Sharpe and Sortino ratios on live performance of your models

As numerai models only earn or burn when the round is resolved, you may be ignoring the information content of the 20 daily scores in between.

The daily gyrations of a model is similar to the daily account value of a buy-and-hold brokerage account. Yes, you don’t realize profit or loss until you sell, yet a slow and steady climb might be more deserving of your hard earned NMRs.

More critically, historical daily scores are a high resolution indicator of your model’s volatility under different staking regimes, and may influence your choice of stake type. For example, half-mmc would yield higher Sharpe and Sortino ratios for @mdo over the past six months compared to other staking methods.

Feed a list of accounts (yours and those you follow) to the live-sharp-sortino notebook. It will read the daily scores from numerai and calculate sharpe and sortino ratios on their live daily-score performance.

So, put this in the notebook:

# your models and others you want to bencmark
account_list = [
    "degerhan",
    "era__mix__2000",
    "hiryuu",
    "integration_test",
    "mdo",
    "sorios",
    "themicon",
    "arbitrage",
]

and receive:

annual_return_corr annual_return_halfmmc annual_return_1mmc annual_return_2mmc sharpe_corr sharpe_halfmmc sharpe_1mmc sharpe_2mmc sortino_corr sortino_halfmmc sortino_1mmc sortino_2mmc rows
themicon 1.333991 1.710942 2.082832 2.811966 2.572024 2.838774 2.938870 2.965025 4.005048 4.448977 4.652587 4.872778 130.0
degerhan 1.780967 2.237819 2.679758 3.523582 2.564452 2.661918 2.703920 2.711582 4.292114 4.627113 4.731770 4.605466 37.0
hiryuu 1.164212 1.507592 1.850411 2.532070 1.987121 1.854426 1.766267 1.660042 3.529113 3.270079 3.053915 2.822324 130.0
mdo 0.958362 1.103332 1.243571 1.509940 1.844011 1.904945 1.873802 1.729242 3.301595 3.398901 3.190205 2.735015 130.0
era__mix__2000 1.369557 1.743672 2.112315 2.833632 2.362132 2.362064 2.322188 2.242782 3.169988 3.030299 2.871511 2.664848 130.0
sorios 0.977431 1.110031 1.239880 1.491600 1.867191 1.791917 1.648394 1.399152 2.507887 2.460163 2.356310 2.064312 130.0
arbitrage 0.885076 0.941292 0.995323 1.096883 1.617012 1.522430 1.400110 1.169640 2.887307 2.723927 2.332565 1.768363 130.0
integration_test 1.041502 1.104208 1.165780 1.285710 1.492098 1.437492 1.373536 1.243987 2.374332 2.202366 2.008255 1.720881 130.0
22 Likes

Very interesting! Thank you for sharing!

@degerhan Thanks for sharing this.