My First Ever Trading Strategy

[josephsaad91]

Hey ASF!!

I'm a newbie to the market, have been reading up and researching for about a month now and have had a crack over the last two days at a trading strategy and back-testing it for the ASX share market as a mid term strategy (about 1-2 month holds)

If i ever implemented this strategy, it would be manual (not automated trading) as i have technical restricting with my current back-testing platform; i can't figure out how to have multiple stop/closing criteria, thus i have to manually oversee the trade for now.


My scope was broad, Any ASX stock that was:
- at least 100 mill cap
- price > $0.50

Trades capped @ $5000 to increase diversity (i hope it does lol)

Stop loss @ -18% (this is currently my catch bucket for that technical restriction i have)

Date range: 1/10/2013 - 08/08/2014
(i know the data is a bit less than a year, but it's all i could get with my platform)
(any trade with "period end" ends up forcing my trade to close)

Personally, i don't know what a good return is for mid term hold trades, but considering this year has been pretty stale for ASX, my hunch is i'm on the right track?

What you guys/girls think?





Cheers,
Joey

 

[VSntchr]

Well the obvious fall down is that a 1-2 month holding period backtested over a period of 1 year, one time, is not a sufficient sample to draw any conclusions from.
You need to backtest over multiple periods, to cover many different market conditions, sentiments etc...not to say that you need to find a system that works at all times, but rather finding when the system works and when it doesn't.

If systems trading is the direction you want to head in, then backtesting is the first place to start - so your on the right track.

As for "mid-term hold" trades expected return, I would think that a return of 19%pa if achievable consistently is very good.

 

[Wysiwyg]

Backtest gotchas :-

Didn't delay trades until next day
Didn't add brokerage or allow for slippage (186 x 40 (buy $20/sell $20) = $7440


18% of 5000 = 900 per losing trade
52 losing trade x 900 = $46800 loss conservative

Result is that large outliers skewed results to positive return, there were less than 6 losses sustained consecutively which with a predominantly range bound market and 33 positions open is not real.

One has to be true to oneself when back testing and be as accurate to real trading as program will allow. Otherwise it is extremely misleading and will result in losses in reality. No discounting the feel good factor when back testing though.

 

[josephsaad91]

Just not sure exactly what you mean on a few things:

- "Didn't delay trades until next day"
- " there were less than 6 losses sustained consecutively which with a predominantly range bound market and 33 positions open is not real."

Just to be clear, i haven't actually disclosed my strategy just gave some context to the result. good point on brokerage, id say cmc/bell direct numbers @ $9.90- $15. the -18% losses would never have gone through if i could overcome my tech restriction, will post up revised results as i can get them aha

 

[AverageJoe]

Backtest gotchas :-

Didn't delay trades until next day
Didn't add brokerage or allow for slippage (186 x 40 (buy $20/sell $20) = $7440


18% of 5000 = 900 per losing trade
52 losing trade x 900 = $46800 loss conservative

Result is that large outliers skewed results to positive return, there were less than 6 losses sustained consecutively which with a predominantly range bound market and 33 positions open is not real.

One has to be true to oneself when back testing and be as accurate to real trading as program will allow. Otherwise it is extremely misleading and will result in losses in reality. No discounting the feel good factor when back testing though.

My personal experience with system trend trading long only and I did it successfully pre-GFC is because of that Gigantic bull trend! I had backtested with software on the dot com boom bust and it performed in a robust fashion.

It is too easy to be aggressive with no. of position opened at any time and I had 10 with the max 10-1 leveraged in the backtest when developing the system and the backtested results were fantastic! I even spend a big part of the effort tweaking the parameters for optimum performance! It worked fantastic in real life for a year until 2008 GFC destruction and then I suddenly realised I did not designed a system my trading psychology could handle.

The lesson I learned is be careful aiming for profits at the expense of huge draw downs and the number of consecutive losses before the mind starts to question the system. Along the way I have also discovered that data mining which is what backtested system trading effectively is can only give you so much. You have to believe that past replicate the future.

Sadly I have gone back to discretionary price action trading with S/R lines and candlesticks. I discovered along the way system trading is not for me and much more successful with discretionary trading.

Good luck in your journey.

 

[Wysiwyg]

Just not sure exactly what you mean on a few things:

- "Didn't delay trades until next day"
- " there were less than 6 losses sustained consecutively which with a predominantly range bound market and 33 positions open is not real."

- "Didn't delay trades until next day"

You get the signal to buy on daily close prices and buy the next day at open price. If you don't delay buy then buy price will be the open price on the same day as the signal and that is not possible in reality.

- " there were less than 6 losses sustained consecutively which with a predominantly range bound market and 33 positions open is not real."

3% drawdown of $4728 is less than 6 x $900 ($5400) losses in a row. From February to July the Index stayed in about 200 point range. From the start of the backtest there were two considerable dips in the market.

