Hi @toms91 - I have started with Numerai tournament about one year ago when $NMR went from about $25 down to $8 and possibility of floor at $1 was discussed here. So I got used to it and volatility is not bothering me. Just average performance of your models can bring you 50%+ annual return. More over if you decide to buy predictions at NumerBay, you can choose above average returns (currently model #1 is in fact purchased with 600%+ annual return). So this type of returns (possibly combined with buying $NMR for staking during dip) can easily compensate for volatility.
If you focus your attention to ML/#data-science and behaviour of stock markets you will get your reward - guarantied, even measured in $NMR. If you will be able to convert it to monetary value is still debatable.
Seems a slightly strange way of looking at things to me.
Hardware sales or cloud providers don’t change their prices from dollars if you want to train large models so your fixed costs of participating are always the same while your payout moves all over the place. If you can ever sell the NMR at all.
Actually I have $0 hardware/automation costs: in always free Oracle Cloud I have small flask server receiving webhooks and triggering swarm of Kaggle notebooks (max 10 concurrently running, each 4 CPU/30GB RAM) doing training and submitting.
That may explain my strange way of looking at Numerai. Additionally, I do not believe that size or method of model can give you superior performance (I would need some evidence to change my mind) because input data are extremely noisy. Also signal (if present) is ever changing. Simply put: it may be just a random walk down Wall Street.
Take this example. You started staking models a month ago:
NMR Close Price April 8th 2023 = $19.13
NMR Price at time of writing = $14.76
Change -22.8%
Before even getting any payout at all its cost you 22.8% of your investment to supply a hedge fund with data. Data they can then use to make money with.
If it was based on a real major currancy, or a stable coin if we much use crypto, that volatility would be a fraction of the size and something you could actually develop on top of a base foundation of stability.
No @toms91 your point is clear. The reason I started this topic is that there is risk that yours and everyone’s $NMR stake can go to zero value due to characteristics of crypto business. Some guys (@quantverse, @wigglemuse) are hearing here complaints about volatility quite regularly, complaint which I was not raising.
My main concern is disconnect in incentives of fund and prediction providers and risk asymmetry (we may lose everything, fund will own stocks portfolio).
“If we get to $1 billion in AUM we’re going to have no problem paying our users in any currency they want. Right now we like NMR, but in the future it could be any currency or payment.”
In times like these, it’s good to resurrect this thread.
The users take risk when they buy NMRs, but what is Numerai’s risk? Numerai chooses the users’ payout/paycut and how the users’ NMR are burned. At the same time, Numerai owns the majority of the coins, so they have a lot to gain from the risk the users take on joining the tournament (users buy NMRs → NMR rise in value) without risking anything. At the same time Numerai is the one who makes the rules. Isn’t it a conflict of interest?
If they paid us in USD I wouldn’t mind so much paycuts and constant changes in the rules, because it’s their tournament after all and they have a hedge fund to run, but as things are with NMR I am pretty disappointed and worried.
If you have a better point of view, where a NMR based tournament is not unfair, would you mind sharing it with us? I would really like to have more faith in Numerai, the tournament and NMR, because It’s been a really interesting/rewarding experience so far and I would like to keep it going.
I forgot to add one related point. Since the users take risk when buying NMRs, the stake weighted meta model would have a value in itself even if all the staked models were submitting prediction based on the benchmark models. For that reason, any payout metric that takes the SWMM model for granted and rewards only a model additional contribution (e.g. TC or MMC) to the meta model is unfair, because it doesn’t take into consideration the risk taken to build the SWMM in the first place.
I had a dream. All users unstaked their models and added them to NumerBay. Numerai had to buy the models from NumerBay to keep the hedge fund running. The users stopped being exposed to the NMR risk and Numerai could pay only for the worthy models. The reward was still in NMR, but there was no risk associated with it (no staking) and the payout (model price at NumerBay) was fair, that is determined by the market (model performance/price tradeoff).
The idea is that Numerai, who owns large amount of NRMs, doesn’t take any risk in buying the model predictions, and who sells the predictions can convert the NMR to Fiat money right away. So no risk, unless who gets the NMR really wants to be exposed to NMR and hold them hoping for a future price increase.
