Feed back on Trading Startegy

vocuer

I have been working for a few weeks on a new trading strategy.
The strategy is based on monthly closing of QQQ.
It is based mostly on a bok I read, their strategy uses monthly closes of SPY's and ATR.
I added some signals: crossing of lower BB 12,3 , combination of WRSI, Full Stochastics and MACD, and $NAA200R crossing from below.

Entries=All entries were based on monthly closing price
1-$NAA200R crossing the 20 line from below
2-QQQ closing above mid BB.
3- 3 Oscillators signaling by crossing their oversold band from below

Exits & shorting=All exits were based on monthly closing price, ending short trades was a times based on stop losses being hit, subsequent reversal into a long trade was based on monthly closing price.
IGNORE ALL EXIT SIGNALS IF ATR FLAT OR DOWN, EXIT ONLY IF ATR IS UP
1-QQQ closing below mid BB
2-3 oscillators crossing oversold line from above, shorts always with stop loss(not waiting until months end.


I backtested the strategy from 2002 to 12/201startin with an original 100$ investment, I did it multiple times.
The math was very tricky. I got at different results, but they were all good, ranging from the 100$ becoming 500$ to 1100$.
The best results were with the strategy described above (prior strategies did not have stop losses for short trades,
I backtested in real time (I started at the end of 2002 and added 1 week at a time so I would not know what came ahead.

The results were:
100$ in 11/2002 became 1145$ in 12/2018(trading long and short.
There were 10 total trades: 7 winning and 3 losers.All 5 long trades were winners, of the 5 short trades 2 were winners and 3 were losers.
Investin on QQQ's in 11/2002 and holing until 12/2018 100$ became 530$
Long trades only 100$ became 889$
Best long trade 87.59% 11/2002-12/2007
Short trades only 100$ became 129$, most short trades were loses less than 15%, but some of the successful ones were massive ( 2007-008).
Best short trade 37.97% from 12/2007-04/2009

As you can imagne this being a trend following strategy gave great results in bulls & bears, but not great in sideways markets.

I have never developed a technical trading strategy before, I would greatly appreciate feedback.

Thanks in advance

vocuer

The settings for the oscillators were determined before the study, they were all set to their standard duration (period of 14 for WRSI % full stochastics, and 12 for the MACD.

It was all done through visual inspection, advancing the chart 1 month at a time.
Practically all of the indicators were examined as stand-alone before using them together.

The original study was done with the entries recommended in the book (crossovers of EMA) which I changed to Bollinger Bands 12,3 and exiting only when the signal occurred after 2 months of increased volatility (ATR) an 2 months of outperformance by utilities/spy;
This produced very good results and it lead me to include a new entry (the downward cross of the lower Bollinger Band. This signal occurred n 5 times betwen 2002-2018, 4/5 were good

The first study with indicators was done with Nasdaq % of stock above 200 EMA, it gave great entry signals (frequently earlier and better than the mid BB crossover) but no reliable exits.Over all the results were good, the added signal generated bigger gains per trade.

Then I tried the 3 oscillators which gave good entry and exit signals, again exits only when ATR is upwards. I found this visually easier than xlu/spy, and it gave the same results. The signals required the simultaneous crossing by all 3 oscillators. The oscillators were tested going forwards month by month. During all these studies I was blind to what was coming
.
I could not get any data before 2002 (first year in which $NAA200r was available.

I ended up with something I was not looking for. This started as a search for a system with more frequent trades and less time in the market, but similar work with weeklies did not pan out good results. I was desiring to find a system tha kept me out of the markets for long periods of time (avoid an October 1987)

There is no guarantee with any system that what worked before will work in the future. But yes, I do plan to trade with this system. However, money management will be very different. In the study, I start with 100$ that become 1145$ 19 years. All the original money remains invested all the time.
In real life I aim at getting 7-10% per year; so I will most likely start with 1/10 of the money, and if the $$ pool increases the then subsequent trades will change accordingly. While I was doing the testing I was avoiding looking at the price action, so that it would not tempt me to start tweaking the signals Of course on real life this would be much more difficult. I also imagine that the results would vary greatly depending on when you started trading. There were long periods when the $ was sitting in the market without doing much. The problem is that I was able to identify a trendless market retrospectively and never retrospectively. I had the intention of identifying these periods so to trade from mid BB to top or bottomBB
.
Drawdowns were not calculated, but at simple inspection they are lower than buying and holding the qqq's.
The losing trades were
2008 -13.61%
05/2010-10/2010 -0.667%
08/2010-09/2010 -9.32%

The stop losses improved the performance on the short side. even though the losses were not large they were more frequent than the wins

Thank you very much for your comments.
Your input is greatly appreciated.

Are you aware of any other forums where I can post this work?

markd

You could try the magazine Technical Analysis of Stocks and Commodities.

markd

Very interesting.

I do have questions and a comment.

Did you determine the signals - like the indicator crossovers - occurred by visual inspection of the chart, or did you run a scan of the criteria for each month end closing price?

If you used visual inspection, were the crossovers all co-incident in the same month, or had some crossovers occurred in recent prior months?

When you say, exit only when ATR is up, does "up" mean that ATR ticked up for that month, or ATR subjectively looks like it is trending up?

What I'm getting at is, could there have been some unconscious subjectivity in selecting entries and exits? If you ran scans to determine signals, that should be reliable, but visual inspection would raise the question of interpretation in real time - when money is on the line, would you make the same interpretation?

Another question comes to mind about curve-fitting. Were the indicator choices and their parameters chosen after looking at ALL the data? If so, it may be that the choices are valid only for the period studied, and may not work in the future. One way to test that, if you have access to the data, would be to run the system again against older data, say 1970-1990, or 1950-1970.

Also a comment about holding periods. It seems trades are very infrequent. As a practical matter, if you have important money on this system, would you have the confidence to hold through periods of under-performance or drawdowns? (I don't see that you mentioned drawdowns in your statistics - i.e., length of time you would have been holding a loss, and what per cent.)

Great work. Thanks for sharing!