Change the test period = change the result.

 

[Wysiwyg]

Just not sure exactly what you mean on a few things:

- "Didn't delay trades until next day"

To be more helpful, post a stock that you bought on test. The code, the day of buy trigger, the day of sell trigger and I will show you.

 

[Wysiwyg]

This is a highly successful strategy if employed at the start of a bull market and switched off at the end of a bull market. The test period on the All Ords is same as your period. The result was due to one stock outlier, LNG, which profited $21000, Take that anomaly out of the equation and the system is at a loss. 10% stop loss and consecutive losses of 37 is unrealistic. Fail!

 

[tech/a]

Joey

Firstly well done for taking your first steps into system design.
I personally think its one of the best things a new trader can do to jet propel his learning process.
Spend enough time doing it and you'll soon workout what works and what doesn't---and most importantly WHY


But the road is long and fraught with many traps.

Wysi has pointed out a few.

The taking a trade in the exact day the entry is seen in a data set is a big mistake and I think every budding systems designer makes it.
Not including brokerage another.
Insufficient data another.
Forward referencing conditions another
Over optimization another
Poor testing---not using out of sample data---not forward testing--not understanding that a method will not follow its blueprint if market conditions alter significantly relative to the data that was tested originally.

AverageJoe

Points out the obvious and every time I see it it shake my head. long only methods are by nature designed to perform best in bull markets. If the way over perform the index then they are a good system.
If they sit out bearish periods or in some way cut back to maximize return on risk then they are also doing their job.
To expect a long only system to perform brilliantly in any market is like expecting a car to get to a destination as quickly and efficiently through Mud/Sand/Water with no fuel as it would travelling on bitumen with plenty of fuel.
If you want to design a system that works well in all market conditions---pick an instrument that you can short.
My choice is index futures.

Finally.

If your getting started I can guarantee you that your trading ideas (Which you are protecting with non disclosure)
will have been tested by the 1000s out there doing the same thing--LEARNING

Unless you can program loops in your testing---can overcome forward referencing, have the language capability of an experienced programmer---then your systems are likely to be similar to 100s of others. There IS NO NEED to protect you proprietary conditions and variables.

I have been down your road.

I have traded one of mine live (TechTrader) on the Net for 7 yrs and it has been dissected by more people than I could possibly know. It traded in a raging Bull market and turned $30K on margin to $365,000 when I stopped it in 2007.
After reading Howard Bandy's work ---who is a Doctor of Mathematics---it became clear to me that this Builder (me) was never going to be as good as these guys when it comes to systems design.
Now I pay people to design and critique a whole host of ideas and methods. These guys are amazing and the results can be truly eye opening to say the least. Worth every cent. (I'm currently investing around $23000 a year on this).---and yes at sometime there will be an attempt to recoup my investment not only through the systems that are developed and designed but from those eye openers as well!!

I suggest reading Radges book "Unholy grail's" which goes a good way with the basics--TechTrader one of mine is in there.
Then Howard Bandy's works.

 

[tradezy]

Hi Tech/A,

Great information!!

I haven't heard of that term "Forward Referencing Conditions". Can you explain what this means?

Thanks,

Tradezy

 

[tech/a]

Hi Tech/A,

Great information!!

I haven't heard of that term "Forward Referencing Conditions". Can you explain what this means?

Thanks,

Tradezy

Many budding system developers when they find the Peak and Trough functions think they have found the holy grail of conditions. Anything with a peak or trough can be found and incorporated in a formula.
Such as a % change from a pivot high or low or Divergence which requires a higher high or lower low.

These reference the ZIGZAG function.
The Metastock primer explains it better than I


http://www.metastockproducts.com/page_info.php?link_name=ch4_zig_zag

When you use a data set (historical data) the peaks and troughs are KNOWN.
So results are perfect.
In realtime the last leg to identify the latest peak or trough is dynamic it can't be known until the next leg is dynamic.

 

[darkhorse70]

23k a year John, thats insane!

When I read books such as market wizards I have just now realized that many of these successful traders developed systems. Some or most assisted by programs.

Were most of them assisted by computer programmers?
What about newb traders such as myself and others who dont have those sophisticated computer skills?

Ate we still capable if designing great unique systems?

 

[tech/a]

23k a year John, thats insane!

When I read books such as market wizards I have just now realized that many of these successful traders developed systems. Some or most assisted by programs.

Were most of them assisted by computer programmers?
What about newb traders such as myself and others who dont have those sophisticated computer skills?

Ate we still capable if designing great unique systems?

Yes of course. (don't know about unique but certainly profitable.
Many years ago I designed the one mentioned above. Did very well.

We (Kris and I) are working on the very issue you and many traders face.
Will keep you and interested people in the loop.

 

[darkhorse70]

Sounds good John.