Larry,
I tried first on the Q's. I used a buy signal when the price went over the 26 week EMA and the average was sloping up. I sold when it hit the SAR below the 26 week EMA while heading down. Reversed for shorting. No stops and took a full position starting with $10K. I used the sell signal I did with the following thinking: If the price hits the sar above the average the trend is still intact, if the price dips below the average it may be a mild correction, if the stock hits the sar below the average it means the stock dropped below the ema and couldn't rally sufficiently back up so the trend is likely broken.

Results: Started in May 99 when stockcharts had the data.
4 trades each profitable so no drawdown. The last trade is long and technically still going but I used a price of 37 to calculate results. Today I would have 41,380 from the 10K or 33%/year.

That looked a little too pat so I did it on the SPY using dat from Jan 93 to present. Same rules.
There were 13 trades, the last long and still going but I ended it with a price of 112.
7 trades were winners and 6 were losers. Today the 10K would be about $31,500.
There were 2 big losses so I thought a stop might help. I looked at the MAE for each trade and found that all the winning trades had MAE's less than 1 point except one that had an MAE of 6. All of the losers had MAEs over 2.
There were now 20 trades. The big win turned into 6 little losses and a big win and one of the small losses turned into 2 small losses. Of the 20 trades only 7 were profitable. Using today's price the 10K would have grown to $38,500, about 13%/year compounded.
I'd like your (or anyone's thoughts) about this system's value and what I should do next (try it on other markets, tweak it differently?)
Thanks,
Howard

Howard,

It looks like you're getting close. I'll take a look at it this weekend to see if I can suggest some improvements.

Larry

Howard,

One more thing while it's on my mind. I know the data since 1986 so well I don't even have to look at it. And I can tell you that it will be very easy to design a moving average system that did well in 1987, 1988, 1989, 1991, 1995, 1996, 1997, 1998, 1999, 2001, 2002 and 2003. It will be very difficult to do well in 1986, 1990, 1993, 1994, and 2000. The bear years of 2001 and 2002 can be greatly enhanced by simply taking profits when the S&P trades 20% below its 200 day moving average. That rule should be part of any system we design.

Larry

Larry,
Actually that trade did very well. It was the big winner that turned into several small losses. The problem was getting it started. I stopped out 5 times before getting in to stay. Once I was in (nov2000) I didn't get out until march 02. Got back short in april and did well until got out in nov. I guess that could have been better to get out a month earlier using your stop, now that I look at it while I'm writing this. It would have been another 12 points.
I'll try it again with that added.
h

Larry,
I redid the test with the stop at 20% below 200d mva. One big win of 3000 turned into 2 big wins totaling 7800. After the 11 year period the 10K was 40K instead of 31K. Boosting the average from 11% a year to 13%. After getting stopped out in July 2002 the market moves up for 5 weeks and then declines another 6 weeks for 10 points. I don't see how I could reenter the trade except after the first down week in aug. but that seems pretty iffy. What do you think?

These findings may be an understatement of your improvement because it gives me over 9000 more to buy more shares of the next big down move. If that were to happen today for instance I could afford about 80 more shares. It's just that your improvement enters into the picture so late (2002).
I'd still like to do better than 13%. Any ideas?
Thanks for your help,
Howard

Howard,

I've been up since about 4 am (central time) working on Point & Figure stuff. Man, it's more time consuming than I realized it was going to be. But it's going to be worth it. So I haven't looked closely at what you're doing yet, but I'm going to (also, the ideas of Gary and Joe).

But I do want to make a couple of comments about what you're doing. Since 2000 you should find three instances of the S&P trading 20% below its 200 day moving average (see Strategy 4 for exact dates) - two are in 2001 and one is in 2002. I think when you factor those in your rate of return should be significantly better. And because of the nature of bear markets in that they fall very fast and get very oversold, it just makes sense to take profits according to that rule.

Also, 13% is not bad. You're not going to find a moving average system that performs as well as the COT. But you don't need to because you're just looking for something to compliment the COT and to smooth out the volatility during the times the COT is going against you. At least that's my understanding of what you're trying to do. For example, if you were following your system now that part of your portfolio would be long the market which would greatly reduce the volatility of your overall portfolio.

Having said all of that I think there are improvements to be made with your system. And I'll get a chance to look at in the next couple of days.

Larry

Larry,
Thanks. The dates are the same as I had. Stopped out for profit in March and Sep 01 and again in July 02. In both the 01 cases there is no opportunity to reenter as the average flattens and priced go above the average before it heads down again. So there's a clear signal for a new trade. But in 02 the price after July doesn't reach the average which is still declining. The price almost immediately goes back above 20% below the average. I can't figure if there's a mechanical way to reenter the trade. I don't want to use my judgment (cause I'm not that experienced). Any suggestions.
Howard
PS Thanks for your work on p and f. Don't feel I'm expecting a quick answer.

