Need help debugging system - Aussie Stock Forums

# Thread: Need help debugging system

1. ## Need help debugging system

Hi all,

I have been playing around developing a system of my own, as I feel this is the best way to truly understand what is going on.
As I am a software developer, and have access to asx data, I decided to write my own program that would search the stocks and run them against certain criteria.

My trading system to date is -

Run program at 3:30pm each afternoom
Stocks are examined for the following criteria:
- sp gained more than 5% throughout the current day
- my purchase (\$3000), is less than 1% of total volume for the day
- stock has had positive growth over the last 3 days, of which each day is less than today's gains
- each of the last 3 days closed within 2 increment of the day's high
- total volume today is greater than 5 times the average of the last 3 days
- sp gain today is less than 20%
- last sale price was within 1 increment of the day's high

The program takes approx 5 minutes to run.
On completion I then have a list, usually consisting of 5-12 stocks. This list is ordered in terms of volume, then market turnover. I manually watch these stocks for the next 10 minutes, as the market draws to a close, concentrating on the top 3.

If any of the top 3 enjoy any further gains, I take this as buying pressure, and make my purchase accordingly. Usually at about 3:55pm.

On open the next morning, I sell the stock once it triggers an exit alarm.
These are:
- stock has remained within 1 increment after 1 hour has passed, and volume is low compaired to previous day
- stock has opened lower by less than or equal to 2 increments
- stock is climbing, but has started to retrace
- end of the day has come

I have been paper trading this for 2 months, and have the following outcomes:
Win ratio - 70%
Average Win Amount - 4.5% (incl. brokerage)
Average Loss Amount - 6% (incl. brockerage)

Based on the above outcomes, the average profit per trade is 1.35%.

From simple mathematical analysis, if I increase the win ratio, increase the average win amount or decrease the average loss amount, then the system would return a much higher profit per trade.

I have in the last 2 weeks gained confidence in my system, and decided to get my feet wet, only to find that I have had 1 win out of 10 trades. This equates to an average of a loss of -4.95% per trade. That hurts....

One thing I have noticed myself is that most stocks returned by my program are in the range of \$0.04 - \$1.00. This results in fast movements through increments. When it works its great (some paper trades 30% in one day), when it fails its bad (have lost 8% by going to get a glass of water...).

Can anyone please provide some constructive points about how I can improve this system, or where the nearest bin is so I can start again...

thanks

2. ## Re: Need help debugging system

Hi stock_man

Quite a neat system you have there - one thing I would remove from your output list is the very small penny stocks as these are the most volatile I guess? Just feedback thats all.

3. ## Re: Need help debugging system

Best advice you're likely to get is not to drink water...try bourbon...lol

Interesting system though....have you got one for lotto and horse racing too?.....you may get the same results.....all systems are fallable.

4. ## Re: Need help debugging system

Stockman..
Sounds like a sound start. Did you have all the rules finalised before you started paper trading, or were some of them optomised during the trading trial?
What charting package do you use, some have backtesting capabilities so you can test the method over a larger time sample and get a more accurate picture of results. Dare I say that if you have Metastock (Bullcharts??) you could ask Tech/A about the tradesim backtesting add-on.
Cheers..
Kauri

5. ## Re: Need help debugging system

Originally Posted by Kauri
Stockman..
Sounds like a sound start. Did you have all the rules finalised before you started paper trading, or were some of them optomised during the trading trial?
What charting package do you use, some have backtesting capabilities so you can test the method over a larger time sample and get a more accurate picture of results. Dare I say that if you have Metastock (Bullcharts??) you could ask Tech/A about the tradesim backtesting add-on.
Cheers..
Kauri
I started of with just a few simple ideas and rules, and over the last 2 months have come up with this system. I have to say that it is completely different from where I began!

But this is where I am getting stuck - trying to improve my results from here. I feel that for the return I am getting, the risk is too high - especially from my real testing! (10% win rate...).
I don't actually have a charting package, and only use comsec protrader for software (apart from my own ), so don't have access to back test.

6. ## Re: Need help debugging system

I think you exit a little early and a little too easily, sure use stop losses and if the trend is against you get out quick, but let your winning trades run for a while to try and get some big profits going, always selling the next day will reduce your chance to get any large gains. You'll also reduce your brokerage by buying and selling a bit less.

7. ## Re: Need help debugging system

I agree that letting my trades run a bit longer would be beneficial to times when the trend is strong. This has been proven to me when I have started to see the same stocks occur 2 or 3 times in the last 8 weeks.

One of my aims of this idea was to exploit the T+3 settlement days, and not tie down my capital by having it locked away in a stock. ie: If I buy and sell the next day, I never actually loose the capital, but rather only see the profit/loss in my account.

8. ## Re: Need help debugging system

Firstly your sample size is far to small to have confidence in a system

I personally wont trade anything (Other than discretionary trading) unless I have 20000 10 stock portfolio's in the systems test and 100% success rate.
Many other factors come into play.

Your method is very short term and as such is very likely to be whipsawed badly as you have found.

