Page 1 of 1

Why don't these guys trade stocks?

Posted: Sat Mar 06, 2004 2:01 am
by trendguy
While researching some of the famous mechanical trend followers of the last 30 years I quickly noticed that none of these guys trade in stocks ... namely Richard Dennis, John Henry, and Bill Dunn.

I asked myself ... why is this?

It would make sense for a trend follower to trade any product that trends and stocks definitely trend, some of the time.

The only answer I can come up with is that it's not possible to apply a purely mechanical approach/system to the stock market. I ran some tests on some well known trend following systems, moving average crossovers and breakout systems. I found that it is very possible to have hundreds of signals generated on a daily basis using these simple trend following approaches and it would be next to impossible to trade all of these signals, some of the time.

Some famous equity trend followers, such as William O'Neil, use fundamental data filters to reduce the list of potential candidates but this seems to go against the mechanical trend following approach that Dennis, Henry, and Dunn apply. These guys only use price as an input.

I would appreciate any insight into this issue.

Posted: Sat Mar 06, 2004 9:14 am
by lperepol
trendguy,

Try using larger time frames for moving averages and breakouts. Trade small portions -- diversify. Separate longs from shorts. Long side usually does better. When you get multiple signals choose siganls at random and create a baseline. Then implement a chooser later. Trends are hard to predict. Try a 52 week (250 day) breakout to start with and loose stops and loose exits.

I too am testing systems using stocks. Results are better than buy and hold strategies on major indices.

Read over Ed's risk paper and FAQ http://www.seykota.com/tribe/

Lawrence

Posted: Sat Mar 06, 2004 10:40 am
by trendguy
When you get multiple signals choose siganls at random and create a baseline. Then implement a chooser later.
This is not a completely mechanical trend following approach. I am talking about something that can be backtested. If you choose your signals at random, how can you backtest this? How do you know your results will be the same in the future? How do you know that buying 52 week highs with loose stops does indeed work?

Posted: Sat Mar 06, 2004 11:41 am
by edward kim
if you believe in the concept of persistence, then long term trend followers like Dunn will be more likely to trade futures than stocks. long term trend following is based on many small losses and and fewer big wins, so they are dependent on heavily trending markets. futures are more likely to act this way than stocks.

Posted: Sat Mar 06, 2004 11:42 am
by lperepol
This is not a completely mechanical trend following approach. I am talking about something that can be backtested. If you choose your signals at random, how can you backtest this? How do you know your results will be the same in the future? How do you know that buying 52 week highs with loose stops does indeed work?
>> How do you know your results will be the same in the future? I don't.
>> How do you know that buying 52 week highs with loose stops does indeed work? It works in back testing --> so it must work now.

Not mechanical? How would you know that the chooser you select is not better than choosing stocks at random when multiple signals exist?

Suppose you are using a MA crossover system and 500 equities are at the crossover point, signaling a buy. Your account can handle only 50 positions. How do you choose the 50 to enter into your account?

I say select the 50 at random, and create a baseline, something to compare against. If the results of the chooser are better than the random chooser than you have a good chooser.

I have not found a mechanical chooser that does better than picking at random. The system I use generates far more buy signals than my account can handle and I am too lazy to manage portfolio sizes over 300 positions. Another approach is to lower your position size and buy them all. That also seems to work well -- back test it!


Lawrence

Posted: Sat Mar 06, 2004 2:25 pm
by Ted Annemann
  • 1. Why don't big name hedge fund managers trade stocks?
To answer this question is to presume to be a mindreader. I cannot read the minds of John Henry et al, so all I can do is guess. Their reason may be: they did a lot of computer analysis and decided they preferred futures instead of stocks.
  • 2. Is it possible to apply a purely mechanical trend following approach to the stock market?
Obviously the answer is yes. Whether the results are profitable, and sufficiently profitable to be interesting, is harder to know :roll:

Posted: Sat Mar 06, 2004 8:09 pm
by trendguy
Ahhh. You are looking for certainty. The backtest would involve a statistic comparing many distributions of returns.
Not certainty, just hard and fast rules.

