When I say tap reading I mean the action of price and volume. I look at every quote and every print, trying to spot an aggressive buyer or seller. TS monitors about 500 NYSE stocks tick-by-tick, and when it spots a known pattern, it trades. It also collects every tick (actually, "the result tape") for my after-market analysis, and I do back-test TS based on historical (again, tick-by-tick) data to find/confirm patterns.
Thanks for your reply spec. Is there a web site for this TS? Also if you have a database of each TICK!!! wow the size must be incredible! How much data far back do you look back when you do your backtest.
There is no such site. I was gathering information bit by bit. Tape reading is a lost art, and it's very hard to do it these days with electronic market. There is almost zero information how specialists work. Those who claim that they do tape reading by looking at T&S don't even see a half of the picture. There some excellent threads about tape reading on ET: http://www.elitetrader.com/vb/showthread.php?s=&threadid=56373&highlight=tape+reading http://www.elitetrader.com/vb/showthread.php?s=&threadid=3282&perpage=6&pagenumber=1 NYSE rules (http://rules.nyse.com) is an excellent source for those who trade on that exchange! And of course, the most important part - huge amount of time spent on observing the tape in real-time and after-market. I don't have a "huge" database (I think it's just several GB). Every day I pick up stocks with interesting behavior (news, large activity, etc), and download all ticks for that day (I use ACTIV financial for market data). A couple of words about TS itself. I chose to code it by myself because there is no software/API that is suitable for tape reading. In addition, I have full control over the code, and my system is really fast and scalable - I can monitor every tick for pretty much all stocks on NYSE on my dual-core x64 AMD with a moderate load... It's freaking hard work, but it's fun to do something that pretty much nobody does
NYSE ticks/second = message sent to you of x bytes and processed: May 2006 - 4,000/sec Jan 2007 - 7,000/sec Jun 2007 - 24,000/sec That's about 500,000,000 ticks/day... Which might be 5 or 10 or 20 GB sent over your internet connection... To process in real-time every day. A LOT of serious firms are doing this... Probably many 100s. They are using high end data vendors like Thomson or Reuters or Bloomberg... And dedicated lines and systems... But, IMO, it's pretty much pointless overkill. The best approach is always to SPECIALIZE in a market niche of 50 or 100 or 200 stocks... Apply CLASSIC arb techniques... And become the BEST at what you do. I would trade ANYONE in the world... Head-to-head... In my specialty... And be confident of winning.
The last time I measured load it was about 6K update messages per second on about 2K symbols. Not a peak for my system, but of course, I do not do it on daily basis. I agree it's overkill to scan the whole market, I specialize on NYSE only stocks that do 1M-5M avg volume daily. That's about 400 symbols where I look for any aggressive player. That's my edge.
I do the opposite... I trade about 400 stocks that average 5K to 100K daily volume... And probably trade 50 to 100 of these on any given day. The volume adds up. Nobody gives a sh*t about these stocks... Including Specialist or MM firm... Which views these issues as a pain in the butt to manage. That's my edge. So I could easily maintain a tick-by-tick database for my stocks... And it would be roughly 1% of total NYSE daily ticks... Or a very manageable 5,000,000 messages/day. But I can't imagine why I would do this... Unless I had a $500,000 annual tech budget... To blow on exotic, ultra-low margin black box algo trading. I will probably go that route within a year or two... Because I'm a software engineer turned trader... But this is really very sophisticated territory... Nothing trivial about it at all. Best of luck with your work... What you are doing is very cool... Kind of applying leading edge techniques to Old School methodologies.
Hmm, good for you if you know how to trade those stocks - spread is huge, price may drop/spike on a zero volume, and those stocks rarely do good moves. In my tape reading the volume is paramount - it shows the commitment level of the buyer/seller. On the other hand, stocks over 5-6M too hard to follow since prints are all over the place. Just my 2 cents... Cannot say much about MM, but the specialist can be your friend as long as you trade with him on the same side - there is still some flow coming from floor brokers and the specialist sees it. We are on the same page here
A 50,000 average volume stock... actually trades very efficiently 90% of the time. Price spikes in my stocks... Which have NOT been chosen randomly... Are usually a sign of market inefficiency... And, very selectively, using professional judgement... An opportunity to exploit. You need a good spread in your favor to double your Quantitative Edge... I flip for $0.04 to $0.07 all day long... And am paying about $0.006 round trip Once you get above 100,000 average volume... You begin to see a lot of black box trading... And if you wanna be professional about it... You gotta be running algos... With a minimum tech budget of about 200K/year... (Which buys you 2 engineers + infrastructure)... So you really need profits of 500K to 1000K annually... To support even a low-budget Algo Operation. Scalping is far more optimal than playing "moves"... Because while waiting for a $1.00 move that may or may not happen... Or will go against you 45% of the time... You can always scalp for $0.05 dozens of times... Regardless of whether it ever moves or not. The Specialist or MM is never anyone's friend. And I don't care how the MM is manipulating the market today... Because it's always a short term effect. So it's either a small cost of doing business... Or it works to my advantage when things revert to mean... Overall... I'm 100% sure that I've been taking real money away from the MM for 15 years. The MMs ** know ** that we are Mortal Adversaries... And have been f*cking with my orders since Day One. The changes of the last 6 months... With more market fragmentation... Have definitely made it MORE profitable for me.
I've designed my system to keep the position as long as thee is strong buyer/seller there. It may be $0.10 or $3 move. I found it useful to watch large number of stocks because sometimes a dull, inactive stock makes 2 point move during market hours because of news, and I want to get involved as early as I see it! Again, cannot say much about MM - I never traded NASDAQ, but the specialist can be observed and followed - those guys know what they are doing!
I see this is a trend of software engineers going to trading I'm there too, although I'm in the very beginning of the transition Speaking of which, can we share some [c++] code? I'd be interested in that FIX layer you talked about. I can give in exchange zero-locking threading class... or may be something else? Drop me a couple lines if you are interested. Thanks!