Forums  > Trading  > Virtu  
     
Page 1 of 1
Display using:  

fastmoney


Total Posts: 3
Joined: Mar 2017
 
Posted: 2017-03-11 07:38
I work in quant trading doing intraday stat-arb (~hours) in equities for a well-known firm. I'm studying competitors and can get my head around most of their business/trading styles, but Virtu is one I have trouble understanding.

On the face of it, it seems like what they do is so simplistic that their edge should be competed away, yet they deliver extremely consistent profits. This makes me question whether seeking quant alpha is worthwhile or whether my understanding of their style is right.

My understanding is that they don't hold unhedged risk, engage in esoteric statistical hedging, or try to predict price movements. This is obvious when you see they never have a down day and their CEO claims they don't do hedges he can't understand. They don't employ many quant researchers compared to competitors like HRT or Tower.

Some of the guys I work with were around for the halcyon days of HFT and tell stories of running mechanical strategies that made a ton of money, but I don't know many people doing that today. The economics of these trades are obvious, but the real question is why one firm dominates them while others are doing more predictive modeling, longer holding period, slightly lower Sharpe strategies. I can think of a few explanations, but not very satisfying ones:

-They are faster than anyone else.

-They trade so much volume and get the best fees, lowest clearing costs, etc.

-They have some orthogonal non-quanty understanding of limit order book/flow dynamics that other firms lack.

-They have access to more markets than competitors and can identify more mispricings this way.

-Running the entire business focused on simple trades without silos lets them extract economies of scale of some sort vs. firms where small trading teams run a small set of products with different trading styles.

-Other firms simply chose to play a different game because they believe these trades are unsustainable. A predictive edge doesn't require as many expensive investments in technology compared to an arbitrage/scalping edge.

rftx713


Total Posts: 73
Joined: May 2016
 
Posted: 2017-03-11 17:46
I'm a bit confused by that +0 strategy mentioned in the excerpt. Would that require being at the front of the queue with both a buy and a sell order, and then making a cancellation decision on the opposite order type if it doesn't tick up (or if models predict a downtick)? Is the article suggesting that the order types facilitated this or that this is simply a strategy that opens up when you have faster systems on average than the rest of the market?

a路径积分


Total Posts: 77
Joined: Dec 2014
 
Posted: 2017-03-12 02:18
*repeat post*

a路径积分


Total Posts: 77
Joined: Dec 2014
 
Posted: 2017-03-12 02:18
*repeat post*

a路径积分


Total Posts: 77
Joined: Dec 2014
 
Posted: 2017-03-12 02:18
One of the funny things in this space is that people always regress success against trading skill, knowledge, experience, secret sauce, speed, whatever technical attribute you can find.

To be honest, one of the biggest drivers of Virtu's success was that Vinnie was a formidable leader, who managed to raise an incredible amount of funding, and he picked the right person to build the tech stack whose technical prowess exceeded Mark's starting 10. Long before Silicon Valley became hip or universities started teaching quantitative finance, he had a knack for plucking talent from Harvard, bringing them out to the west coast where he turned them into cash-generating automatons.

fastmoney


Total Posts: 3
Joined: Mar 2017
 
Posted: 2017-03-12 19:16
"I'm a bit confused by that +0 strategy mentioned in the excerpt. Would that require being at the front of the queue with both a buy and a sell order, and then making a cancellation decision on the opposite order type if it doesn't tick up (or if models predict a downtick)? Is the article suggesting that the order types facilitated this or that this is simply a strategy that opens up when you have faster systems on average than the rest of the market?"

According to the link, this originally started in futures markets without special order types. Figuring out when to be in the market, GTFO or puke is unclear and you can't just hand-wave away those decisions. It's very light on details and put forth by a guy who thinks order types were some big conspiracy, so I take it with a big grain of salt. It just seemed superficially similar.

I'm not even sure that it could work even at no latency, let alone just faster than average. Even if you are ahead of depth, someone can take out the entire stack at which point you're in the hole instantly. No amount of speed can prevent that. Is the risk/reward even favorable? Why would an inefficiency like this continue to exist rather than being competed away?

fastmoney


Total Posts: 3
Joined: Mar 2017
 
Posted: 2017-03-12 19:35
"One of the funny things in this space is that people always regress success against trading skill, knowledge, experience, secret sauce, speed, whatever technical attribute you can find."

Do you mean it's wrong to assume that someone is more successful because they're far superior in one dimension vs. small advantages in many dimensions?

What strikes me as funny is that firms with approaches that are almost polar opposites can both be very successful. Many firms only do predictive modeling and rarely make two-way prices. You don't see that in most industries. Usually competitors become more and more alike over time. In trading it seems like the best end up finding a very specific niche and dominating it, almost willfully blind to what others are doing.

NeroTulip


Total Posts: 995
Joined: May 2004
 
Posted: 2017-03-13 01:24
I wouldn't say polar opposites, it's more like they are orthogonal.

