Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Quants Run Wall Street Now

wsj.com

Post reply on HN
Print view View original

61–70 of 165 posts

Re: The Quants Run Wall Street Now

#61

Post by SEJeff »

kasey_junk wrote: post #52
dsacco wrote: post #43

Earlier quoted context omitted.

Virtu only lost money trading one day out of 1278 trading days between 2009 and 2014. In the most uncharitable analysis (1278/2; or the lost day happened in the middle), they had a 0.5^639 chance of doing that. Maybe you disagree with 0.5 per day. Let's make it 0.9! ...But that's still 5.7 x 10^-30. How many firms do we need to exist for this to emerge by chance? This and website bug bounties being sold on the black…

> Virtu only lost money one day out of 1278 trading days Lost money trading . If they made $1 trading that counted though I assure you that day was a loser from a business perspective. I also have not lost money trading in 3 years, simply by not trading at all. It doesn't really detract from your broader point, which I don't have a strong opinion on but that stat is a pet peeve of mine as it's fairly meaningless.

This is massively incorrect. If Virtu only made $1 trading, they would still make massive amount of $$$ via exchange rebates, as they are a designated market maker. In fact, they aren't the only market maker that does this. A market maker guarantees they'll take trades on both sides of the book, so long as their strategies aren't grotesquely losing, they'll always come out ahead due to exchange rebates.

http://www.investopedia.com/articles/active-trading/042414/w...

Source: I worked for Virtu's predecessor Madison Tyler for 4 years and for Virtu after it merged with Madison Tyler for ~8 months and have first hand experience with HFT.

Re: The Quants Run Wall Street Now

#62

Post by tzs »

SirLJ wrote: post #49

Everyone can try to do it, but you need to spend the time and educate yourself like in any other profession or hobby and the first step is to go to eBay and search for historical stock market data, you can buy 20 years of data for less than $100 and you can test all trading ideas for free and without losing a single penny...the barrier for entry is very low, some Python knowledge + Linux machine and the data and off…

If you are going to use machine learning on the data, though, make sure you know what you are doing unless you are just using someone else's complete package. It's really easy to screw up machine learning.

I recall an example given in a class I took. (I may be misremembering the details, though).

Some people were trying to apply machine learning to currency trading. They had a bunch of data. They normalized the data (a common step in machine learning), and divided it into training and test and validation sets, and trained their model. Everything looked great, and they were getting excellent results on the test set.

When they went live with real money instead of making the nice profit predicted, they lost a lot of money.

Their mistake? They normalized the whole data set up front. What they should have done is split it into the training, test, and validation sets, and normalized each of those individually. Normalizing before splitting compromises the independence of the sets, biasing the learning.

(I must admit I never did quite understand this. Normalizing is optional. As far as I understand if one does an arbitrary transformation on one's data as a whole that should not actually make learning go bad, at least as long as the same transformation is applied to all the input when you go live. So when they did a normalizing step on the whole data set, why wasn't that just like doing any other arbitrary transform? There is serious dark magic here...)

Re: The Quants Run Wall Street Now

#63

Post by seibelj »

zackmorris wrote: post #48

Anyone have hard numbers on if/by how much quants outperform old fashioned techniques like flags and finding stocks that tend to go the opposite direction from the one you're interested in? I don't know the terminology for all this but I know there are limits due to uncertainty so even the best algorithms may not do much better that someone guessing. Do quants do 10% better, 2x better, 10x better? My Dad and I had a…

I don't do this, but there is a place without any rules, 24/7 trading, wild volatility, and tons of dumb money. This place is cryptocurrencies. Supposedly it's a lot easier to trade in these markets, less competition, less money needed. I used to laugh at my brother, but he turned $20k into $300k, so it is possible.

Re: The Quants Run Wall Street Now

#64

Post by matheweis »

SirLJ wrote: post #49

Everyone can try to do it, but you need to spend the time and educate yourself like in any other profession or hobby and the first step is to go to eBay and search for historical stock market data, you can buy 20 years of data for less than $100 and you can test all trading ideas for free and without losing a single penny...the barrier for entry is very low, some Python knowledge + Linux machine and the data and off…

What about options? So for I haven't been able to find any good source other than CBOE and their resellers, which are all quite expensive.

Re: The Quants Run Wall Street Now

#65

Post by dluan »

My question for these new would-be quants is, how do you see your work?

I know there are certain types of smart whizzes who see working for Google/Facebook/Amazon as some sort of intellectual step down, or where in finance the value to society is some handwavy "market-making" argument (depending on how philosophical you get), but presumably these types going into this are also tenure-track, research producing scientists. You could spend your time disproving the Higgs at Cern, or you could optimize pennies (albeit in ever more challenging ways).

