Forums  > General  > numerai  
     
Page 2 of 3Goto to page: 1, [2], 3 Prev Next
Display using:  

Nonius
Founding Member
Nonius Unbound
Total Posts: 12706
Joined: Mar 2004
 
Posted: 2016-05-10 00:05
well, to be fair my wondering samurai, he has or had a good set of neurons. I've kept in touch with him since 83. he was actually, and ironically the first mofo who mentioned to me about some motherfucker called Jim Simons. I was like, what the fuck. he was like, mofo, this dude is the gas., I was like, whatevs. then I tried to learn Chern Simons and shit. that lead me to some dark shit and shit.

Chiral is Tyler Durden

Nonius
Founding Member
Nonius Unbound
Total Posts: 12706
Joined: Mar 2004
 
Posted: 2016-05-10 00:13
then, bruv, when I was like thinkin my brain can't handle chern and simons and attiyay and shit, and donaldson and smale, I was like, mkay bro, I'll just hang at the pool with some motherfuckers and shit. it went down hill from there.

btw, why is there so much hostility here? Goldarak and Expresso trading barbs and shit bro. it ain't right.

Chiral is Tyler Durden

tradeking


Total Posts: 7
Joined: May 2016
 
Posted: 2016-05-10 02:14
EspressoLover:
> In this toy model, the vast majority of your trades are negative EV. Medallion's alphas are worth billions, but 50% of Medallion's alphas are worth nothing. And this toy model is very close to what Numeral is dealing with. There only hope is to data mine marginal information that others don't have, while keeping non-marginal information out of their models. That's just not going to happen. Almost all the major sources of alpha are pretty much solved. Discovering something that RennTech, Citadel, Two Sigma, KCG, HRT, et al. don't know is nigh impossible. (And whatever is new to discover, is almost certainly not a well-defined feature in a "curated dataset"). Collectively the major quant shops have near perfect LOB alpha, near perfect residual reversion, near perfect order flow models, etc. Even being wildly optimistic Numeral may get 2-3% marginal information relative to any major traders, and capture 80% of the non-marginal information. But without anything special on the monetization or infrastructure side, that simply doesn't mean squat.

Your theory might be true for HFT alphas, but judging from the size of the test dataset that is provided by numerai (~50k rows), they are most likely searching for daily alphas. Pickings are still plentiful for alphas of this range so their approach could be fruitful even without a #1 ranked R-squared.

ronin


Total Posts: 219
Joined: May 2006
 
Posted: 2016-05-10 10:35
> it went down hill from there.

Same story for all uf us, just different parts of maths/physics we couldn't handle any more.

> Goldarak and Expresso trading barbs and shit bro. it ain't right.

I think they are still discussing numerai - it does that to people. Your friend is *so* much more interesting.

"People say nothing's impossible, but I do nothing every day" --Winnie The Pooh

radikal


Total Posts: 258
Joined: Dec 2012
 
Posted: 2016-05-10 17:01
> mathematician friend in the states who is getting PAID to do research on behalf of some porn-king who has a new theory of gravity

What comes up must...

I think I'm came off as more cynical-prick than I had intended. I think there's some probability of success. And investors and creators are perhaps right to chase it. Hard to know what will work but fortune favors the bold salesman.

There are no surprising facts, only models that are surprised by facts

Mat001


Total Posts: 33
Joined: Aug 2014
 
Posted: 2016-09-11 13:54
LOL

fomisha


Total Posts: 29
Joined: Jul 2007
 
Posted: 2016-09-12 20:33
in a competitive industry, in the end, it comes down to price. data work can be done cheaper abroad. so, this model has legs. no, they will not do better than ren tech and two sigma but they will do some of the same things cheaper. and they will be able to process datasets that it does not make commercial sense to processes at the cost of labour on wall st.

I am not sure how they will deal with the problem if someone tries to laundry insider information through them.

EspressoLover


Total Posts: 254
Joined: Jan 2015
 
Posted: 2016-09-13 00:03
> data work can be done cheaper abroad. so, this model has legs.

