Deadline June

[Yamato]

Will you please comment on my long, thoughtful and detailed post? Particularly considering that I was answering your previous post.

 

[Adamus]

Regarding the discrepancy, I don't think it's big. But there is a discrepancy. If an order is triggered even just 2 seconds later, which is the case with my two servers probably, then (on my simple systems) you're going to have that kind of discrepancies: 20% of 1-tick discrepancies and 1% of orders not triggered or going in the opposite direction (long instead of short). So if that's a big discrepancy, which it isn't in my opinion, then you're never going to be happy with any comparison between 2 different sets of data, because when back-testing on 2 different sets of data you're definitely going to have more discrepancies than what you have between two identical servers/platoforms/systems 8000 kilometers away whose orders are triggered or received by IB at a distance of 2 seconds, due to either a different computer clock (which in turn causes different moving averages and prices), or different distance from IB.

Basically you saying "like it or lump it", correct me if I'm wrong?

 

[Yamato]

I wanted to know your opinion on my post, and my explanation that what you called "huge" differences on my servers (and your data sets, too, presumably) are not "huge" and are not avoidable. I showed in detail where they come from, what they are due to, and how small they are, even though they constantly happen, with that 20% frequency (and the more serious ones, in 1%). Also, you didn't tell me if and how big your differences are, and what type of differences you're having. Let's say you have 100 trades. On one data set you have 50 long trades and 50 short trades. On the other data set, what happens? Let's say you have an execution price difference. How many ticks on average?

Or feel free to ignore my posts, and next time I'll write something like:
Right on, bro. I'm sure everything will work out...

 

[Adamus]

The reason I don't quote all your post is because I only want to refer to part of it. Many of your posts touch on multiple subjects and often include youtube clips and images from wikimedia, therefore there is no point in quoting it 100%. Unless of course you happen to be in a dialogue with a paranoiac who thinks you are trying to misrepresent him. Despite my legendary status :clap: here on the T2W forum, it seems I haven't earned the right to be considered above misrepresenting what someone else said. Unless it was for humorous purposes of course.

I am actually desperately short of time since it turns out that day trading does in fact take all day or at least it does when you have my level of skill.

Plus everything is a mess round here and I've just lost a profitable system. I guess I just rewrote it and left the unprofitable code in there and forgot. Plus I have a problem with NinjaTrader and IB connecting to the IB data farms. NinjaTrader say it's IB's fault and IB says it's NinjaTrader's fault. So I'm constantly posting on their support boards trying to get a resolution, although I don't actually need it now since I subscribe to IQ-Feed.

So I wasn't ignoring what you said deliberately. What you wrote was interesting and I thought I thanked you for it and the reason why I haven't got any stats on the difference in trades between one dataset and the other is because I didn't think of doing it and it's going to take a bit of time, but thanks for suggesting it and I guess I should do it.

 

[Adamus]

Just got a reply from IQFeed about the way they compile their minute bars from the live feed (compare it to IB):

Our data is true tick by tick.

So every single trade is used to make up the minute bars.

 

[Adamus]

And NinjaTrader is came back with info about their Historical Data Manager, which by their account does what any sentient being would expect it to, unless of course you are an IB programmer:

Yes, it will save every tick that we receive from the connected connectivity provider.

 

[Adamus]

Comparative Analysis of System Performance using different Intraday Data

The main reason for carrying out an analysis of system performance using different data is to provide an estimate of the extent of the degradation of PNL figures from backtesting to real-time trading, and an overview of the nature of the changes to be expected, e.g. what % of trades will be the same, what % nearly the same, what % will just not correspond to the other PNL at all.

I'll edit this intro more as I think of things tonight, and maybe even start it.

I have put together a list of all the systems I have that are profitable and noted what the backtest results and against which data I backtested them. I need to trade them live to get a representative sample of trades that provide a PNL result set that I can use in this analysis.

I shall put all the PNL into a spreadsheet and use an average of the PNL for all systems over each data. So I should have an average PNL for all systems for the following data:

(1) Disktrading tick data from 2000-2010
(2) FXCM 1min bar data from 2006-2010
(3) IB sampled 1 min bar data from Jan '10 to Aug '10
(4) Tenfore 1 min bar data from 2009-2010
(5) Live trading

While running the live trading I could also attempt to collect an unbroken history of live feed tick data from IB using the NinjaTrader Historical Data Manager, but that depends on being able to keep my hardware, sofware and internet connection going for that long. All my attempts so far have failed.