"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)

a路径积分


Total Posts: 77
Joined: Dec 2014
 
Posted: 2017-03-13 04:57
No, I mean it's naive to assume that when someone beats you at low latency trading, it means that they are faster or smarter than you, and people outside our scope of work always seem surprised when someone says that their differentiator is something simple, like better leadership.

jslade


Total Posts: 1064
Joined: Feb 2007
 
Posted: 2017-03-13 05:30
Haim Bodek is an ass. 0+ is simply describing what market makers do, using indicators built on the order book. OMFG, I used probability theory, good engineering practice and colos instead of ... the Riemann zeta function and the Zohar. Obviously I am guilty of great sins. No fair no fair!

+1 on Virtu having good leadership. While luck and timing is important (RGM, Getco), if there is one thing I've learned about ventures, the human factor is pretty much everything that isn't luck and timing.

"Learning, n. The kind of ignorance distinguishing the studious."

a路径积分


Total Posts: 77
Joined: Dec 2014
 
Posted: 2017-03-13 07:28
I think I met Haim at an event in Tribeca once. He sports a bald hairdo right? He seemed like a sweet guy. Now Hunsader is a funny one. For some reason he felt that voting for DJT was going to wipe out the influence of large banks and EMM firms, and of course they installed Giancarlo as chairman.

EspressoLover


Total Posts: 221
Joined: Jan 2015
 
Posted: 2017-03-13 10:45
> The economics of these trades are obvious, but the real question is why one firm dominates them while others are doing more predictive modeling, longer holding period, slightly lower Sharpe strategies

What makes you think Virtu dominates? You don't think Citadel, KCG, Jump, etc. have O(100mn) strategies with O(0) down days?

I suppose Virtu's relatively unique in that they don't even seem to try to do the quant-y stuff. Maybe my understanding of the firm's off, but they only care about doing purely mechanical strategies. Rejecting the quant stuff seems more about business focus, rather than some sort of secret sauce. Chasing quants can get really expensive really quick, and that style tends to go through more hit-or-miss periods.

Plus Virtu's reputed corporate culture isn't well suited for high-level alpha work. You can slave drive people 16 hours a day, and the market data parsers will still get written. But burning people out doesn't lend itself to deeper creative work. The best quant firms tend to be like RennTech, more university department than investment bank.

radikal


Total Posts: 253
Joined: Dec 2012
 
Posted: 2017-03-14 00:51
I respect quite strongly what they've accomplished, though most of what's said here re culture/strats is fair.

I wouldn't discount how powerful they're omnipresence in every pool, exchange, ecn, etc has been for them. Everyone is at a LOT of places, but Virtu has been the most aggressive.

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

EspressoLover


Total Posts: 221
Joined: Jan 2015
 
Posted: 2017-03-14 08:28
> Everyone is at a LOT of places, but Virtu has been the most aggressive.

Which brings up my biggest quandary about Virtu. Why the hell don't they run an internalization pool? (At least as far as I can tell they don't, and if they do it's either not under their name or not big enough to be in league tables). Given that they have such high market share in equities, wouldn't it make sense to wholesale retail order flow? Internalization is already barely any different than market making on lit venues. The major barriers are institutional clout and infrastructure, which Virtu has in spades.

The only explanation I can think of is that they can't match SIG, Two Sigma, ATD, et al's payment-for-order flow rates. Without alphas they can't price improve enough to win any contracts from major brokers. Still you'd think they'd be able to grab at least some of the retail wholesale market. It'd be worth it to make some push. Dark flow is just so much more profitable than lit flow.

a路径积分


Total Posts: 77
Joined: Dec 2014
 
Posted: 2017-03-16 02:31
EL, there could be other options, such as buying KCG.

https://www.wsj.com/articles/virtu-financial-makes-bid-to-acquire-kcg-holdings-1489609032

radikal


Total Posts: 253
Joined: Dec 2012
 
Posted: 2017-03-23 01:20
@EL -- that's a really great point. I'm not I've got much insight; too much time spent watching the robots.

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

katastrofa


Total Posts: 357
Joined: Jul 2008
 
Posted: 2017-03-28 13:16
I don't think they're buying KCG for their ex-Getco business.

HitmanH


Total Posts: 422
Joined: Apr 2005
 
Posted: 2017-03-28 23:29
Agreed K.

Virtu's business model - and as many have said - they have excellent leadership - has been based not on the best alpha/prop models - but the widest flow in the assets that they trade.
They're buying KCG for the ex-Knight business model - of connections to so many tier-2 brokers, to other sources of flow etc - that they can offload very quick and efficiently - and that is that
(yes - some synergies in terms of comms networks - but that is secondary)

xyz1453


Total Posts: 2
Joined: Aug 2014
 
Posted: 2017-04-19 15:05
Interesting read: Insights Into High Frequency Trading From The Virtu
http://online.wsj.com/public/resources/documents/VirtuOverview.pdf
http://www.valuewalk.com/2014/11/high-frequency-traders-virtu/

kabakkaya


Total Posts: 1
Joined: Apr 2017
 
Posted: 2017-04-29 20:38
For those that are interested, here is a decent write-up of what really went on during the early days of Virtu and how it actually came to be:

http://www.amsterdamtrader.com/2014/02/tibra-stole-source-code-getco.html#comment-19465

katastrofa


Total Posts: 357
Joined: Jul 2008
 
Posted: 2017-05-01 21:17
Getco had the same idea when they bought Knight: "we will take the franchise and move it to our wonderful trading platform" ;-)
Previous Thread :: Next Thread 
Page 1 of 1