Maybe. But if that's the value proposition, Numeral is a hell of a way to get to that point. Opening offices and hiring researchers in Bangalore, Buenos Aires and Belgrade seems like a lot less complex way to take advantage of that. Throwing in all the SV buzzword crap like "distributed research", "cryptographic anonymity" and Bitcoin is just introducing a lot of untested points of failure into the core business model.

For that matter, there's no barrier stopping Akuna or AQR from already shipping quant work overseas. For the most part they don't. To the extent these firms have overseas offices, they're almost always either for back-office software or front-office work exclusively dealing with local markets. So I'd be pretty suspect about how much untapped global supply of cheap talented researchers there actually are.

Alpha and trading research tends to be way more complex and have a way higher quality threshold than most technical work. But third-world off-shoring struggles to produce acceptable results on even basic software projects. Again, I think there's a reason that Bangalore isn't a quant powerhouse. And I don't think Numerai will be changing this anytime soon.

eeng


Total Posts: 15
Joined: Dec 2014
 
Posted: 2016-09-13 09:40
I always thought that the reason not to off-shore quant jobs was IP protection and law enforcement related. Just remember what happened to the Russian GS guy that appears on Flash Boys, would it have been the same in Bangalore?

AndyM


Total Posts: 2321
Joined: Mar 2004
 
Posted: 2016-09-13 10:53
>>Maybe. But if that's the value proposition, Numeral is a hell of a way to get to that point. Opening offices and hiring researchers in Bangalore, Buenos Aires and Belgrade seems like a lot less complex way to take advantage of that. Throwing in all the SV buzzword crap like "distributed research", "cryptographic anonymity" and Bitcoin is just introducing a lot of untested points of failure into the core business model.

But if the trade is not global labour arbitrage, but naive young quants with stars in their eyes arb, then it's perfect as is.

I used to be disgusted; now I try to be amused...

ronin


Total Posts: 219
Joined: May 2006
 
Posted: 2016-09-13 17:17
> But if the trade is not global labour arbitrage, but naive young quants with stars in their eyes arb, then it's perfect as is.

It's neither of those.

Naive young quants take more in handholding than they produce, for years.

But unlike blind data scientists from Bangalore, at least naive young quants get their hand held, and over time they become less naive and less young, but they start paying their rent.

If you have a recipe for extracting a few million per year in pnl from a naive young quant, then you are on a different plane from the rest of us. Respect.



"People say nothing's impossible, but I do nothing every day" --Winnie The Pooh

AndyM


Total Posts: 2321
Joined: Mar 2004
 
Posted: 2016-09-13 20:42
Of course there's no actual arb. It pretty much goes without saying that you cannot produce anything of value in this way. The site and concept are clearly a joke. But they're spinning a story that some people are biting at, so in the vapourware universe...maybe they can sell the sizzle, rather than the steak.

I used to be disgusted; now I try to be amused...

tbrown122387


Total Posts: 34
Joined: Feb 2010
 
Posted: 2016-09-19 05:47
A lot of discussion about efficacy, but a few things that I didn't see mentioned:
1. They're looking for a distribution of a 1-dimensional random variable at a time. This destroys information if you're trading more than one asset (can't reconstruct joint). But perhaps the training data csv is not representative.
2. They're modelling a bernoulli rv. This destroys information (sigma field generated by this rv is coarser than the original). But perhaps there's some reason for this I'm not taking into account.
3. the multiple accounts hack. But perhaps they are very good at spotting fraud accounts.

Mat001


Total Posts: 33
Joined: Aug 2014
 
Posted: 2016-09-25 11:00
If you look at the leaderboard, multiple accounts are evident. They use boosting to improve logloss. Some members have thousands in yearly earnings but no actual earnings. The whole concept has many holes. Lack of transparency is one of them.

NeroTulip


Total Posts: 997
Joined: May 2004
 
Posted: 2016-09-25 15:29
Why are multiple accounts evident?

"Earth: some bacteria and basic life forms, no sign of intelligent life" (Message from a type III civilization probe sent to the solar system circa 2016)

Mat001


Total Posts: 33
Joined: Aug 2014
 
Posted: 2016-10-09 11:58
Why are multiple accounts evident?

