HFT Myths

[volpunter]

awesome? Lol, while I do not agree with many of Michael Lewis's comments, how is this an awesome discussion when you sit two hft poster children together to hype their business? It is like sitting 2 democrats or 2 republicans together to discuss politics. Awesome!!!

P.S.: By the way Hyme is the guy who was fired because his firm went bust. Of course he is bitching about the big hft firms that put his mid sized firm out of business, but at the same time he is whoring himself out as hft proponent because that is what feeds his family right now as publicity poster boy as well as to hedge his future prospects. Talk about being a hypocrite.

http://insider.thomsonreuters.com/link.html?cn=share&cid=1222178&shareToken=MzpmMDg0MTEyMC01NWY5LTRjZDItYjQzOC1iNTNiMzAyMDJhN2I=

Awesome discussion between Haim Bodek and Manoj Narang on HFT. No sensationalism, just facts presented by practitioners.

 

[hft]

There are lots of literatures on applying Stochastic Control Theory to market making. Do you know firm that has successfully apply Stochastic Control Theory to market making?

Not personally, though I'm not as heavily into the pure quant-driven side of HFT.

 

[hft]

hft

To what extent do you guys use machine learning and AI in your systems vs using static rules?
for example, earlier you mentioned comparisons between the bid and the ask size to forecast the next tick with a given level of accuracy. To what extend would you use a fixed interpretation (say, if the bid is 300% of the ask, then the tick must go up) vs using an ML algo to let the system figure that out on its own?

It's almost always a combination of both ML and human-driven adjustments ("gray box"). Never truly static - the only static parameters are really defining the manner in which the ML system trains and adjusts. Some are truly black box and ML only, but that probably sounds more complex than it usually is. For instance, my black box algo lost money today because of this event, so I'm going to put in a filter to react accordingly in this type of situation in the future. Now it's still 'ML/black-box', but really just an automation of human input.

 

[hft]

90% of scratches?Do the strategies use some sort of the offset orders routine?

Yeah that's pretty common. Try to get in front of the line, if you don't like your fill then just flatten at the same price you were just filled at.

 

[hft]

what do you mean with "your feed"? Are you a liquidity provider? And if yes on which platform/ECN?

Thanks

This was in reference to quoting directly to big banks (GS/JPM/etc) through an ECN like Currenex. The bank would get my feed as a stand-alone endpoint to trade against.

 

[hft]

that would never happen "naturally". Remember that the single reason why futures trading is centralized is REGULATION.

Futures trading consolidation isn't due to regulation. For example, you can trade currency futures on CME as well as ICE and others, but people naturally choose to trade at CME, making the network/distance issues moot as compared to equities. It makes trading quite simpler. There are other issues with futures trading that, by definition, mimic the equities fragmentation problem though: the underlying(s). So yeah, all currency futures are traded at CME, but now you have a host of spot FX endpoints to worry about. The problem isn't quite so bad for most retail traders though.

My point was that equities consolidation won't happen naturally as you said. But how do you determine who gets to capture all equities trading? Scrap all of them and have a government-ran matching engine, hah? RegNMS was designed to even out the playing field for retail, but it's an absolute mess that just managed to make an already complex problem even more complex.

 

[stock777]

hft talk a good game but how much of it is nothing but pure frontrunning.

 

[Brahmastra]

hft talk a good game but how much of it is nothing but pure frontrunning.

Your a plonker!

 

[volpunter]

without putting my signature under anything Michael Lewis says...after having read up a lot on the exchange/ecn specifics of IEX looks to be to be the currently fairest and best approach to how to facilitate the exchange of risk between buyers and sellers of risk. And yes, we need again one centralized market place where everyone meets and competes for best price. How this is done I do not care much about, it could be an auction system, it could be a shared operation between the major exchanges but it needs to get done at some point. The equity markets at the moment in the US are really not much different from the wild west of spot fx when you think of bucket shops and "rogue brokers" that basically manipulate and run their own feeds. Yes there are subtle and important differences between that and equity exchanges but there is also a lot of overlap. We need to again establish a centralized market place.

Futures trading consolidation isn't due to regulation. For example, you can trade currency futures on CME as well as ICE and others, but people naturally choose to trade at CME, making the network/distance issues moot as compared to equities. It makes trading quite simpler. There are other issues with futures trading that, by definition, mimic the equities fragmentation problem though: the underlying(s). So yeah, all currency futures are traded at CME, but now you have a host of spot FX endpoints to worry about. The problem isn't quite so bad for most retail traders though.

My point was that equities consolidation won't happen naturally as you said. But how do you determine who gets to capture all equities trading? Scrap all of them and have a government-ran matching engine, hah? RegNMS was designed to even out the playing field for retail, but it's an absolute mess that just managed to make an already complex problem even more complex.

 

[Brahmastra]

One of the myths,that algos affect trading metamorphoses into theory suggesting that the manual trading is greatly disturbs the algo trading!So,message to the retailers is,stop fking bothering my algo trading with you S/R bolshet!!!