No. Long short market neutral global equity hedge fund.
How many tournaments are there?
One tournament:Kazutsugi Historically, Numerai has had up to seven tournaments.
How long is each round of the tournament?
one (1) week
When are round earnings paid out?
Weekly, four (4) weeks after the start of the round
Does everybody get the same data?
Yes, the dataset download link is the same for everybody. Everyone gets the same data.
How much USD or NMR is awarded?
Check out the https://numer.ai/rounds to see the prize pools for current round.
How are the weekly payouts determined?
See the Payouts section of the Tournament Walkthrough
Does Numerai consider my best submission or only my latest submission?
Numerai uses only the latest submission.
Which of my submissions are considered for payout?
Only the last submission made by an account in a given round is considered for earnings
How do I withdraw earnings?
When logged in, see the "Earnings" button at the bottom of your user information control panel.
How do I get an ether wallet to withdraw earnings into? How do I sell ether?
There are many ways to do this. The easiest is Coinbase. Set up a free account there. You can withdraw your Numerai ether earnings into your Coinbase account and sell the ether for USD and transfer to a bank account.
How fast are USD or NMR withdraws from my account?
Expect less than 1 minute for the transaction to first appear on your account. How quickly the funds become available to sell from your account depends both on the service you're using as your wallet, as well as the current load on the blockchain. Typically it's less than 15 minutes.
What is Numeraire?
Numeraire (NMR) is Numerai's crypto token. Learn more here.
Where can I get NMR to stake in the tournament? Uniswap, Bittrex, Poloniex, Shapeshift, Airswap, 0x, Upbit, DDEX, Paradex
You can also switch out your ETH for NMR using NMR at your Numerai wallet
Is there a Telegram? Yes, here's the NMR official telegram
-ln(0.5)used in the original NMR whitepaper?
-ln(0.5) = 0.6931471805599453... This is the logloss you'd expect to get from making completely random guesses. A logloss lower than this is better than random guessing.
What are the features and ids in the dataset?
The features and ids of the dataset is intentionally obfuscated to remove human bias from the data science.
Why isn't the data for past rounds included in the training data of subsequent rounds?
There is a trade off between the length of the test data (used in our backtest) which gives us a lot of information to build the meta model and the length of the training data which gives users more information. Particularly, we like that our test set Sharpe is calculated on a large set of data. If we turned the test data into training data, we wouldn't have much to go on unless the staking tournament was working perfectly and users got extremely good at estimating p and communicating it through stakes. This is subject to change in the future.