Friday, June 12, 2009

Mailbag: Bot Trading Manages Risk at the Open

The Open and its frenzied activity can reward as well as punish traders. The difference between these two groups comes down to experience certainly, but also the insight and quick decisions that new tools (especially automation) can bring to bear.

From the Mailbag (Support Forum) came a question about such topics that I want to answer here. First the question then our response:
Mean Reversion and poor fills:

I'm using a mean reversion strategy that looks for stocks with spreads >9c. I have been automatically placing limit orders at bid+1c at the open, and have been getting killed by adverse selection.

If the stock moves up, I only get 1/10 of my order, and if the stock moves down, I get the entire order. I did learn that I'm making my time the order is in force too long (2 min) - losers move down in that time span, and will automatically fill my limit order.

So if I tighten the TIF (Time in Force) to 10 seconds, I probably won't get filled on as many losers.

But there's still a big problem here:
  • Set limit at bid - only fills on losers
  • Set limit at ask, or market order - Lose 9-19c per trade (1-4% on stocks from 6-9 dollars!)
So if there's a stock with a large spread, how do you trade it profitably?
Our Director of Trader Education, Jamie Hodge ( responds:
For as long as I can remember, either myself, or other traders I know have engaged in mean reversion off of the open.

Whether it be manual, or robotic, one rule has always held true, the larger the spread, the larger the reward, the greater the risk. There aren’t many trading rules that have stood the test of time in light of recent market activity, but this one has.

As a manual scalper of yester year, I would run from large spread stocks as well as thinly traded issues. Today, with the assistance of robotic execution, I welcome the big and the thin.

Here is a snapshot of one of my strategies that I launch on the open every morning:

For this strategy the minimum spread is set to .40 and the maximum spread is $2.

[Editor's Note: this strategy runs from the open until 388 minutes before the close; each position entered into lasts only for 15 minutes or a stop loss - whichever comes first; Understand these backtest results from The OddsMaker in more detail by reading the User’s Manual:]

Fill rates can average anywhere from 10 to 30% depending on opening conditions, mkt gap, etc.

So, looking at the daily summary, on May 11th, 12 signals were generated off of the open, so I more than likely was filled on 7 or 8 of the 35 possible trades.

From a statistical viewpoint, it doesn’t matter which orders get filled, they all have the same probability of working out.

The same strategy may also be flipped and run for shorts with favorable back test as well. Typically I’m bidding or offering a nickel inside, that’s where you want to get hit on a large spread stock.
I hope this helps.


Russell Miller said...

Awesome Post, Thanks

Anonymous said...

So where is the strategy? I only see the oddsmaker results?


D TradeIdeas said...

At the moment, he is not sharing it. But I can tell you that the strategy consists of only 1 alert - the % down for day alert. You'll have to experiment and define the filters that best suit your trading plan profile. I'll ask him if he wants to share this one.

Anonymous said...

You guys should give some back. Don't you get ideas from your users and integrate them into your own trading? I saw a strategy on your bot demonstration that looked very familiar to something I had discussed with someone on your support team.


D TradeIdeas said...

I'm surprised to hear this feedback.

We provide a lot of direction and inspiration on how to model strategies. If you can't find the numerous examples on this blog of ways in which we interpret ideas and patterns into usable strategies, then you should also visit our Support Forum.

The Support Forum consists of subscribers who in some cases describe their desired trading style and responses from Trade Ideas Support who attempt to offer strategies that come close to the original requests.

The devil is in the details of course and our subscribers details are their own. If something 'looked very familiar' I'd challenge you to be more specific. My point: You'll find the flow of ideas goes in the other direction here, at the Support Forum, and in dealings with our staff.

Thanks for the comment