Posted: Mon Mar 08, 2004 8:48 pm
by BigBrad
As others I cannot answer as to why futures and not stocks. Based on my back-testing I definately believe that a mechanical system can be created for the stock market that has above average returns and ratios and what all.

A different issue with Stocks is the margin requirement. With stocks you can only basically buy 2X what you have. So a 50K account could only buy 100K of stock and then be fully loaded. If position size was set at say 2% or 1000 risk and you used a volitility sizing parameter that was $2 per share, then you'd buy 500 shares of stock. If that stock was a $60 stock then you'd be $30K for that one position. At best you'd be able to get 6 total positions. At that point, to get better diversification you'd want to have fewer stocks in the portfolio or take a smaller size. But when you go to smaller positions slippage and commision become greater and eat into profitability.

Fixed fraction risk and stocks

Posted: Mon Mar 08, 2004 10:07 pm
by ksberg
A different issue with Stocks is the margin requirement. With stocks you can only basically buy 2X what you have. So a 50K account could only buy 100K of stock and then be fully loaded. If position size was set at say 2% or 1000 risk and you used a volitility sizing parameter that was $2 per share, then you'd buy 500 shares of stock. If that stock was a $60 stock then you'd be $30K for that one position. At best you'd be able to get 6 total positions. At that point, to get better diversification you'd want to have fewer stocks in the portfolio or take a smaller size. But when you go to smaller positions slippage and commision become greater and eat into profitability.
Interestingly enough, if you traded with 8% to 10% stop loss per the CANSLIM method, the sizing parameter is $4.80 to $6.00 per share, and you'd buy 417 to 333 shares of stock. This scheme allows you to diversify a tad more. At 25% stop loss (a feasible addition for certainlong-term stock systems), you could conceivably diversify even more, buying only 133 shares of that same $60 stock in that given account.

I don't think sizing stands in the way of a mechanical stock system.

Cheers,

Kevin

Posted: Tue Mar 09, 2004 8:01 am
by Mark_et_Lizard
Hi thread,

I traded stocks all through the 90s, I got about 80% of the way to being fully automated. I think if Veritrader eventually gets automated execution it could easily be done with it. I cut this from a chat I sent someone a while back so forgive the bad grammer and typos

Here are some of the things I used that worked for me:
Buy only, no shorting
buy the second time through of the all time highest high for the entire history of the stock (this alone will keep you out of almost everything that isn't moving. I just don't mean a 200 day high or 1000 day high but the all time high. wait for the second time, if it follows through the very next day or a week later. if it doesn't go through for a few weeks then wait for the second break above the new all time high)
trail a tight stop, you'll cycle through a lot of stocks but you want to look for something that going to trend really well.
pyramid, once you get a stock trending buy every new break out, wait for a few days retracement and put buy stops above the high(no need now to wait for two times through )
keep % risk small, if you can use a big account you can risk 1/2 % or even a 1/10 of a percent on each entry (you want to be able to cast a wide net, buy everything thats moving up, then start to hone in on the real winners

lastly if you can code all that up and automate figure out how to make veritrader do it please let me know I'd be very interested and might even start trading stocks again.

Posted: Tue Mar 09, 2004 12:48 pm
by trendguy
Here are some of the things I used that worked for me:
That type of system can work well if you add a market trend filter, test it and you will see.

Posted: Wed Jun 02, 2004 11:31 pm
by si
Mark_et_Lizard wrote: keep % risk small, if you can use a big account you can risk 1/2 % or even a 1/10 of a percent on each entry (you want to be able to cast a wide net, buy everything thats moving up, then start to hone in on the real winners
I think this is key, given the high margin setup and too many choices. Otherwise, you are forced to filter some of the signals out, and may miss the real winners or not be able to meaningfully pyramid early enough and either of those essentially kills a turtle-style system. It may be possible to construct other systems though.

--s.