Tuesday, September 04, 2007

NYSE and Program Trading

This is part 2 of a series of posts started Sept. 1st on order flow dynamics. This post covers program trading (PT) .. that portion of trades accomplished through automatic executions driven by set buy/sell algorithms. Also see http://www.programtrading.com/ for more info.
A couple things strike me about this data:
Volume of trades - PT accounted for fully 80% of the NYSE trades during August. That's a lot. One thing to keep in mind is that there is no single trading algorithm used to trigger trades, but rather thousands of them, based on a variety of parameters including support/resistance, time of time, technical indicator signals, arbitrage with the futures, pair trades, etc. And most of these algorithms are dynamic and scalable, ie., triggers at 6:30am don't resemble triggers at 7:00am. What PT takes out of the equation is the foible of human emotion trying to second guess or finesse a trade. Nothing personal, just good risk management. And the reason there are so many PTs is simple mathematics . . . if you want to buy 250,000 shares of WM and do it at 100-200 shares per trade so as not to show your intent, it's going to take about 1600 trades.
Size of average trade - this is interesting because it is only 2/3 the size of the average retail trade. Where are all those 1000, 10,000 and 50,000 share orders? It's hard to tell and program trading is designed specifically to make it that way. There's a very successful local company (San Clemente,CA) that markets and manages software called "The Executioner". The clients are those trading more than 1M shares a day (you get a price break if you trade more than 10M). The software is designed to take a 1M share order and break it down into 100 and 200 share orders and trickle them into the pipeline at the most advantageous time. There is similar software available from a number of vendors designed to accomplish the same ends, although you don't read about it Active Trader or Technical Analysis of Stocks and Commodities. The net effect is little disruption to the price . . which benefits both big buyers, who don't care to show their hand and pump the price, and big sellers, who don't want to show their hand and drive the price down. Front running orders has become a lot more complex lately unless you've got a Sun8000 computer and some very sophisticated (and pricey) software. So, as you sit in front of your trading screen, watching level 2 quotes flow by displaying an apparently endless stream of 100 share orders, you're really watching a flood of program trades interlaced with a trickle of few others.
Implications - I routinely use the TICK, NYAD, VIX and volume to gauge short term momentum and trend strength because I assume the specialists and market makers (MM) can't manipulate these numbers. This is a popular view among many bloggers and traders. Trying to determine market strength by looking at L1 or L2 data is not practical, since MM aren't obligated to show their true order size - the same option I have available to me as a retail trader. So I'm always looking for a clue, an edge, an indicator, a tell that the MM can't finagle and that will provide me an undistorted peek at the true character of the market. I rely heavily on the MACD of the TICK and NYAD to guide my day trading (along with pivot targets), and while it is hard to fudge the NYAD I am now rethinking the effectiveness of the TICK indicator based on this current research. Dr. Brett's hybrid TICK/volume indicator looks like it still has legs, and I will explore this issue further in coming posts.


Christian Gross said...


As I pointed out in TraderFeed what you are seeing is called the IceBerg order. My broker has it. It is an order that shows up on the order books as 100 to 200 shares, but is executed MUCH MUCH larger.

The exact stats never show up on the order book, but they do show up AFTER the fact in the exchanges.

Thus as I pointed out in TraderFeed's post, volume at the time of trading is pretty irrelevant. Volume AFTER the fact is still relevant, but because it is AFTER the fact the information is pretty uses.

bzbtrader said...

I mention this issue in the Implications paragraph. It's also my understanding that an "iceberg" order refers to masking the size of the bid and offer volume, relative to the true size of the position intended to trade. However, the L2 time and sales data are true numbers, or the SEC would view the manipulated data as fraud.

Christian Gross said...

You are absolutely correct. The problem, and this is what people have told me is that if you want to track the true sales and data you better have a computer sitting beside the exchange.

People tell me that they want to start buring software in EEPROM to speed things up even more.

So what this tells me as an overall answer is that unless you happen to be an institution following volume is going to generate more noise.

BUT that is my opinion.