I would add a trend filter initially say trading above a 40 day EMA.
Id also look for a longer timeframe,then you will need to test a much much larger sample.
Profit is in the EXIT so look at exits for increase in profitability.
However trading stock which has not entered a positive trend on the long side is suicide.

9. ## Re: Need help debugging system

Originally Posted by pacer
Best advice you're likely to get is not to drink water...try bourbon...lol

Interesting system though....have you got one for lotto and horse racing too?.....you may get the same results.....all systems are fallable.
You're a funny guy Paceman .............. might go and get myself a bourbon right now !!

Stockman, The principles sound good. If you mix Tech's and Senecas advice you may well be on to something sound ......... I agree with tech re the computer savvy ...... wish I had that as well ... Cheers Barney.

10. ## Re: Need help debugging system

Thanks for the feedback guys.

I guess the struggle I am finding is whether this is becoming a system (eventually...), or just an educated gamble?

The only reason GDN didn't appear on my list yesterday (40%+ gain today), was the volume didn't quite fit my formula. Something to ponder on...

Am I right in my idea that as stock is pushed towards its high on close, on a positive day, that there is good buying pressure, which should help it to open higher the next morning?
The only thing I have to combat here is when profit takers start to sell, and the stock stagnates or falls, based on the fact the stock has had good prior gains.

I feel I am starting to see the light of holding stock for more than overnight.

11. ## Re: Need help debugging system

I thought I might let everyone in on my motivation:

I have a bet going with my brother about who can double \$4,000 the quickest, by whatever means necessary.

My aim is to double this by April next year (6 months). Its a big ask, but its better to aim for the stars and land on the moon, than never take off at all.

Just for the curious, he is doing it through starting a business selling computer hardware.

12. ## Re: Need help debugging system

Originally Posted by stock_man
Can anyone please provide some constructive points about how I can improve this system, or where the nearest bin is so I can start again...
You are describing a price/volume breakout system with what sounds like pretty much a perfect entry strategy.

You are then describing your exit as wanting the stock the next morning to pretty much immediately keep climbing, and as soon as it doesn't out you go. In essence, you're trying to select for those few outliers which go ballistic (eg GDN today). Tech/A is demonstrating this style of trading very effectively in another thread.

An observation I'd make - many perfectly profitable breakouts retrace a bit before going ballistic again. You may wish to try giving the trades a tiny bit more room to work out before shutting them down.

The one warning sign that all is not good with your system is that the average loss is higher than the average win, so you're relying on a very high win rate to make a profit.

13. ## Re: Need help debugging system

Originally Posted by MichaelD
The one warning sign that all is not good with your system is that the average loss is higher than the average win, so you're relying on a very high win rate to make a profit.
Thanks Michael for you comments. I agree with you about the loss being higher than the win. This is due to the fact that I allow the loss to go a little bit under before pulling the plug, as most stocks seem to be volatile throughout the day. Then add on your brokerage, and the loss goes further. Eradicating this would significantly change the end results.

14. ## Re: Need help debugging system

Stockman

With \$4k you need to trade just one stock.
You need to cut your losses very quickly,when trading this way I use 2 or 3 price points.IE if its a 10c stock then 1.5c.
You need a reward to risk average of 3:1 or greater.
You need a win rate of over 60% so 6 wins to 4 losses.

You need to find stocks that are on the move EARLY in the day, and hop on.
If they close toward their high provided they havent risen more than 30% in one day or 50% in 2 days then Id hold overnight.
Make either the High of the day you bought OR the Close of the day you bought your trailing stop for the next day.
If it gaps on open then sell if it comes back to the high.
If it opens at the high or opens at the close and trades back to the close of the previous day then sell---quick.

If you wish to survive trading in this kemekazi mad style of trading then thats the best you'll be able to do.You could double your money quickly but if you fail to take your sells youll lose the lot in no time.

Its really high risk and not really for novices.
You really have to be highly disciplined and understand Price action well above novice standard to make a dollar.
You have to watch the screen constantly.
I can make a good buck from it but I hate it.
Ive pretty much had enough of this last 2 mths trading very short term and will place the funds shortly in System 4 a new longer term portfolio method which again I'll trade on Margin.

Stevo
You maybe interested to know its a weekly one.

Anyway Stockman hope that helps.

15. ## Re: Need help debugging system

There was a post several years ago on the IC forum (would still be in achives) about closing prices and 3 day settlement.

From memory it went along the lines that stocks that have 3, 5 or 8 consecutive days of a higher close tend to decline the following day.

Don't know if it's worth checking out, as one of your rules seems to include day 3.

Cheers

16. ## Re: Need help debugging system

Thanks Tech/A for your awesome insight. When I first started this system, I was actually working on something similar, ie: catch stocks as they start to build momentum early in the morning. But I was having difficulty in determining if this was actually a solid price gain, or just small peaks. The analysis I was using was past performance (ie: previous gains in the last few days), coupled with large volume and quick price gains within the first 30-60 minutes.

As this is a very white knuckled sort of ride, I found my idea above, which was to identify these stocks EOD, and just skimp a little from the top before they peak. Profit wise, all I need is 2% profit per trade over 82 successful trades (with compounding profits) to reach my target.

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts
•
Aussie Stock Forums