meanreversion
Senior member
- Messages
- 3,398
- Likes
- 538
broker = IB
data = eSignal
software = Amibroker
data = eSignal
software = Amibroker
meanreversion
Senior member
- Messages
- 3,398
- Likes
- 538
Yes, eSignal data will go back 3 years if you upgrade, BUT this data cannot be used on 3rd party software, which is beyond moronic.
meanreversion
Senior member
- Messages
- 3,398
- Likes
- 538
I'm asking them now.
But how retarded is it .... I can get data going back 6mths in Amibroker but not 3yrs..
But how retarded is it .... I can get data going back 6mths in Amibroker but not 3yrs..
Adamus
Experienced member
- Messages
- 1,898
- Likes
- 97
3 sets of historical data
I'm going to try posting some results here from backtesting the same trading system on different data:
I expect that table will have the format scrambled by trade2win but hopefully it shows the differences.
This is 10 pairs on one year of data of 1 minute bars using my version of the 3Ducks system.
IB and FXCM data give pretty much the same sort of results surprisingly. Often FXCM data is just as different from IB as DTC.
This helped me appreciate what I'd known theoretically for a long time but hadn't internalised, that backtesting results are nothing. I just hope that the systems that backtested and then forward test profitably also turn a profit in real time.
NB NinjaTrader spits out the average of the max drawdown for the basket - not the max max drawdown.
This system is good, because it wins on all three data, but I have some systems which looked great on one data but were just miserable on the other sets.
I guess the message here is that it's good to have two sources of data to backtest on.
I'm going to try posting some results here from backtesting the same trading system on different data:
Code:
[FONT=Courier New]Provider Profit Avg Trd Total Net Max. DD # Trd Hit Avg Win Avg Loss
Factor[/FONT][FONT=Courier New] Rate
FXCM 1.15 52.11 69925.70 -11032.15 1342 28% 1480.85 -491.76
DTC 1.05 21.20 29044.57 -12602.23 1370 27% 1526.17 -527.42
IB 1.16 57.63 76593.70 -9724.35 1329 28% 1488.63 -488.29
.
.[/FONT]I expect that table will have the format scrambled by trade2win but hopefully it shows the differences.
This is 10 pairs on one year of data of 1 minute bars using my version of the 3Ducks system.
IB and FXCM data give pretty much the same sort of results surprisingly. Often FXCM data is just as different from IB as DTC.
This helped me appreciate what I'd known theoretically for a long time but hadn't internalised, that backtesting results are nothing. I just hope that the systems that backtested and then forward test profitably also turn a profit in real time.
NB NinjaTrader spits out the average of the max drawdown for the basket - not the max max drawdown.
This system is good, because it wins on all three data, but I have some systems which looked great on one data but were just miserable on the other sets.
I guess the message here is that it's good to have two sources of data to backtest on.
Adamus
Experienced member
- Messages
- 1,898
- Likes
- 97
Forex Morning Trade
Brilliant! Now I'm worried all most posts from now on will be in Courier New. :jester:
Just lurking on the ForexMorningTrade journal to see what they're up to. I've been watching that thread for a while now and I can't work out whether it's really just a little cabal of users who never post anywhere else on t2w or if it's an elaborate marketing ploy with multi-nics by the system vendor who started it.
They're all newbies with 0 - 100 posts and they all love the system, but they occasionally post stuff that you just wouldn't ever make up if you were trying to fake a whole thread. It's a shame it's a black box system because it looks interesting.
But just recently one of the posters there put up a really interesting PDF about day-of-the-week and hour-of-the-day trade timing, which is something I've got on my list of things to do.
forexcheatsheet.pdf
it reads bizarrely like a get-rich-quick website but it isn't trying to sell anything.
There's something about this concept which bugs me.
NinjaTrader has the facility to see how much profit is generated per hour of the day so my natural inclination is to put up a filter to make my systems trade at those times of day and not others. But this feels like over-optimisation.
If there's an obvious link between time of day and profitability, I'd use it, but I can only think that it makes sense for a couple of markets. For instance, I'd be happy to trade JPY-AUD only midnight to 6AM and EUR-USD 9AM to 17PM (GMT+1).
I don't even trade JPY-AUD though. Or USD-MexPeso.
JPY-USD though should be tradeable all day long.
I guess when I've got some real results I'll look at it again.
Brilliant! Now I'm worried all most posts from now on will be in Courier New. :jester:
Just lurking on the ForexMorningTrade journal to see what they're up to. I've been watching that thread for a while now and I can't work out whether it's really just a little cabal of users who never post anywhere else on t2w or if it's an elaborate marketing ploy with multi-nics by the system vendor who started it.
They're all newbies with 0 - 100 posts and they all love the system, but they occasionally post stuff that you just wouldn't ever make up if you were trying to fake a whole thread. It's a shame it's a black box system because it looks interesting.
But just recently one of the posters there put up a really interesting PDF about day-of-the-week and hour-of-the-day trade timing, which is something I've got on my list of things to do.
forexcheatsheet.pdf
it reads bizarrely like a get-rich-quick website but it isn't trying to sell anything.
