Wednesday, March 28, 2007

Dead Cat Bounce - The Filter

The "dead cat bounce" is price action after a panic drop over one or more periods - generally price jumps and then resumes the decline soon after.

One of my readers - IIO - pointed me to another site where a guy is using a filter he calls the Rocket and on first glance it appeared to be similar to my BOB. But it isn't.

The difference between the "Rocket" and BOB is that the Rocket was founded on a price basis including percentage decline and BOB was developed on a volume basis. That's why I'm always insistent that volume on the second day of the triad be higher than on the first, and for a set-up - significantly higher. The BOB was also designed for a sustained rally because it was originally conceived to support swing trading. The Rocket was designed to catch the proverbial dead cat bounce and as such works extremely well.

There is a great difference between price based filter and a volume based filter so while the two filters may appear to be similar they are as far apart as night and day. I discovered BOB when I was doing an extensive analysis of the effect of volume on the next and subsequent day price changes.

The guy on that Blog keeps his code secret so I derived a filter from the several charts he had on the site containing stocks he attributed to the filter he named "the Rocket". I call this one - what else - Dead Cat Bounce -

show stocks where close is between 1 and 35
and average volume(90) > 150000
and close 1 day ago is more than 8% < close 2 days ago
and close is more than 5% < close 1 days ago

For the current test period (11/06 - 03/06) the back test saw 67 stocks (based on taking the highest volume stock when two or more were selected on a single day) and produced a win rate of 41% with a 263% ROI and a 4-day net of 4.23 and a 30-day net of -1.71%.

Except for the win rate that's impressive. Then for the known good period 10/06 - 02/06 the back test saw 66 stocks and produced a win rate of 42% with a 509% ROI and a 4-day net of 6.33 and a 30-day net of -.01%.

This period was also a little lacking in the win rate department but if anyone were to use this filter and actually play the stocks that are output I suggest you set the stop close - the stock is either going to go up immediately or decline immediately and in the later case you want out quickly. Also this ROI is based on the fact that the highest volume stock each day of all stocks output was used for the test. I suggest that you adopt that approach also.

The code is clear - it is looking for deep declines in the stock market and it wants them over two days. I'm sure there are ways to modify this to raise the win percentage some but I have to say - for that kind of ROI I'll take a few stop outs.

I'm going to play with this for awhile and if I can find a lower risk solution I'll let you know.

Also stop by Stock Monster - I read some yesterday and even if the guy doesn't share his code he does share his picks, writes well, and writes often.


slotmonkey said...

Here's the actual code which he did reveal later

show stocks where count(RSI(2) 2 day ago above 90,60) is above 0
and RSI(2) is below 20
and average volume(10) above 250000
and close one day ago more than 8% below close 2 day ago
and close more than 4% below close 1 day ago
and close above 0.99
and 60 day slope of close is above 0
and market is not otcbb

The "reverse rocket" has even more impressive performance. The code is the reverse of the rocket's. If you need it let me know.

Could you please explain the first line in the code? I'm not that good understanding advance code like that.

The problem with the rocket filter is that it can go for days without a selection. That's why I've started trading the reverse also and have made some nice coin last couple of weeks. Am also considering play the otcbb rockets just so I could get more action.

I don't believe you've ever posted the code for your breakout filter. Do you mind sharing that one with me?

I read you everday. Keep up the good work.

Marlyn Trades said...

Thanks for the code - I'm pretty happy with my "reverse engineering" I hit the first decline perfectly and was only 1% off the second. Not bad for an old guy

"show stocks where count(RSI(2) 2 day ago above 90,60) is above 0"

He wants to make sure that the RSI(2) was greater than 90 2 days ago and he wants to count how many times that occurred in the past 60 days. Since he's only asking that the count be about 0 I'm not sure why. I'll try it and also try "and RSI(2) > 90 2 days ago" and see if that gets the same results - I'll let you know.

slotmonkey said...

My hat's off to you, I'm amazed at how close your rev. eng. came to his parmeters. Actually, you can loosen it up from 10%, 5% to 8%, 4% to squeeze out more plays without hurting the bottom line too much.

I want to monkey around with the other parmameters to see if it could generate more plays without a big sacrifice on the bottom line.

Asking the count be above zero is to look for stocks that have been in an uptrend and hit some sort of a peak at least once in the last 60 days, is my guess.

Could you please also verify that holding for one day is better than two.

Once you backtest his filter, the resulting win percentage is going to put a big smile on your face, guaranteed.


That old dead cat bounce.