Skip to main content

Tournament Format

SignalNet runs a continuous, overlapping tournament where contributors submit stock predictions and earn rewards based on accuracy.

Structure​

  • Frequency: New round opens every Monday
  • Universe: S&P 500 (~503 stocks)
  • Horizon: 20 trading days
  • Overlapping: ~4 rounds in-flight at any time
Week 1:  Round 47 opens → Submit by Friday
Week 2:  Round 48 opens → Round 47 resolving...
Week 3:  Round 49 opens → Round 47 still resolving...
Week 4:  Round 50 opens → Round 47 resolves! Results published.

What You Submit​

Rankings — a continuous score for each stock from 0.0 to 1.0:

  • 1.0 = most bullish (you think it will outperform)
  • 0.0 = most bearish (you think it will underperform)
  • Must cover all stocks in the universe
  • No ties allowed

CSV Format​

ticker,prediction
AAPL,0.85
MSFT,0.72
GOOGL,0.63
TSLA,0.91
AMZN,0.45
...

Scoring​

Your submission is scored on three metrics after the 20-day resolution period:

MetricWeightWhat It Measures
IC50%Spearman rank correlation of your predictions vs. actual 20-day returns
TC30%True Contribution — your signal's unique value to the meta-model
MMC20%Meta-Model Contribution — how much the ensemble improves with your signal

Final Score = 0.5 × IC + 0.3 × TC + 0.2 × MMC

Why Three Metrics?​

  • IC rewards raw predictive accuracy
  • TC rewards signals that are different from the crowd
  • MMC rewards signals that help the ensemble

A contributor who copies the meta-model would score high on IC but zero on TC and MMC. Unique, accurate signals earn the most.

Staking​

Stake $SIGNAL tokens to back your conviction:

  • Minimum: 100 SIGNAL per round
  • Maximum: 10,000 SIGNAL per round
  • Payout formula: Score × Stake × 0.25
  • Max loss per round: 25% of your stake

Example​

ScenarioScoreStakePayout
Strong signal+0.045,000+50 SIGNAL
Good signal+0.025,000+25 SIGNAL
Neutral0.005,0000
Weak signal-0.025,000-25 SIGNAL

Provisional Scores​

You don't have to wait 20 days for feedback:

  • Daily updates: After each trading day, we publish provisional IC scores based on partial returns
  • Not final: Provisional scores can change significantly before resolution
  • Useful for: Early signal quality assessment, engagement during the wait

Diagnostic Rounds​

Practice rounds using historical data with instant feedback:

  • No staking required
  • Scored immediately against known historical returns
  • Great for testing new models before going live
  • Results don't affect your live ranking