Locust 

From my point of view 1000 trades per day per instrument is already well within the boundaries of HFT. Which contract are you talking about and what are the stats like avg. trade profit before cost and max position size. I was looking at a pretty simple approach based on 25 tick bars and positive pyramiding. Given a volatility level above 15% you are well in the green zone. Locust

I'm not trading 1000 per instrument. I just have a new strategy I'm playing around with that generates that in simulation. Like I said, I'll probably never implement it live. It generates a ton of scalps and isn't really worth it after commissions. Unless I can tweak it to grab a few extra ticks. I'm also concerned about something going haywire at that volume. Not to mention, I have this strong negative feeling about lining my broker with that much money per month. Something doesn't seem right about that!

The one I'm actually using live... Is a more realistic 100 round trips a day, but even that is really pushing it in my opinion. I'm currently fiddling with it to make it more efficient and not take so many trades per day. While it gets several decent sized winners, it gets a lot of small losses. If only I could figure out how to sit out those losers and still take the winners! That is the tricky part that I've been struggling with for weeks.

Right now it is about 50/50 with winners vs losers. Profit factor is about 1.7.. That is an average over the last 6 months. This is not very good in my opinion, but unlike my other strategies (that were 70%+ winners and 2 to 3 Profit factor), it doesn't require any optimizing. It reads the tape and makes the trades and I don't have to do a daily optimization. It feels more clean and less likely to have operator error (it took a ton of variables in the optimization process and if I put in the wrong number, I would get losses that day). This latest strategy doesn't have as good returns, but more or less runs on its own with no optimization needed.

I did a lot of research into HFT that is being done by the actual HFT firms (not guys sitting in their home offices - like me ). Somewhere, I read that their algorithms generated about a 50/50 win/loss ratio. So, I'm right in line with that. And okay with it. Of course, their volume is so high that they also get rebates. So, maybe 50/50 isn't that great, now that I think about it! lol

lookOutBelow 

I'm not trading 1000 per instrument. I just have a new strategy I'm playing around with that generates that in simulation. Like I said, I'll probably never implement it live. It generates a ton of scalps and isn't really worth it after commissions. Unless I can tweak it to grab a few extra ticks. I'm also concerned about something going haywire at that volume. Not to mention, I have this strong negative feeling about lining my broker with that much money per month. Something doesn't seem right about that! The one I'm actually using live... Is a more realistic 100 round trips a day, but even that is really pushing it in my opinion. I'm currently fiddling with it to make it more efficient and not take so many trades per day. While it gets several decent sized winners, it gets a lot of small losses. If only I could figure out how to sit out those losers and still take the winners! That is the tricky part that I've been struggling with for weeks. Right now it is about 50/50 with winners vs losers. Profit factor is about 1.7.. That is an average over the last 6 months. This is not very good in my opinion, but unlike my other strategies (that were 70%+ winners and 2 to 3 Profit factor), it doesn't require any optimizing. It reads the tape and makes the trades and I don't have to do a daily optimization. It feels more clean and less likely to have operator error (it took a ton of variables in the optimization process and if I put in the wrong number, I would get losses that day). This latest strategy doesn't have as good returns, but more or less runs on its own with no optimization needed.

LookOutBelow,

if you can get a 50/50 rate with profit factor of 1.7 (after commissions?) and 100 trades a day I consider that alright.
With the right position sizing/money management , this gives a good return if you can let it compound.

Are you familiar with the work of Van Tharp on the huge effects of position sizing on trading results?
If not then definitely have a look at his book "The Definitive Guide to Position Sizing".

I would not mind the fact that your broker gets a big chunk of the 'pie' and consider it a cost of doing business. Provided, of course, your piece of the 'pie' allows consistent equity growth.

The fact that the strategy is profitable without optimisation makes it a lot more robust. Have you tried it on different markets?
Do your 70% systems generate a comparable trade frequency?

Stevenvdv

stevenvdv 

LookOutBelow, if you can get a 50/50 rate with profit factor of 1.7 (after commissions?) and 100 trades a day I consider that alright. With the right position sizing/money management , this gives a good return if you can let it compound. Are you familiar with the work of Van Tharp on the huge effects of position sizing on trading results? If not then definitely have a look at his book "The Definitive Guide to Position Sizing". I would not mind the fact that your broker gets a big chunk of the 'pie' and consider it a cost of doing business. Provided, of course, your piece of the 'pie' allows consistent equity growth. The fact that the strategy is profitable without optimisation makes it a lot more robust. Have you tried it on different markets? Do your 70% systems generate a comparable trade frequency? Stevenvdv

Hi Steven,

I've heard of Van Tharp before, but I haven't seen that book. I'll have to look it up. Thanks.

The 1.7 prof factor is after commissions. As far as using this strategy on other instruments. I'm currently adapting it to ES and will move on to others in the future. It takes some tweaking, but it seems to be adapting well. I don't have any real figures yet. Not until I work through some kinks.

As far as my other strategy (the 70% one), it trades far less frequently. Like 1-10 RT's a day. It is on my list of things to tinker with. I think I can improve upon it from what I've discovered while building this newer strategy.

Thanks

HFT trading is for well capitalized accounts. Commissions are very low as is the ultra-low latency used.