Have these Renaissance/TwoSigma type firms really ever delivered social value that isn't a new trading model with a limited shelf life? Is there new tech? Network, hardware, software innovation that spills over into the real world?

Re: The Quants Run Wall Street Now

#66

Post by tzs »

yourapostasy wrote: post #54
SirLJ wrote: post #49

Everyone can try to do it, but you need to spend the time and educate yourself like in any other profession or hobby and the first step is to go to eBay and search for historical stock market data, you can buy 20 years of data for less than $100 and you can test all trading ideas for free and without losing a single penny...the barrier for entry is very low, some Python knowledge + Linux machine and the data and off…

Why isn't this data out on torrents? Is it somehow copyrighted?

I have a hard time seeing how it could be copyrighted, at least if we are talking about comprehensive stock listings organized in the obvious way (e.g., a table of prices organized by date, where the stocks included are chosen by some straightforward criteria).

In the United States I'd expect this to be covered by Feist Publications, Inc., v. Rural Telephone Service Co., 499 U.S. 340 (1991).

[1] https://en.wikipedia.org/wiki/Feist_Publications,_Inc.,_v._R....

[2] https://www.law.cornell.edu/copyright/cases/499_US_340.htm

Re: The Quants Run Wall Street Now

#67

Post by fizixer »

meddlepal wrote: post #38
crimsonalucard wrote: post #34

The value of something across the span of milliseconds has no meaning. At this temporal level of resolution it's just a gamble no different from games like poker. Why deploy brain power on trying to extract this value? What a waste of intelligence.

If you haven not been keeping up, since about sometime in the mid-1980's acquisition of Money became the defining virtue of goodness and success. That's all that matters now.

The problem is that it's more like sports than math/science: a tug-of-war (zero-sum game) instead of win-win.

You're pitting your intelligence against the intelligence of other quants (not against the complexity of some natural system or something). You have 3 phds? they bring in 4. You bring in 6? they bring in 10. You bring the processing down to 50ms, they go 45ms; you go 5ms, they go 2ms. The game never ends.

It's even more depressing when you realize that, even if you win, you're essentially propping up, what I like to call, a "fluff market", instead of adding-value the way Silicon Valley and the manufacturing industry does.

Re: The Quants Run Wall Street Now

#68

Post by taeric »

10165 wrote: post #56
themodelplumber wrote: post #51

Earlier quoted context omitted.

How is this different from using 'links' or 'w3m'?

Using links is both better and easier. amp html pages look great in links.

This made me wonder if it would render well in emacs using eww. Surprisingly well rendered, actually. Odd that it doesn't have any pictures. But easy to read.

Re: The Quants Run Wall Street Now

#69

Post by tptacek »

SEJeff wrote: post #61
kasey_junk wrote: post #52

Earlier quoted context omitted.

> Virtu only lost money one day out of 1278 trading days Lost money trading . If they made $1 trading that counted though I assure you that day was a loser from a business perspective. I also have not lost money trading in 3 years, simply by not trading at all. It doesn't really detract from your broader point, which I don't have a strong opinion on but that stat is a pet peeve of mine as it's fairly meaningless.

This is massively incorrect. If Virtu only made $1 trading, they would still make massive amount of $$$ via exchange rebates, as they are a designated market maker. In fact, they aren't the only market maker that does this. A market maker guarantees they'll take trades on both sides of the book, so long as their strategies aren't grotesquely losing, they'll always come out ahead due to exchange rebates. http://www.in…

For whatever it's worth and on the off chance that this short-circuits a really common message board pathology, and allowing that I'm only commenting here because I read all of Kasey's comments: he's got HFT experience too.

I think he's just saying that particular metric is missing context: a lot of crappy businesses could claim never to have lost a dollar trading, while still as businesses losing tens or hundreds of millions of dollars chasing, and not finding, profits.

I don't think he's arguing specifically that Virtu lost money.

Re: The Quants Run Wall Street Now

#70

Post by finid »

zackmorris wrote: post #48

Anyone have hard numbers on if/by how much quants outperform old fashioned techniques like flags and finding stocks that tend to go the opposite direction from the one you're interested in? I don't know the terminology for all this but I know there are limits due to uncertainty so even the best algorithms may not do much better that someone guessing. Do quants do 10% better, 2x better, 10x better? My Dad and I had a…

Sorry to digress, I should have stopped at the first paragraph.

No, it's good digression, and it brings back fun and painful memories, as I made and lost money at that time.

Trading on margin is something rookies should avoid. It's the easiest way to lose money.

The thing I don't understand is this noise about shorting being unethical. You have more to lose than anybody in the game, and you're not breaking any rules.

Post reply on HN