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
      • Alpha
      • Meta Portfolio Contribution (MPC)
    • 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
  1. Numerai Signals
  2. Scoring

Alpha

One of the primary scores of Signals using the 60D2L chili target.

PreviousDefinitionsNextMeta Portfolio Contribution (MPC)

Last updated 1 month ago

CtrlK

CORR, FNC, and MMC are all imperfect metrics for paying Signals users. This is because users can be good at these scores, but backtest poorly. To better align the Numerai Signals tournament with the Numerai hedge fund, the performance of a submission must align with it's backtest performance.

The Numerai research team invented 2 pieces of data that provides a way to do this:

  • Neutralizers Matrix: a 200-column condensed matrix used to produce a neutralize signal

  • Sample Weights Vector: a weight vector used for sampling a signal to produce a set of stock weights that accounts for real-world trading constraints such as adv

Using these, we can create a set of "neutralized weights" from a signal as follows:

Given the following data:

  • signal s (NaNs filled w/ 0.5)

  • neutralization matrix N

  • weight vector v

We first create normalized signal s’:

s` = normalize(rank(s))^1.5

Then generate neutralize predictions and sample them:

neutral_preds = s` - (N @ (N.T @ (v * s`)))
neutral_weights = neutral_preds * v

Finally, we use the 60D2L target chili to derive the final alpha value:

alpha = neutral_weights @ target_chili_60

This value provides us some insight into how a signal might perform in a backtest and more closely aligns the Signals Tournament with the hedge fund.