LogoLogo
TournamentsHedge FundDiscordGitHubForumDocs
  • Numerai Tournament
    • Overview
    • Data
    • Models
    • Submissions
      • Model Uploads
      • Compute Heavy
      • NGROK Example
      • CRON Example
    • Scoring
      • Definitions
      • Correlation (CORR)
      • Meta Model Contribution (MMC)
      • Feature Neutral Correlation (FNC)
      • Grandmasters & Seasons
    • Staking
    • Bounties
  • Numerai Signals
    • Overview
    • Data
    • Models
    • Submissions
    • Scoring
      • Definitions
    • Staking
    • Signals + QuantConnect
  • Numerai Crypto
    • Overview
    • Data
    • Submissions
    • Scoring
      • Definitions
    • Staking
  • Community
    • Discord
    • Forum
    • Twitter
    • Youtube
    • Apps & Content
      • Office Hours with Arbitrage
        • Office Hours Recaps: Season 1
          • OHwA S01E01
          • OHwA S01E02
          • OHwA S01E03
          • OHwA S01E04
          • OHwA S01E05
          • OHwA S01E06
          • OHwA S01E07
          • OHwA S01E08
          • OHwA S01E09
          • OHwA S01E10
          • OHwA S01E11
          • OHwA S01E12
        • Office Hours Recaps: Season 2
          • OHwA S02E01
          • OHwA S02E02
          • OHwA S02E03
          • OHwA S02E04
          • OHwA S02E05
          • OHwA S02E06
          • OHwA S02E07
          • OHwA S02E08
          • OHwA S02E09
          • OHwA S02E10
          • OHwA S02E11
        • Office Hours Recaps: Season 3
          • OHwA S03E01
          • OHwA S03E02
          • OHwA S03E03
          • OHwA S03E04
        • Office Hours Season 4
        • FAQ in Office Hours
        • Cited resources
      • Structure of Numerai by Wigglemuse
  • NMR
    • Coinbase
    • Uniswap
    • Etherscan
    • Github
  • Connect
    • Index
Powered by GitBook
On this page
  • What is CORR?
  • How is CORR computed?
  • CORR on the website
  1. Numerai Tournament
  2. Scoring

Correlation (CORR)

What is CORR?

The simplest scoring metric on Numerai is the correlation between your model's predictions and the target. Numerai uses a special variation of correlation we call "Numerai Corr". At a high level, this metric is designed to be a good proxy for actual portfolio returns if the predictions were used in live trading.

How is CORR computed?

def numerai_corr(preds, target):
  # rank (keeping ties) then Gaussianize predictions to standardize prediction distributions
  ranked_preds = (preds.rank(method="average").values - 0.5) / preds.count()
  gauss_ranked_preds = stats.norm.ppf(ranked_preds)
  
  # make targets centered around 0
  centered_target = target - target.mean()
  
  # raise both preds and target to the power of 1.5 to accentuate the tails
  preds_p15 = np.sign(gauss_ranked_preds) * np.abs(gauss_ranked_preds) ** 1.5
  target_p15 = np.sign(centered_target) * np.abs(centered_target) ** 1.5
  
  # finally return the Pearson correlation
  return np.corrcoef(preds_p15, target_p15)[0, 1]
  • First the predictions are gauss ranked. We do this to match our live trading process where all model predictions are standardized this way before being ensembled together in the meta model.

  • Then the target is centered around 0. We do this to match the gauss ranked predictions which are now centered around 0.

  • Finally both the gauss ranked predictions and the centered target are raised to the power of 1.5 before calculating the Pearson correlation. We do this to accentuate the tails as hedge fund tends to only trade the stocks with highest or lowest predicted returns.

The key takeaway here is that your prediction's distribution does not matter. You are only evaluated on your prediction's ranks. And your score depends more on the tails than a typical rank-correlation.

CORR on the website

You may see different variations of CORR on the website

  • CORR20V2 - the latest CORR score against the 20 day version of main target

  • CORJ60 - correlation to the 60 day version of the auxiliary target named Jerome

  • CORT20 -correlation to the 20 day version of the auxiliary target named teager

PreviousDefinitionsNextMeta Model Contribution (MMC)

Last updated 3 months ago