HitmanH


Total Posts: 443
Joined: Apr 2005
 
Posted: 2016-12-12 13:41
I used to rate USV... Until Now..

NeroTulip


Total Posts: 997
Joined: May 2004
 
Posted: 2016-12-13 02:48
C'mon, Numerai is genius as a way to commoditize the skillset of machine learning PhDs. And since the data is encrypted and they don't even know what they are working on, they can't ask for a raise by threatening to take their model somewhere else. An arbitrary large share of the profits can accrue to Mr Craib rather than to the data scientists who created the models. It is the ultimate quant sweatshop.

It's just a bit mysterious why they would need funding for the management Co though. The costs are really minimal.




"Earth: some bacteria and basic life forms, no sign of intelligent life" (Message from a type III civilization probe sent to the solar system circa 2016)

NIP247


Total Posts: 540
Joined: Feb 2005
 
Posted: 2016-12-13 14:45
And now in Wired...

On your straddle, done on the puts, working the calls...

NeroTulip


Total Posts: 997
Joined: May 2004
 
Posted: 2016-12-13 15:34
"If the fund reaches a billion dollars under management, Craib says, it would pay out over $1 million each month to its data scientists."

So, about 1.2% out of the 2/20 for the geeks. And fatal error, he should have said "when", not "if", to keep the Silly Con Valley thing going.

"Earth: some bacteria and basic life forms, no sign of intelligent life" (Message from a type III civilization probe sent to the solar system circa 2016)

katastrofa


Total Posts: 369
Joined: Jul 2008
 
Posted: 2016-12-13 20:02
PhDs have been commoditized long time ago.

goldorak


Total Posts: 1000
Joined: Nov 2004
 
Posted: 2016-12-13 20:45
> It's just a bit mysterious why they would need funding for the management Co though. The costs are really minimal.

You always need good lawyers...

If you are not living on the edge you are taking up too much space.

JTDerp


Total Posts: 44
Joined: Nov 2013
 
Posted: 2016-12-14 01:09
They just made everybody's favorite doom p0rn site: http://www.zerohedge.com/news/2016-12-13/new-hedge-fund-2-20-model-trades-purely-anonymously-submitted-ideas

"Meanwhile, Numerai is not the only game in town when it comes to anonymously sourced trading strategies as Steven Cohen's new Point72 Asset Management has helped fund another competitor, Quantopian.

'Perhaps the most successful of these new ventures, Quantopian, is already onto its Series C fund-raising round. Quantopian has attracted hundreds of millions of dollars from the venture capital firm Andreessen Horowitz and the new investing firm run by Steven A. Cohen, Point72 Asset Management.
With Quantopian, data scientists submit investing algorithms that asset managers around the world can license and use to trade. The algorithm’s authors get a cut of any profits. The investment ideas are not predictions of specific price patterns, but instead are mathematically based models that can be applied to different investments.'

Yes, we can certainly understand why the idea of 'anonymously' sourced trading ideas would be appealing Mr. Cohen."

The clouded mind seeks; the emptied mind finds.

NeroTulip


Total Posts: 997
Joined: May 2004
 
Posted: 2016-12-14 06:32
The real killer, and I think Goldorak will appreciate this, is that they do not talk about multiple hypothesis testing, White's Reality Check, or anything like that. Whaddaya say Goldorak?

"Earth: some bacteria and basic life forms, no sign of intelligent life" (Message from a type III civilization probe sent to the solar system circa 2016)

goldorak


Total Posts: 1000
Joined: Nov 2004
 
Posted: 2016-12-14 13:39
I love the smell of lost dollars in the morning. Evil Smile

Anyway looking at my own affairs I don't know if it is even worth taking care of that kind of gory details...

Thinking of it, I actually may have something to sell to these guys.

If you are not living on the edge you are taking up too much space.
Previous Thread :: Next Thread 
Page 2 of 3Goto to page: 1, [2], 3 Prev Next