Larry, Do you have data to test things back beyond 1987. I can't get spy on big charts but can get $spx. It looks like I would have been long with no warning for the crash. That's one of the things I want to take into consideration. As everyone's pointed out, the drawdowns will be easier with a trend following system in place along with COT, but if the trend following system can have as big a drawndown as 1987, I'm wondering how it would effect the results. Imagine if 1987 happened again next Tues. COT would do very well, but I'm not sure it's an equal trade off.
h

Howard,

You don't want to test with SPY anyway. You want to test with SPX. When backtesting with SPY or QQQ you can sometimes get bad ticks (trades that didn't happen). With anything before 1990, I use Big Charts. It goes back to 1970. If you're talking about 1987 in regards to your 26 week EMA test you would have come out smelling like a rose. SPX traded below the moving average about a week before the crash (it looks like at about 310). A week later the crash took it down to about 210. You would have been a hero. In both 1987 and 1929 there was plenty of warning before the crashes occured.

Larry

Larry,
I might have not done this right. I was using weekly charts and exiting trades only if the price hit the sar while it is below the 26 week average. The week before that it went below the average would not have triggered my exit. Should I add another exit a certain percentage below the average? The week before looks like it was only 8% below the average. Should I have been using a daily exit?
h

Larry,
I've spent the last several days testing a system based on the 26w EMA for $SPX. Would you please give me any critique or suggestions for further refinement?

The basic system: Buy at open after the first week that the weekly price range is completely above a rising 26w ema. Sell when price falls to average again. Reverse for short. 17 year results on 10K, fully invested, no leverage: 46 trades, 21 winners and 26 losers, ending capital was 35.5K.

Adding COT signals: I added COT signals by only buying if COT was long and exiting trades if COT went short, reverse for shorting. 17 year results on 10K, fully invested, no leverage: 47 trades, 31 winners, 16 losers, ending capital was $61.8K. A lot better but I noticed that it really didn't provide a smoothing retun on the COT system because it didn't follow trends the COT was against.

Removing COT signal and using 2% stop: I examined the MAE for the basic system and found that all but 2 winners had MAE's below 2% and half the losers had MAE's over 2%. Using a 2% of opening price stop the 17 year results on 10K fully invested, no leverage: 46 trades, 20 wins, 26 loses, ending capital was $73.4K.

As above and ending a short trade if price goes below 20% below 200d SMA. Results on 10K fully invested, no leverage: 46 trades, 19 wins, 27 losses, ending capital was $101.8K.

As above using position sizing risking only 1% of capital per trade starting with 100K. Results: trades as above,unleveraged ending capital $341K for a profit of 241K.

Larry, thanks for your time. Any suggestions or critique is welcome.
Howard

Howard,

I will take a look at it. It sounds like you're doing good work.

Larry

Howard,

When you say that you “sell when price falls to average again”, I assume you mean when it falls below the average.

You wrote, “As above and ending a short trade if price goes below 20% below 200d SMA. Results on 10K fully invested, no leverage: 46 trades, 19 wins, 27 losses, ending capital was $101.8K.”

That’s 14.63% per year. That’s not bad at all for an unleveraged moving average trend following system. What was your largest loss and your average loss in terms of dollars and percentages?

You have a method of taking profits when the market gets extremely oversold. Maybe you should have a method of taking profits when the market gets extremely overbought. You could use RSI, or Stochastics, or NYSE Bullish Percentage. There are a number of tools you could use. But you’ll need to experiment with it because the market can stay overbought for a long period of time. Or you may want to take partial profits when the market gets extremely overbought and let the rest ride in case it keeps going.

It looks pretty good. The main issue is whether you’re comfortable with it.

Larry

Larry,
Yes, I meant that if the price went below .5% below the 26w EMA, I ended the trade. I thought about the taking profits on overbought, I'm going to look at it further.
In any case the issue of yahoo, bigcharts, and stockcharts is vexing. Just to get on the same page I redid my test using stockcharts from 1990 on for $spx.
The trades were a little different.
I had 50 trades with a significant draw down from 1987 to 1992 of 21%. That was the biggest draw down resulting from 15 losers out of 16 trades (one the win was very small). It would have taken a lot of faith and patience to wait that one out.
It seemed that the system did great because it captured big trends and took small losses when not trending. I decreased the MAE to 0.5% and actually did much better (with more losing trades of course). Risking 2% of capital per trade with that MAE gave me at 17 year ending capital of $2.2M, but $1M was made this year on this up trend.
There are only 9 out of 50 winning trades so this is not something I'd like to leverage with options. But increasing to 2% risk did require margin averaging 120% of capital with each trade. Unleveraged using 1% risk the ending capital was $607K.
I'm going to redo the Nasdaq one with stockcharts and apply the overbought to that one first because of the greater volatility.
I'd also like to find a way to increase the reliability (maybe requiring 2 weeks above the average before I buy, or requiring a certain slope to the average). Can you think of a way I could cut down on the trades during the periods when the market isn't trending?
Thanks,
Howard

Howard,

The philosophy behind any tend following system is that there is always something that’s trending. So one way to do it is to allocate your trades among several stocks in different sectors that are not likely to necessarily correlate with one another. That way while some are going sideways, and whipsawing you, the others are trending.

Another way to do it is to use an indicator that is supposed to help you identify when the market is trending and when it is not. The two main indicators that Stock Charts has for that purpose are ADX and the Aroon Oscillator. For the purpose of identifying trends, I think Aroon is a little easier to use.

The more you learn to use the P&F charts the better you’re going to like them. But I don’t know how to tell you how to backtest them, so you will just have to track them over time.

Larry