horizons, score scales and per-round capital differ across tournaments; these charts do not rank atomic committed-capital returns or substitute Classic scores for Signals. See the current metric definitions.

Positive MMC is a coin flip in Classic, a slim edge in Signals, and a near-coin-flip-with-fat-tails in Crypto: 56% of Crypto submissions clear positive MMC versus 48% on Classic, and Crypto’s mean per-round payout/stake (0.37%) is nearly double Classic’s (0.19%) despite a far thinner field. Three tournaments, three risk regimes — and the choice of which to play is mostly about which one your data can survive.

Numerai runs all three concurrently on one platform. Classic (tournament 8) provides free obfuscated features. Signals (tournament 11) requires participants to bring their own financial data. Crypto (tournament 12) targets cryptocurrency tokens and is not traded by Numerai’s hedge fund. This post compares all three on participation, stake, returns, and score difficulty using nmrdash data.

Each tournament feeds Numerai’s meta-model differently. For background on each format, see How Numerai Works; the tournament overview tracks live aggregate stats.

Participation

Classic runs far ahead of the other two tournaments in staked model count.

Staked model count over time for Classic, Signals, and Crypto, with Classic climbing to roughly 7,000 models, crashing below 3,000 in early 2025, then recovering to just over 5,000, while Signals and Crypto each sit under 1,000 — Enlarge chart

Classic climbed from near zero in 2020 to roughly 7,000 staked models by late 2024, crashed below 3,000 in early 2025, and recovered to just over 5,000 within this sample. Signals fell from its peak toward the mid-hundreds; Crypto grew into a similar range. The model participation chart provides newer counts with its own source dates.

Return Distributions

Per-round payouts (payout divided by stake) show how often models win or lose relative to what they put up.

Density histogram of per-round return for Classic, Signals, and Crypto, all centered at zero with Classic the tallest and narrowest and Signals the flattest with the widest shoulders — Enlarge chart

All three distributions center on zero, but the shapes differ. Classic (n≈280,656 round-model observations) is the tallest and narrowest: its zero bin tops density 50, so most rounds land within roughly ±1%. Crypto (n≈8,243) peaks near density 42 at zero and stays tight, with almost nothing outside ±5%. Signals (n≈11,101) is the flattest, peaking near 30 with visibly wider shoulders; its ±2% to ±5% bars are the tallest of the three, so Signals rounds swing harder per round than Classic or Crypto. None of the three shows meaningful mass beyond ±6%. The current staking return distribution uses a separate recent settled sample, with tournaments and staking regimes kept distinct.

Score Difficulty

Median MMC per round shows which environments make positive scores hardest to come by. MMC (Meta Model Contribution) measures how much a model improves Numerai’s ensemble beyond what the meta-model already captures.

Median MMC by round number for each tournament, with Classic hugging zero and Crypto spiking to plus 0.05 and minus 0.02 between rounds 800 and 1,050 — Enlarge chart

Classic’s median MMC hugs zero across the observed history, a side effect of a large, diverse field smoothing per-round noise. Signals runs slightly wider. Crypto swings the most: median MMC spikes above 0.05 and below -0.02 in several rounds between round 800 and 1,050, consistent with a thinner field where a handful of models can move the median.

The asymmetry under the surface matters more than the noise. Of all resolved submissions, 56% of Crypto rounds clear positive MMC against 52% on Signals and just 48% on Classic. Crypto’s wider tails (4× the MMC standard deviation of Classic) come with a directional bias — when the median moves, it tends to move up.

Where the Money Is

Total staked NMR shows the economic weight behind each tournament.

Stacked area chart of total NMR staked by round across Classic, Signals, and Crypto, with Classic holding the clear majority and Signals and Crypto appearing as a growing wedge from about round 700 onward — Enlarge chart

Classic holds the clear majority of staked NMR through the full history. Signals and Crypto appear as a growing wedge on top of the stack from about round 700 onward, together reaching roughly 15–20% of total stake by round 1,200. The stake gap is wider than the participation gap because Classic attracts the largest individual stakers.

Which One Should You Play?

Classic is the lowest-infrastructure route and the most crowded field. The free data and large community lower the barrier to entry, but the shared feature set makes differentiated MMC harder.

Signals is attractive only if your external data survives neutralization. The field is smaller, but the data-sourcing burden is much higher.

Crypto offers a thinner field, but its smaller sample and non-hedge-fund-traded status make historical payouts less comparable to Classic.

Takeaways

Classic leads on participation, stake, and history, but its dense field also compresses scoring edge — the 48% positive-MMC rate is the lowest of the three. Crypto’s 56% positive rate looks attractive until you remember its MMC standard deviation is 4× Classic’s; the average is higher because the field is thinner, not because the alpha is freer. Signals sits in between on both. Diversifying across all three is possible but expensive in both data sourcing and operational overhead.