disclaimer: no affiliate ties whatsoever

From a ML perspective using low bias but high variance independent predictions cancels out the variance so I think this could be a major thing. Looking to see if there are any objection I've missed. Sample size is garbage but the underlying logic based on ensemble algorithms isn't.

https://cindicator.com/

whitepaper link: https://cindicator.com/Cindicator_WhitePaper_en.pdf

Use the altcoin thread

yes they are the first to apply maths to markets or combine human intelligence with AI

I would, but the topic would be forever lost.

In regards to news:

I mean, I'm having hard time finding the scam part of this except the obvious offshore lol zones and adverse selection with ICO. And somehow this thread went straight to page 4 of google search in 1 day.

So much wrong with this

1) No benchmark
2) ML is already pretty crowded, they are not innovating here.
3) When someone implies those returns, run
4) Market is already a voting machine, are they not doing any pricing, just trying to win by fitting well?
5) You need to pay a fee to invest, so they need additional outperformance.
6) Even if you do invest, you'll essentially be paying for the privilege of being frontrun
7) Even if you do invest and win, you need to pay a performance fee
8) Incentives aren't aligned (forecasters/company, forecasters/investors, company/investors)

Also, definitely selling securities imo

https://venturebeat.com/2017/08/29/t...ings/view-all/

Ensembles outperform individual models in all machine learning competitions. So using a large amount of individual forecasters will likely be better then the best individual forecaster.

There are a bunch of competitors in this space and its early days, dunno about valuations . What I like about these projects is that there is a economy with people that are forecasting, people that are inputting the data, people that create the ensemble. So everyone can give data to the blockchain, blockchain anonimizes it, machine learners do their thing on (part) of the data. If data was valuable u are rewarded. So there will be some people that will be specialized in getting data from different sources, some people specialed in machine learning. Ensemble layer that is automatically created. Everyone doing their part but No1 has acces to the whole thing. If you have 8000 people getting data from different sources, creating different features in the end you should be able to out-perform all the other algos.

People are also creating projects for this thing in healthcare sector which is kind of cool. Everyone can input all their medical history, blood values and these things. Idea is to get a superbig worldwide anonymous database where all the machine-learners can then train their models on. AI does the ensemble and everyone gets back predictions based on their personal data. Blockchains are nice in those areas where you can good results with enough input data but the data is sensitive so you dont want anyone to actually have the full acces to it

the white paper is two pages of anecdotes illustrating wisdom of the crowds then a hundred pages of abstract formulas a la science journals that either a) are meaningless or b) give away secret sauce

this project is a microcosm of what financial markets are and what hundreds of billions of dollars have already been deployed to do: look at ex-post data and determine what translates to ex-ante returns.

nothing differentiates this project from a hedge fund - or a kid in a basement -
and i would wager significant money they can't maintain excess returns (if they had any kind of benchmark) over any significant period.

but ICO blockchain wayofthefuture etc etc this will raise billions no question

Ill take the under

God these things are really perfect if you are some huckster finance sales guy. I mean nobody really understands how currency works, every other asset is expensive and it has a nice futuristic feel to it. And the best thing, you can quote ridiculously high numbers like the global money supply to estimate total addressable market size. Anybody who tries to call you out has to start some complicated explanation to prove you wrong (and few will want to pay attention to those 'buzz kills').

Perfect recipe for a massive bubble.

Yep. And the 1000 bagger return on cryptos in general gives the upside arguments weight. Perfect scam fodder.

Ok, do we then concur there is a short term asymmetric high potential?

I think the underlying "technology" here is completely uninteresting/worthless. ICOs are about writing words in sequences that tickle the brains of people wanting to pretend they're buying into something other than a pure scam/ponzi.

You have < 1% that any trading alpha will come from this. "Is this ICO written such that it's likely to scam other people to buy into the ponzi after me" is the only question you need to ask yourself.

In general, I think human traders have 5-10 years left before ML cleans up all the easy edges. It has already, a little. But that has nothing to do with the proposal in the OP.

The concept isn't worthless in principle but the odds of them outdoing what any random ml engineer could do because zomg ensemble and 'clusterization' is close to 0

U dont think u can get better data input this way? So much sources available. Obv there is the price data which everyone uses but then theres twitter, FB, news websites, private sources. Seems difficult for random ML engineer to utilize everything