 

[Yamato]

Yeah, I understand that you're busy and can't reply to everything. But it was the second or third time you ignored my thoughtful posts, so I was almost getting offended by it (I later re-wrote my post to make it friendlier). Next time I'll try to write a little less so I won't get upset if you fail to appreciate... or rather "forget to show appreciation".

Until then, right on, bro. Cool thread. :clap: :cheesy:


I hate this crap, but at least now I am a regular user and I won't be paranoid about my valuable posts being ignored.

 

[Yamato]

Re: Comparative Analysis of System Performance using different Intraday Data

The main reason for carrying out an analysis of system performance using different data is to provide an estimate of the extent of the degradation of PNL figures from backtesting to real-time trading, and an overview of the nature of the changes to be expected, e.g. what % of trades will be the same, what % nearly the same, what % will just not correspond to the other PNL at all.

I'll edit this intro more as I think of things tonight, and maybe even start it.

I have put together a list of all the systems I have that are profitable and noted what the backtest results and against which data I backtested them. I need to trade them live to get a representative sample of trades that provide a PNL result set that I can use in this analysis.

I shall put all the PNL into a spreadsheet and use an average of the PNL for all systems over each data. So I should have an average PNL for all systems for the following data:

(1) Disktrading tick data from 2000-2010
(2) FXCM 1min bar data from 2006-2010
(3) IB sampled 1 min bar data from Jan '10 to Aug '10
(4) Tenfore 1 min bar data from 2009-2010
(5) Live trading

While running the live trading I could also attempt to collect an unbroken history of live feed tick data from IB using the NinjaTrader Historical Data Manager, but that depends on being able to keep my hardware, sofware and internet connection going for that long. All my attempts so far have failed.

I'll be doing similar tests as explained here:
http://www.trade2win.com/boards/trading-journals/85510-my-journal-2-a-127.html#post1226634

 

[Adamus]

trend following the swiss franc

How can you seriously write a trend following system that trades the swiss franc when your results are based on stuff like this?

 

[DionysusToast]

Adamus - why do you need to automate ?

Why not now learn how to trade ?

 

[Adamus]

Adamus - why do you need to automate ?

Why not now learn how to trade ?

I don't need to. I want to. If I can trade one market efficiently by eye, and then I can automate what I'm doing, I can clone my edge and apply to 100 markets.

Don't tell me you would have got that trade just trading by eye? The only reason my algorithm got it was just pure chance. Later on, that market killed me going the wrong direction into that bar and who knows where the backtested stop loss would have got me out in reality?

EDIT: look at it on the 5min:

 

[brettus]

I don't need to. I want to. If I can trade one market efficiently by eye, and then I can automate what I'm doing, I can clone my edge and apply to 100 markets.

Don't tell me you would have got that trade just trading by eye? The only reason my algorithm got it was just pure chance. Later on, that market killed me going the wrong direction into that bar and who knows where the backtested stop loss would have got me out in reality?

Isn't having a automated trading system that works exactly the same as knowing how to trade, except you dont have to click the buttons?

 

[Adamus]

Isn't having a automated trading system that works exactly the same as knowing how to trade, except you dont have to click the buttons?

To be honest I don't really know what DT is actually saying.

 

[DionysusToast]

I don't need to. I want to. If I can trade one market efficiently by eye, and then I can automate what I'm doing, I can clone my edge and apply to 100 markets.

Don't tell me you would have got that trade just trading by eye? The only reason my algorithm got it was just pure chance. Later on, that market killed me going the wrong direction into that bar and who knows where the backtested stop loss would have got me out in reality?

Adamus.

I am 40. I run a software company. I have been programming since I was 13 years old. That's roughly 27 years programming.

I totally understand where you are coming from but you are wrong in your assessment about being able to automate what you can see by eye.

I too took the path you are taking now. I took it because it was what I knew. The problems is 'what I knew' had nothing to do with what will make you profitable.

Global Futures support automated trading. Less than 1% of their system traders are profitable & the ones that are profitable change the parameters/systems day by day. In other words - they still use their discretion.It doesn't matter how much smarter you are than me. It doesn't matter how much smarter you are than Global Futures customers.

The bottom line is that trading is not a problem with a programmatic solution. If you can understand this, you can make money.

If you want to make money day trading and you are using price information alone, then you are missing out on some of the most important information to making money short term.

Lets say for instance you see the best bid at 70 contracts but you see 500 trades go through at that level - without the price dropping - this tells you someone probably has an iceberg order at that level. On the other hand, let's say there's 1,400 contracts at the best bid and 200 contracts go through and the price ticks down - this tells you someone has pulled their orders. If you are attempting to trade short term without this information, then you are fighting a losing battle.