There's something about this concept which bugs me.
NinjaTrader has the facility to see how much profit is generated per hour of the day so my natural inclination is to put up a filter to make my systems trade at those times of day and not others. But this feels like over-optimisation.
If there's an obvious link between time of day and profitability, I'd use it, but I can only think that it makes sense for a couple of markets. For instance, I'd be happy to trade JPY-AUD only midnight to 6AM and EUR-USD 9AM to 17PM (GMT+1).
I don't even trade JPY-AUD though. Or USD-MexPeso.
JPY-USD though should be tradeable all day long.
I guess when I've got some real results I'll look at it again.
meanreversion
Senior member
- Messages
- 3,398
- Likes
- 538
Adam, I haven't combed through the recent posts, are you live on a system yet?
Adamus
Experienced member
- Messages
- 1,898
- Likes
- 97
I was but I pulled the plug on them after getting clobbered on the results for 2 weeks and simultaneously discovering that the variability in the data I was just talking about was all against me for the systems I had live.
I've got them back on simulation after giving them all a wash and brush-up looking at the results on all 3 historical data sets I've got.
Lame perhaps but I lost the confidence I had in it all. It's hard to switch the system back on when you don't trust the results anymore and it just lost you a $1000 the day before.
meanreversion
Senior member
- Messages
- 3,398
- Likes
- 538
How many degrees of freedom do you have in your system?
Have you read "Trade your way to financial freedom" by Van Tharp? I'm in the process of re-reading it.. he's interviewed many "master" traders in his time, and his conclusion is that the system is about 10 pct of what makes a successful trader.. 60 pct is psychology and the balance is money management.
Have you read "Trade your way to financial freedom" by Van Tharp? I'm in the process of re-reading it.. he's interviewed many "master" traders in his time, and his conclusion is that the system is about 10 pct of what makes a successful trader.. 60 pct is psychology and the balance is money management.
Adamus
Experienced member
- Messages
- 1,898
- Likes
- 97
You've forgotten the hidden 90% - hard work. I'm not making any claims about that though, taking into account the amount of distractions around me.
Tell me what a degree of freedom is and I'll tell you how many I've got.
If you're referring to me "giving my systems a wash and brush-up" for the different data - then I don't know how to count the things I did. I made substantial changes to the algorithms I used to avoid as much reliance on highs and lows, which tend to be the cause of the differences between the historical datasets.
Actually on one of the systems I have that was quite profitable in backtesting and even in simulation but which blew up badly, I haven't found a good way to change what it does yet. I have to avoid spending all my time on analysis of this situation - I think it's actually easier to implement some of the new ideas I've got and create totally new systems.
Tell me what a degree of freedom is and I'll tell you how many I've got.
If you're referring to me "giving my systems a wash and brush-up" for the different data - then I don't know how to count the things I did. I made substantial changes to the algorithms I used to avoid as much reliance on highs and lows, which tend to be the cause of the differences between the historical datasets.
Actually on one of the systems I have that was quite profitable in backtesting and even in simulation but which blew up badly, I haven't found a good way to change what it does yet. I have to avoid spending all my time on analysis of this situation - I think it's actually easier to implement some of the new ideas I've got and create totally new systems.
meanreversion
Senior member
- Messages
- 3,398
- Likes
- 538
A degree of freedom is simply a parameter or variable you can change. So, for example, a system which stakes 1 pct of equity when a 10 day MA crosses over a 50 day MA has 3 degrees of freedom. Generally, any system with more than 5-6 degrees of freedom (I know it's not many) will be less likely to work over several markets under varying conditions.
There is no question that developing a system takes hard work. I spent about 3-4 months researching and backtesting before trading a medium term trend system, at the start of this year.
There are a limited number of ways to trade. You're either looking for breakouts/trends, or range behaviour/reversions to the mean. Let's say you choose breakouts.. whether you buy on the breakout of a Bollinger band, or a Donchian channel, it's kind of the same thing.. over time the systems will look broadly the same.
The real challenge comes in sticking to the system. If you design a system with positive expectancy, then you might save money by skipping one signal, but if you keep doing it you will eventually come a cropper. This is the psychology side, and it's far more important than the system itself.
Think about backtesting.. you're trying to find something that has performed well in the past. Almost by definition, it's unlikely to perform as well in the future.
So in summary --- don't spend TOO long on the system development side.. that's kind of the easy part compared to the mental challenge of sticking with the system. You mention a system that looked good in testing then "blew up badly".. this is exactly what I mean. You have to suck it up through the drawdown and trust in the positive expectancy.. this is the hard part, believe me.
There is no question that developing a system takes hard work. I spent about 3-4 months researching and backtesting before trading a medium term trend system, at the start of this year.
There are a limited number of ways to trade. You're either looking for breakouts/trends, or range behaviour/reversions to the mean. Let's say you choose breakouts.. whether you buy on the breakout of a Bollinger band, or a Donchian channel, it's kind of the same thing.. over time the systems will look broadly the same.
