If a seat is available you buy one, it just comes down to a relatively straight-forward accounting breakdown and is usually a no-brainer. We still trade through brokers for certain markets where there are no 'seats' though (EBS for instance).
I know you're not primarily an equities trader but I've got a question I'm sure you could help with. I think I have a pretty good grasp on MMing strategies on a lot of the thick liquid names like BAC or MSFT but I don't really understand what kind of signals hfters are looking at when quoting symbols liquid or illiquid with wider spreads like AAPL PCLN GOOG PCYC ESS. I don't really understand the edges that your competing over since queue position isn't so important since you can improve by a penny or hide in front of other quotes. And you can't really scratch a trade like you can on the 1 tick spread symbols. It seems like the alpha signals are kind of vague. Can you shed any light on this? Thanks for your help so far.
From what I do know about equities (again, caveat that it isn't too much), the alphas have less to do with the primary exchange's microstructure (bid/ask volume, trades) than the lead/lag relationship between quotes from other exchanges. I.E. if the latest quote from ARCA is X ticks better/worse and came Y microseconds ago, can/should I alter my market in NSDQ.
For instance in futures market (e.g. CME), if you're not a MM your limit orders are filled depending on your queue position (MM have some pro rata fills guaranteed). Is there a way to know your queue position or to improve your position? How can you estimate if your limit order will be filled? Tks
How do you protect you code inside the company? if someone will be able to look at your code then they can replicate the strategy and your profits will go down?
You can't know for sure on CME. You just make your best guess at it. Keep track of the size ahead of you when you place your order, then whenever the size decreases assume some percentage of people cancelled their order ahead of you vs. behind you. You can overlay all sorts of latency/queuePosition models onto this estimate, but it's still just an estimate.
There's all sorts of ways to encrypt/permission files. Some firms are wide open (less and less these days). Most just use basic linux file permission systems and standard source control. Believe it or not, there's still quite a gap to bridge between source code and profitable implementation. *Implementation* carries with it so many other variables than the code itself. Nevertheless, of course it's possible to replicate this stuff and has been done, both legally and illegally over recent years.
very interesting thread, especially as I work at one of the other firms you mentioned earlier on although personally I'm more interested in what types of MFT strategies you'd consider viable for the at home trader either way, mainly just commenting so I can subscribe to the thread to keep up with what else might be added cheers
Hi, thanks for sharing and taking the time to answer these questions. I have a couple questions: Are HFTs or Algos able to jump the order queue on the CME? Can my order be jumped even if I place my limit order before the HFT algo? How long are HFT orders sitting at the exchange for before they are executed? e.g Do they usually place orders 1 tick away from the bid/ask? 1min? 5min? Thanks
Thanks for the sharing. Could you please give an insight on the following.. Suppose one has a simple system (like you explained on the first pages).. lets say arbing 2 instruments traded on the same CME engine group.. And one tries to capture some very simple microstructure inefficiencies by sending 2 orders down the CME pipe (from aurora collocated box) in 25 microsec (tick to order time - Bid/Ask change message came from CME, signal calculated, order created and sent out) And given the whole project startup cost like $2K (for the box and 10gb network card). Would you say this kind of project is competitive on the HFT arena? Would adding 300 microsec for margining from one the broker side ruin all the chances or will it still have a chance to capture few ticks a day?