Longer term trading is for someone other than me to fill you in on.

As an IT guy, you know full well you can't automate something without the rules - how would you come up with a spec?

Cast the automation aside - learn to trade & then you'll understand why it's not even necessary to automate.

 

[DionysusToast]

Isn't having a automated trading system that works exactly the same as knowing how to trade, except you dont have to click the buttons?

The reason you don't understand it is because you are not yet making money trading.

Trading is not about having an absolute ruleset to initiate & exit a trade. This is why directional trading, placing outright positions with an automated system is beyond the reach of the retail trader.

When you make money, you realise the importance of nuance, interpretation and discretion.

Would-be traders should understand that discretion is the key and not something to be feared. Of course, a mechanical system would be lovely but that's not to say that discretion is hard to use. I think people just fear a path that requires judgement & discretion which is why there's a million scammers & authors selling mechanical systems.

 

[Adamus]

The bottom line is that trading is not a problem with a programmatic solution. If you can understand this, you can make money.

If you want to make money day trading and you are using price information alone, then you are missing out on some of the most important information to making money short term.

Lets say for instance you see the best bid at 70 contracts but you see 500 trades go through at that level - without the price dropping - this tells you someone probably has an iceberg order at that level. On the other hand, let's say there's 1,400 contracts at the best bid and 200 contracts go through and the price ticks down - this tells you someone has pulled their orders. If you are attempting to trade short term without this information, then you are fighting a losing battle.

Aha. This is the argument that caused BSD to throw his toys out of the pram, IIRC. Book trading.

Cast the automation aside - learn to trade & then you'll understand why it's not even necessary to automate.

Are you trying to turn me to the Dark Side?

Talking of icebergs, you obviously have an iceberg of knowledge on this subject, because there is no logic in your conclusion. You haven't destroyed my faith that I can automate trading. You have merely inspired my belief that there is more to be automated than I previously realised.

 

[brettus]

The reason you don't understand it is because you are not yet making money trading.

Trading is not about having an absolute ruleset to initiate & exit a trade. This is why directional trading, placing outright positions with an automated system is beyond the reach of the retail trader.

When you make money, you realise the importance of nuance, interpretation and discretion.

Would-be traders should understand that discretion is the key and not something to be feared. Of course, a mechanical system would be lovely but that's not to say that discretion is hard to use. I think people just fear a path that requires judgement & discretion which is why there's a million scammers & authors selling mechanical systems.

I am a complete discretionary trader and profitable, but i also do believe in mechanical models. Some extremely successful ones can analyze the liquidity as you say and dont even require any technical analysis. Just because you're an IT guy with 27 years experience it doesn't mean you should be better than anyone else at programming and creating a black box. Variations and parameters are programmable. Many would consider automated trading to be more profitable, due to the fact that it takes away the emotion from trading.

Name the variations and parameters and think about whether or not they are programmable.

 

[DionysusToast]

I am a complete discretionary trader and profitable, but i also do believe in mechanical models. Some extremely successful ones can analyze the liquidity as you say and dont even require any technical analysis. Just because you're an IT guy with 27 years experience it doesn't mean you should be better than anyone else at programming and creating a black box. Variations and parameters are programmable. Many would consider automated trading to be more profitable, due to the fact that it takes away the emotion from trading.

Name the variations and parameters and think about whether or not they are programmable.

Actually - I have 27 years coding experience and I am NOT trying to automate.

Bottom line - you have to learn to trade first. Once you learn to trade, you will find there's no need to automate.

 

[DionysusToast]

Aha. This is the argument that caused BSD to throw his toys out of the pram, IIRC. Book trading.



Are you trying to turn me to the Dark Side?

Talking of icebergs, you obviously have an iceberg of knowledge on this subject, because there is no logic in your conclusion. You haven't destroyed my faith that I can automate trading. You have merely inspired my belief that there is more to be automated than I previously realised.

There is plenty of logic in watching orders being pulled/refreshed on the DOM. The problem is you haven't watched it, nor has it been explained to you the way people are sucked into positions over & over again before being run over. I am fairly sure you could learn it though. It's not rocket science. You certainly would not be able to automate it unless you got your head around the games first.

It's not the dark side - we all trade off information. Some people like to pigeonhole information into certain categories and just look within that category. Some people have an information category called TA and shun everything that doesn't fit in there.

Retail traders that automate generally only have information related to the chart to use because that's all that the retail automation platforms give them to play with. They don't give you DOM info & an n minute bar will not show you how the flow of orders is changing over time.

Good luck with it though.