HFT Myths

Thanks for taking questions on this stuff! Have you followed the academic research on HFT, and do you have any opinions on it?

One particular question I have, there is a paper by Baron Brogaard and Kirilenko that use S&P 500 futures data to estimate how much HFTs make trading the ES. They have CFTC account data that lets them figure out who the HFTs are by trading patterns. Do you think this approach makes sense, or are they making statements on their profits based on looking at trades that are often one piece of a multileg strategy? Maybe hedges against stock or option trades, or one side of spreads against some other index futures contract?
 
Quote from jb514:

I'm not a futures trader so bear with me.

Would you say your, and other hft market making strategies apply more towards products that trade like barcodes like the 30 year, as opposed to a more whippy product like crude oil.

How often is the position you have on important to you? If you are long are you immediately adjusting the strategy to try and flatten out? Or does the strategy behave the same whether you have plus or minus 25 contracts and only starts to get biased towards flattening out when you have say +- 50 contracts?
More...

HFT's are very active in both types of products you mentioned. I'm not sure what barcode means but I assume it's slow and thick like ZB vs. the fast and thin markets like oil.

As far as position factors, it really varies across the board among different strategies. There are times when you bias immediately, or bias heavily upon a certain threshold, and lots of other variations of risk management models. As you would assume, most models aren't so concrete on either end of the spectrum and lie somewhere in the middle.
 
Quote from black diamond:

Thanks for taking questions on this stuff! Have you followed the academic research on HFT, and do you have any opinions on it?

One particular question I have, there is a paper by Baron Brogaard and Kirilenko that use S&P 500 futures data to estimate how much HFTs make trading the ES. They have CFTC account data that lets them figure out who the HFTs are by trading patterns. Do you think this approach makes sense, or are they making statements on their profits based on looking at trades that are often one piece of a multileg strategy? Maybe hedges against stock or option trades, or one side of spreads against some other index futures contract?
More...

I have not read it but will take a look. My first question is 'why bother'. My second impression is that it's a bunch of voodoo data mining for data mining's sake.

In answer to your specific question about the trade analysis, yes I wouldn't be surprised if a large portion of that ('large' being subjective, but probably enough to throw off any numbers they might be obtaining) could be multi-leg strategies. Then again I wouldn't be surprised if it was the opposite either! My point is that I can't fathom a way to possibly tell.

In any case, I am interested in how they can possibly come to any conclusions on the above, so I will take a quick read through and get back to you.
 
HFT you suggested that both futures and options are something that somewhat less funded (slower) firms might focus on.

Futures make sense because they are traded in one place so especially if there is no good off exchange hedge, you dont really need fast interconnect (microwaves)

However most us stock options are now tradeable on many exchanges and are clearly priced off of stocks traded on many exchanges. Why did you say options ?
 
Quote from bidask201:

HFT you suggested that both futures and options are something that somewhat less funded (slower) firms might focus on.

Futures make sense because they are traded in one place so especially if there is no good off exchange hedge, you dont really need fast interconnect (microwaves)

However most us stock options are now tradeable on many exchanges and are clearly priced off of stocks traded on many exchanges. Why did you say options ?
More...
I stand corrected. I don't have much experience with options, but I was thinking of ES options on CME in particular. Or even something like SPY options trading where there's a lot more pricing and quantitative analytics involved than futures/equity trading that's so latency dependent.

You're right though, HFT equity options trading seems like it would be hopeless without good intercolocation connectivity.
 
Futures traders need not be concerned about HFT outfits. Some adverts claim the HFT firms are WHY the small traders in futures (and maybe stocks) can not compete because the poor trader is still using a mouse and turtle slow order entry/exit. Pity the small trader compared to the BIG DOGS controlling the game. Well, boys and girls do not fall for that dribble from those that want to sell you THEIR bologna.

This has been a good thread to read...........good stuff. But indeed, the skills necessary to win in a probability game of chance will only come from your skill sets developed in the battleground of the market itself..........anything else is just spinning your wheels and a waste of time and money.

Be yourself, be brave, depend on no one else because everyone else is out for your money. :)

HFT firms are not going to upset the small trader because the small trader can pick his/her spots and do what they do best.
 
hft,

Can you describe the general features of what hardware specs would go into a co-located execution server (i.e. 10GbE, FPGA, typical OS, proc type, core count, etc.) I'm guessing you would be hesitant to go into exact detail, but I am wondering what would go into the $5k-$10k hypothetical server that you described in prior posts.

Thanks for all the effort that you have put into responding to everyone's posts. It has been a great read.
 
Quote from tcdooren:

hft,

Can you describe the general features of what hardware specs would go into a co-located execution server (i.e. 10GbE, FPGA, typical OS, proc type, core count, etc.) I'm guessing you would be hesitant to go into exact detail, but I am wondering what would go into the $5k-$10k hypothetical server that you described in prior posts.

Thanks for all the effort that you have put into responding to everyone's posts. It has been a great read.
More...
Just go to Dell and customize an R620 with the highest end CPU(s) and that's in the $5K-$10K general price range without any super customizations. 10G, centOS, 4/8/16 core arrangements are all common but nothing groundbreaking. FPGA's are another story altogether, but the cost there is primarily in development and utilization as opposed to hardware cost.

I don't think any HFT firm really frets about its server cost (with the exception of equities firms running hundreds of servers at a colocation). Any strategy that's worth anything would pay for its server usage many times over. In fact, it probably costs me more in IT personnel costs to deploy/maintain a server than the cost of a server itself.
 
Quote from Drifter:

What is your opinion on news feeds and do you use them?
More...
I listen to a squawk box for entertainment but it really doesn't help since I don't make any intraday adjustments. Ransquawk is excellent.

News feeds for direct usage (NTKN, Reuters, DJ) are a different game and very interesting. Some people are making boatloads of money using them but I haven't really gotten into that space yet.
 
You must log in or register to reply here.
Back
Top