Your HFT provides DMA or direct exchange connection software for trading. They would likely partner with several collocation vendors such Equinex, 7ticks or Guavatech to get a collocated service.

Here a two companies that will offer you that service:

1) First Futures Software Engineering Pvt. Ltd. - Products and Services for the Capital Market
2) Edge Financial Technologies

Here's the scoop -

You will have to open an account with a brokerage/FCM such as Advantage futures, NewEdge or Velocity Futures to use HFT software.

The company I spoke with provided 2 solutions for direct exchange connectivity and both solutions are aimed at algorithmic/ black/ gray box style of trading where microseconds or milliseconds make a big difference. Most spreading, or arbitrage style strategies need this.

1. A windows based solution with C# API, $3500 per month, month to month commitment.
2. A windows or Linux based solution with hardware acceleration, C# or C++ API. $5000 per month, month to month commitment.

Of course they will tell you that those solutions are faster than all known competitors- TT, CQG, ORC etc. using the following costly solutions -

Solution 1 latency is around 150 microseconds for both price message processing and order routing.
Solution 2 latency is around 60 microseconds (or less) for both price message processing and order routing.


Hope that helps to quiet a few restless souls.

COTtrader
"Ken"

Just to check you mean millisecond and not microsecond, right? I mix them up too.

COTtrader 

HFT trading is for well capitalized accounts. Commissions are very low as is the ultra-low latency used.

Good information. This is well beyond what I need at this point. As my strategies don't require low latency like that. At least any lower than what I have now (I'm using Ninja Trader / Zenfire presently).

Out of curiosity though, and this is what would interest me. Do you know what those commissions are? I'm sure they depend on volume. I'd love to see a rate schedule. But in my experience (haven't checked these guys specifically), they don't advertise that information on their websites.

MXASJ,

HFT trade black boxes that may execute 1,000 trades per second. Exchanges can process trades in less than 500 microseconds (or millionths of a second not one thousandth of a second).

Millisecond = One Thousandth of a second
Microsecond = One Millionth of a second

Microsecond trading is Flash Trading.

COTtrader

MXASJ 

Just to check you mean millisecond and not microsecond, right? I mix them up too.

for what it is worth... HFT does not mean Low Latency Trading(microseconds), but it is always associated with it mainly because you need LLT depending on your strategy... i.e. if you are collecting rebates from the exchange, you need to buy/sell within the same ms .. a retail trader can do HFT with decent latency (sub 2ms), but fees will kill you given that means you are hyper scalping more than likely.. so your broker makes more $$$ than you do...however, that does not mean that you cant find a broker that will cater to that type of investor, and eventually even do all the work for you to lease a seat to lower your transactions even further...

I see value in "Affordable" low latency to both my market data and order execution... most traders will get 50-100ms latency... worst for overseas... so if you can get way less (0.5ms to 2ms) with an automated method and see the same information they see before they do and trade before they do then you gain an edge against those other retail traders, but not against institutional ones... they have the funds to get what they need given their trade sizes.

from my own experience, since I am in IT with financial services and I have in fact designed trading floors and the back end infrastructure required for receiving market data from exchanges(equities) to build ones own books, and I have buddies that have designed the exchange infrastructure (mainly AMEX/NYSE) from end to end; I can tell you that besides the know-how, the only other thing that gives an institutional trader the edge is the technology their company invests on, which is why there is a "race" in those terms within each financial company and no-one shares publicly what they are doing and researching unless they want to be a target...

A retail trader would normally not have the amount of capital to spend on said resources, unless he has a few millions to trade with and it is then warranted, but he can gain a technological advantage just like a pro by investing his resources wisely... btw, I do believe that no amount of technology will make you a profitable trader...

if you have a decent futures strategy that works and can be automated and does a high frequency of trades (> 500+ per day) .. you can colo on The Lakeside Technology Center (350 East Cermak) for anywhere from $100-$5000+ per month depending on the provider and if there is direct connectivity (lesser hops = lower latency)...

you can lower your transaction costs by leasing a seat, or work with an FCM that will give you one commission structure for all your trades i.e. advantage futures and mann(aka MFG) ... their comissions can be as low as $0.03 comission depending on your volume... but that wont be your greatest cost savings, the bulk of your costs are on exchange fees... so getting a seat lease will give you more profits to keep... note that I dont mean an Introducing Broker, but dealing instead directly with the FCM.

Don Miller has a good spreadsheet analysis on his blog that tries to build a case for leasing a seat...

Commission Rates :: Don Miller's S&P Trading Tank

and lastly, but not least, the quality and reliability of your market data feed .... broker provider wont suffice IMO ... IQFeed is good for retail... Reuters is best if you can afford it...

just my 2 cents...

Humans cant trade HFT, you can bactest anything, it can show great profit, but real trading is more complex. You want to belive in that you can close your emotions to a black box, but you cant. And the more trades you do, the more emotional pressure will be, and bacause of that you will not capture the winnings, and you will be in all the losers. Another fact is the filling precentage, and another is the execution speed both human and on connection side. I trade 20-30 /per day, i think 40-50 is the max, more than that is just like you tring to be an algorithm, a computer, with 0speed, 0 emotion.

How do exchanges know if an order was executed by an algorithm or manually by a trader? How do they get these statistics about how much volume was accumulated by algorithms or HFTs?