The real challenge comes in sticking to the system. If you design a system with positive expectancy, then you might save money by skipping one signal, but if you keep doing it you will eventually come a cropper. This is the psychology side, and it's far more important than the system itself.
Think about backtesting.. you're trying to find something that has performed well in the past. Almost by definition, it's unlikely to perform as well in the future.
So in summary --- don't spend TOO long on the system development side.. that's kind of the easy part compared to the mental challenge of sticking with the system. You mention a system that looked good in testing then "blew up badly".. this is exactly what I mean. You have to suck it up through the drawdown and trust in the positive expectancy.. this is the hard part, believe me.
Adamus
Experienced member
- Messages
- 1,898
- Likes
- 97
You are (probably quite justifiably) assuming the worst about my actions - however I did mention the data situation - I had more reason to pull the plug on the system that just a bad 2 weeks.
I subscribed to FXCM because the IB historical data servers were absolutely appalling - I had connectivity issues and even when I didn't have those issues, IB kept on stopping NinjaTrader with "pacing violations" which is the IB euphemism for its bandwidth throttling to prevent customers overloading their puny little historical data servers.
NinjaTrader can't start running an automated trading system without downloading lots of data for it, so I had real problems just starting my systems.
The FXCM subscription solved that issue brilliantly and also gave me access to a month of tick data and 5 years of 1 minute data. When I repeated the backtests for my system, it tested unprofitable. So on one set of data (Disktrading) it was good, and on another set it was bad. As far as I was concerned, I had no credible positive expectancy anymore.
meanreversion
Senior member
- Messages
- 3,398
- Likes
- 538
Fair enough, I understand you've had technical issues. Nonetheless, the original thread title was "Deadline June". I've found it useful to set myself deadlines and to try to get there.. I'm not suggesting for a minute you aren't busting your gut, but when do you think you'll be properly up and running, October?
Adamus
Experienced member
- Messages
- 1,898
- Likes
- 97
I keep forgetting I called the journal 'deadline june'. Take it as an artifact of the T2W journal implementation.
I should have just called it 'The Adamus Journal'.
It did serve a very useful purpose in helping me to get my show on the road. Admittedly my show is not on the road, but I'm at a stage where I can't really improve my working approach anymore by heaping a sense of time pressure on myself. I used to waste days (even years) on irrelevancies. Now my time is wasted just by technical issues and I have learnt to prioritize and focus much better.
As for when I reckon I will get the show on the road, I don't know. Next week probably. It all depends on how that "positive expectancy" is looking.
I should have just called it 'The Adamus Journal'.
It did serve a very useful purpose in helping me to get my show on the road. Admittedly my show is not on the road, but I'm at a stage where I can't really improve my working approach anymore by heaping a sense of time pressure on myself. I used to waste days (even years) on irrelevancies. Now my time is wasted just by technical issues and I have learnt to prioritize and focus much better.
As for when I reckon I will get the show on the road, I don't know. Next week probably. It all depends on how that "positive expectancy" is looking.
Adamus
Experienced member
- Messages
- 1,898
- Likes
- 97
"Ours is an uncertain world, though fortunately not all things are equally uncertain."
I just started reading an article on Bayesian statistics, and when the first sentence makes me stop reading and scratch my head, I usually stop. I should have stopped this time. The second sentence was completely incomprehensible :jester:
Anybody got any interesting references for Bayesian statistics? There's no Bayesian Stats for Dummies book.
I just started reading an article on Bayesian statistics, and when the first sentence makes me stop reading and scratch my head, I usually stop. I should have stopped this time. The second sentence was completely incomprehensible :jester:
Anybody got any interesting references for Bayesian statistics? There's no Bayesian Stats for Dummies book.
Last edited:
meanreversion
Senior member
- Messages
- 3,398
- Likes
- 538
Are you completely comfortable with the mechanical side of things, i.e. order proccessing, trade management etc., it's just the actual trading rules you need to sort out?
Re: 3 sets of historical data
Adamus these results look encouraging! You should forward test your strategy for a month to see if it matches up with the backtests. What lot size are you using in your testing? How many pairs show positive results over the test period? Looks like you might be into something here....
Adamus these results look encouraging! You should forward test your strategy for a month to see if it matches up with the backtests. What lot size are you using in your testing? How many pairs show positive results over the test period? Looks like you might be into something here....
Adamus
Experienced member
- Messages
- 1,898
- Likes
- 97
I'm satisfied I can handle the mechanical side of things, although I'm not comfortable with it. NinjaTrader 7 beta will no doubt cause me some losses and there's no doubt it costs me a lot of time in reporting the bugs, but there's a good chance that they will manage to stabilise the product in the near term future.
I'm also satisfied with the trading rules for the 3Ducks system, although I would be happier if there was less risk and I could probably achieve that if I did some more work on the exits. I'm currently unable to find a trailing stop that I'm happy with.
