Back to Community
An alternative to trade strategies: Erasure Quant

I just bumped into this. Can be very interesting for people familiar with crypto. You can submit stock score predictions and create or destroy value depending on the accuracy of your predictions.

https://medium.com/numerai/introducing-erasure-quant-963e8e74c80c

https://medium.com/numerai/numerai-reveals-erasure-unstoppable-peer-to-peer-data-feeds-4fbb8d92820a

4 responses

Currently undergoing a revamp and will be back online in June.

I gave the first incarnation a shot and was unimpressed.

Documentation was incomplete and full of grammatical errors. Other information they have posted online (such as on Medium) was flat out wrong or contradictory.

The leaderboard/returns charts are full of data errors. For example, they don't account for splits or dividends, so some charts jump 50% in one day.

They have a convoluted system of converting position ranks to portfolio weights. If you're doing any sort of dynamic hedging it gets lost in translation during the process of converting your portfolio weights to ranks and then back to portfolio weights. The basket I submitted performed well locally, but performed much worse on Erasure Quant.

The (not so explicit) "agreement" is such that you're basically giving them your real-time positions for free, which they are free to use, and they can at their option reimburse you. Or rather, not free but at your own cost, since it costs ETH gas to make submissions and NMR if you want to "stake" your submissions. With the "staking" concept, you basically invest your own money in your own strategy on the stock market. It's not clear whether that is more rewarding than investing directly in your own strategy. I think you should always be skeptical of pay-to-play schemes. Pay-to-play blockchain schemes doubly so.

I've gotten like 20 emails notifying me of receiving payments of 0.0 NMR. Yeah, stuff like that.

With the whole blockchain approach they're solving a problem that doesn't really exist (Sybil attacks) while missing most of the primary problems in this space (such as multiple comparisons bias and alpha non-stationarity) and creating a whole slew of new ones. Lets hope they figure it out with version 2, but I'm not holding my breath.

Wow @Viridian I see you've dig into it! Thanks a lot for the feedback.

@Viridian you're being too generous! ErasureQuant was even worse than you describe :)

We had one single junior engineer rush build it by himself but Numerai Signals is going to be a lot better. I think you will like.

Richard
-CEO Numerai