I am at risk of stating the obvious, but just in case: Have you tried coming up with a method and calibrating it on data up to say, March, then "forward testing" with new data until July and see if it still works? I did some initial experiments with systematic forecasting myself (I am trading discretionary) and in those experiments it worked wonderfully until the last step.
Automated Day Trading
- Thread starter Millionaire
- Start date
I am at risk of stating the obvious, but just in case: Have you tried coming up with a method and calibrating it on data up to say, March, then "forward testing" with new data until July and see if it still works? I did some initial experiments with systematic forecasting myself (I am trading discretionary) and in those experiments it worked wonderfully until the last step.
When i started out, i developed a system on 6 months data. That system failed straight away.
The next system i developed was using 3 years data. That worked better and was profitable in live trading but nowhere near as profitable as the back test suggested.
Now i back test almost 20 years of data. This gives me a lot more confidence of what i can expect over the next 12 months of trading. But there is always a chance of getting a year that under performs anything ever seen in the back test.
That is when you get tested as a trader, do you continue trading the system or do you abandon it for something else? Typically the old system, if the edge is still valid but the drawdown was just an extreme outlier, starts to print money just after you abandon it.
"You're sort of tempted after a certain period of underperformance to tinker with your systems and call up research. Then as soon as you implement the new research you find out, well the older system would have performed better"- Jerry Parker.
Last edited:
I should have included my longest drawdown period in my opening post.
This longest drawdown (8 months) came immediately after the best ever month for the system. Perhaps that was the markets way of reminding us not to get too cocky after a winning period.
+20.3 (Best ever month)
-2.0 <- The drawdown starts in this month
-4.7
-1.6
+5.3
+0.3
-5.3
-1.6
+4.1
+8.3 <- New high water mark made here, 8 months after the drawdown started
+0.6
+7.4
+4.1
This longest drawdown (8 months) came immediately after the best ever month for the system. Perhaps that was the markets way of reminding us not to get too cocky after a winning period.
+20.3 (Best ever month)
-2.0 <- The drawdown starts in this month
-4.7
-1.6
+5.3
+0.3
-5.3
-1.6
+4.1
+8.3 <- New high water mark made here, 8 months after the drawdown started
+0.6
+7.4
+4.1
Last edited:
It greatly helps if you know why your system has worked in the past.
Also, I would assume that if you examine one data set... mine it... you are testing on a totally new set of data (?)
I use backtesting to test ideas not to mine or find them.
The ideas i test are derived from live market observations.
Once the idea has been tested to be valid i then use backtesting again to find good parameter values. eg what stop size worked well, which profit taking exits worked well.
Last edited:
This does seem like a good system. One poster wrote that the trades may have been in a fortunate sequence. If you can export the trades into a CSV format file, you could import them into a product like Market System Analyzer (MSA) and perform Monte Carlo analysis and see if re-ordering the trades greatly affects the equity curve. It also has position sizing tools to help decide how many shares/contracts to trade. It costs $350 and has a 30 day, fully functional trial. I have no financial interest in MSA from Adaptrade, just a satisfied customer. Mike Bryant is the owner/developer and supports it well. I usually get questions answered within hours.
Brooks
Brooks
System performance update for August 2018: -0.2%
Total for 2018 so far have been +22.9%
Jan:+9.00% (BackTest)
Feb: +0.40% (BackTest)
Mar: +4.90% (BackTest)
Apr: +1.70% (BackTest)
May: -2.80% (BackTest)
Jun: +7.80% (Live)
Jul: +1.80% (Live)
Aug: -0.20% (Live)
Since going live at start of June the system is up 9.4%
In the past the system has generated monthly returns between -10% and +20%, with about 70% winning months. Very early days yet for going live but the system seems to be holding up in live trading for the first 3 months anyway. That could all change and i could get some nasty losing months over the next few months. No way to tell really. I just have to continue to take all signals (and not turn the system off or change it in any significant way) and hope for the best.
Psychologically it is very hard not to interfere with the system when it is in a losing or protracted sideways period.
Will update at end of September.
Total for 2018 so far have been +22.9%
Jan:+9.00% (BackTest)
Feb: +0.40% (BackTest)
Mar: +4.90% (BackTest)
Apr: +1.70% (BackTest)
May: -2.80% (BackTest)
Jun: +7.80% (Live)
Jul: +1.80% (Live)
Aug: -0.20% (Live)
Since going live at start of June the system is up 9.4%
In the past the system has generated monthly returns between -10% and +20%, with about 70% winning months. Very early days yet for going live but the system seems to be holding up in live trading for the first 3 months anyway. That could all change and i could get some nasty losing months over the next few months. No way to tell really. I just have to continue to take all signals (and not turn the system off or change it in any significant way) and hope for the best.
Psychologically it is very hard not to interfere with the system when it is in a losing or protracted sideways period.
Will update at end of September.
Last edited:
This is a very true statement. I recognize this.
"Don't fix it if it ain't broken". The trick is to be able to identify when it is broken, and when it is not broken.
This is going to be especially true if this is the only system you are trading. I would not be comfortable trading only one system because if it degrades to unprofitably you are screwed.
But it is much easier to focus on 1 system then on 25. Especially for discretionary traders. On top of that I never found a second system that can compete with what I trade now.
Not the numbers of systems you use but the quality of the system is important. Quality instead of quantity.
I can financially survive with the only system I have. I can managed it optimal. And not much stress nor 100 different things to follow for hours while watching the screen.
Last edited: