my journal 2

Status
Not open for further replies.
bbmac said:
On a seperate subject.

I understand some if not all of the rationale you use to decide which of the total available mechanical systems in the portfolio should be trading live to make up the present combination of systems...but in a sense this choice be it with a methodology/rationale/parametres/rules still involves a human decision, ie yours. To that extent the whole rationale of having a mechanised trading portfolio of systems is underpinned by human intervention/choice as to which systems are actually trading at any given point. Ie it is still a discretionary decision in the same way that I make individual trading decisions based on each set-up that presents itself and it's adherance to certain rules/parametres - but still discretionary decisions involving human intervention/choice - ie my own.

Are there any plans, (and indeed could it be done) to mechanise this system choice/mix of the total available via some kind of algorithm that makes those decisions mechanically and may alter the mix as you currently do discretionarily ?

Interested to hear your thoughts on this and whether you consider this to be a a potential weakness / chink in the armour ?

BBmac.
Click to expand...

Replying as I read.

The only problem with discretion is if it requires more work or if it works less well. But in this case, coming up with a formula that entirely automates the systems selection process is:

1) longer than forever picking the systems with a pinch of discretion (and some rules that are fixed but haven't been written - e.g. that the system has been very profitable in back-testing or very profitable in forward-testing or a combination of the two).

2) more profitable than devising and using a formula that (with my present skills) cannot successfully replace my guesstimates

So, until a formula will mean 1) more work and 2) less profitability, then it will be better to not automate the systems selection process. As I said, I am not a Math professor nor any type of academic and I constantly use rule-of-thumb estimates to make money as fast as possible.

And, so, to answer your question, creating a formula right now, with my limited skills, would be a weakness for sure, because it would give me a false feeling of science and precision, and, since the formula would be flawed, it would make things worse.

Let's not forget to say that I have already tried doing what you suggest, several times during the past few years (and subsconsciously I am always thinking about it, without even realizing it), and that is why I am saying what I am saying: I believe I am not ready for that.

But if I had 500 systems instead of 120, and if I were forced to automate because it then becomes more practical to do so, then my formula would be along these lines:
1) enable all systems that have been profitable in forward-testing
2) do a review once every 3 months and change things according to #1.
 
Done with taking care of the internet stick. Now everything works correctly.


These Italian phone companies all have labyrinthine websites. It's a nightmare. I finally figured it out, after having to go the goddamn store to be identified, like at a police station. Damn. It bothered me so much, having to go there with my passport and all.

Now that i am entirely done with the job of checking and double-checking the supplementary secondar internet stick for the careless friends who will come to visit me in September, i can focus on the next parcel. So the next thing to do is go downstairs to the doorman and see if there's more parcels for me.
 
Last edited:
Yeh, someone once referred to it as the 'discretionary part of mechanical/automated trading' and indeed this was discused by you and another poster earlier in this journal.

In a sense your discretionary choosing of the automated systems mix/combination that are traded is potentially harder than a pure discretionary trader's choices, because on the face of it you arguably have more variables with 120 systems to choose from.

For eg: I am a discretionary trader, The principle that underpins my trading edge is confluence. The edge involves pre-identifying potential support/resistance/sbr (support becomes resistance/rbs (resistance becomes support) and then looking for one of a number of repeating indicator and Price Action based patterns (and combinations thereof) on t/f's below it - at certain combinations of the factors that make up the potential support/resistance ( ie: previous price pivots, fibs, trend lines mainly,) with a price action trigger to effect a market entry. I have a set of rules regarding which type of repeating set-ups I look for for at which combination of the pre-identified potential supp/res/sbr/rbs factors and these are divided into ' with next t/f trend (+) after a pullback ' and ' counter trend ' rules. These rules are based on the highest probability incidence of a successful outcome given these combinations over a very large sample, and with experience they have become 2nd nature to me...even to the stage where I am quickly able to rate a set-up on a pre-defined scale A+, A-, B+ or B (which determines the risk used.) If the set-up does not fall into the ratings mentioned it is ignored and I stay flat. My rules are as simple as I can make them and are pretty much set in stone, but I still have to make a decision to enter the market based on the set-ups's adherance to them and subsequent rating. So my discretion arguably comes down to 5 choices, ie enter the market at the relevant risk based on an A+, A-, B+ or B rated set-up or stay flat. In effect the choice like all trading comes down to enter the market or don't enter the market Lol but I only do so under 4 pre-determined and repeating circumstances, is another way of putting it. (Of course this all pre-suposes that my patienmce and discipline is good at all times - It isn't of course and like most, I make mistakes and lapse, but they becomes less frequent with time/age/experience)

But having 120 systems in a sense gives you 120 choices...I appreciate what you say about discretion and guestimates etc but I was wondering how many rules/parametres would you say you used in making this decision ? I know this information is spread across this journal in various posts but are you clear what these parematres are and if you had to say what % of the decision to add a ssytem into the live trading mix are 'hard and fast rules' and what % is 'guestimate/discretionary/experience/gut feeling ?'

Thanks again

BBmac.





travis said:
....So, until a formula will mean 1) more work and 2) less profitability, then it will be better to not automate the systems selection process. As I said, I am not a Math professor nor any type of academic and I constantly use rule-of-thumb estimates to make money as fast as possible.

And, so, to answer your question, creating a formula right now, with my limited skills, would be a weakness for sure, because it would give me a false feeling of science and precision, and, since the formula would be flawed, it would make things worse.....
Click to expand...
 
travis said:
But if I had 500 systems instead of 120, and if I were forced to automate because it then becomes more practical to do so, then my formula would be along these lines:
1) enable all systems that have been profitable in forward-testing
2) do a review once every 3 months and change things according to #1.
Click to expand...

This is a problem that I really struggle with. Like you I trade multiple systems (more acturately I trade multiple instances of ONE system, but with multiple parameters). Currently I trade them all, and I let the position sizing sort out the wheat from the chaff. The sytems that do well trade at increasingly larger size, and the ones that suck would eventualy trade to the point where they hit max drawdown limits.

I've fannied about with a whole bunch of stuff, such as allocating larger positions to systems that are producing greater than average return, or average + sigma, allocating seperate accounts to long and short trades etc etc. The basic problem is, if a sytem has positive expectancy (however small the edge may be) you want to allocate as much capital to it as possible. Starving the method of funds because you have money allocated to something that's doing better often wont work becuase the gains from these crappy sytems running on reduced account balances are insufficient to cover the larger dollar losses that arise from better systems trading larger positions in periods of drawdown. Then again, sometimes this approach does work ! but getting a sufficient sample size to do any sort of analysis is practically impossible.

It could also be argued that the benefits of diversification can be achieved with a relatively low number of systems, and one of the things I often look at is the range of returns I might have theoretically achieved with a half the number of sytems, a quater the number of sytems, one eight the number of systems etc. I think I've come to the concluion that anything more than 20 probably doesnt contribute very much (other than additional work !)

Im my case this stuff gets a bit bizarre becuase I employ a random entry, and so maybe the distribution in returns is a little more random than others experience, although I'm not too sure about that having seen the distribution in other traders results.

I suppose the key point that is whatever method you employ, even if its based on hard and fast rules, the choice of those rules is completely discretionary. Actually the situation is more coplex than that because for example I might want to optimise drawdown rather than returns, but that decision in itself is totally arbitary, and at my on discretion !
 
bbmac said:
Yeh, someone once referred to it as the 'discretionary part of mechanical/automated trading' and indeed this was discused by you and another poster earlier in this journal.

In a sense your discretionary choosing of the automated systems mix/combination that are traded is potentially harder than a pure discretionary trader's choices, because on the face of it you arguably have more variables with 120 systems to choose from.

For eg: I am a discretionary trader, The principle that underpins my trading edge is confluence. The edge involves pre-identifying potential support/resistance/sbr (support becomes resistance/rbs (resistance becomes support) and then looking for one of a number of repeating indicator and Price Action based patterns (and combinations thereof) on t/f's below it - at certain combinations of the factors that make up the potential support/resistance ( ie: previous price pivots, fibs, trend lines mainly,) with a price action trigger to effect a market entry. I have a set of rules regarding which type of repeating set-ups I look for for at which combination of the pre-identified potential supp/res/sbr/rbs factors and these are divided into ' with next t/f trend (+) after a pullback ' and ' counter trend ' rules. These rules are based on the highest probability incidence of a successful outcome given these combinations over a very large sample, and with experience they have become 2nd nature to me...even to the stage where I am quickly able to rate a set-up on a pre-defined scale A+, A-, B+ or B (which determines the risk used.) If the set-up does not fall into the ratings mentioned it is ignored and I stay flat. My rules are as simple as I can make them and are pretty much set in stone, but I still have to make a decision to enter the market based on the set-ups's adherance to them and subsequent rating. So my discretion arguably comes down to 5 choices, ie enter the market at the relevant risk based on an A+, A-, B+ or B rated set-up or stay flat. In effect the choice like all trading comes down to enter the market or don't enter the market Lol but I only do so under 4 pre-determined and repeating circumstances, is another way of putting it. (Of course this all pre-suposes that my patienmce and discipline is good at all times - It isn't of course and like most, I make mistakes and lapse, but they becomes less frequent with time/age/experience)

But having 120 systems in a sense gives you 120 choices...I appreciate what you say about discretion and guestimates etc but I was wondering how many rules/parametres would you say you used in making this decision ? I know this information is spread across this journal in various posts but are you clear what these parematres are and if you had to say what % of the decision to add a ssytem into the live trading mix are 'hard and fast rules' and what % is 'guestimate/discretionary/experience/gut feeling ?'

Thanks again

BBmac.
Click to expand...

Replying/commenting as I read (feel free to reply or to not reply forever).

Are you kidding me? How is it harder to pick profitable systems than to choose, out of the blue, a trade (entry and exit) to make? It's very easy to detect and identify systems that have been profitable. I can expand on this if necessary, but i don't think it's necessary. This "discretionary part of automated trading" does not really leave much to discretion.

Your subsequent example of how "simple" it is to make a discretionary trade is quite eloquent: it is not simple at all. It took you twenty lines to explain how simple it is for a discretionary trader to make his choices. Can I make my "discretionary" choice appear simpler? I think so. Did such and such system make money? Yes? Then, let's enable it.

How may parameters? Maybe just one is needed: sharpe ratio. Back-tested and forward-tested sharpe ratio above 2? Definitely enable it. Conflicting sharpe ratios? If the sum is above 4, then still enable it. This could be, among many others, one of the automation formulas of the process, if I really had to do it. But, as i said, things work better if I still keep this part discretionary.

Yes, I am clear what these parameters are (total money made in back and forward testing, and sharpe ratio achieved during the same two periods). It is all "hard and fast rules", way harder and faster than anything that has to do with discretionary trading. The discretionary input is so small that i could automate the whole process tomorrow and it would still be profitable, maybe even more. Can I say the same about discretionary trading? Not at all.

Were you going as far as implying that discretionary trading is more automated/mechanical/univocal/unequivocal than automated trading? Impossible to do so.

Bring it on.

 
Last edited:
drawdown lasting forever

Today lost another 600 dollars.

Snap1.gif

Snap2.gif

Only thing left to do for me in rome today is enjoy thunders and rain.


 
Last edited:
Tomorrow I should be getting the helicopter.

The best of supertramp is this guy whistling at minute 3:05:


The guy singing has a face and accent like ed norton in leaves of grass. It probably is ed norton, before he became an actor.

 
Last edited:
I found it interesting that your current period of drawdown matches the pattern (thus far) of the last 'significant/prolonged' period, ie

a-b = decline
b-c = recovery
c-d decline lower than point b
d-e recovery higher than point c
e-f decline but point f higher than point d (deepest part of the drawdown)
f-g recovery exceeding point a.

I find it particularly interesting because it is reminiscant of the principle of one the 3 repeating price action patterns I look for that combine with certain indicator patterns to complete an entry set-up (should the correct combo of potential supp/res/sbr/rbs factors be present.) etc.

Have you noticed any repeating patterns in the drawdown or equity advance ?

jfj4gl.jpg
 
Last edited:
Thanks for the reply...it answers my central question perfectly, and your reply is further illuminating.

Semantics, probably, but to be clear .. '.....Were you going as far as implying that discretionary trading is more automated/mechanical/univocal/unequivocal than automated trading?...' my answer to of course, No !

G/L

travis said:
Replying/commenting as I read (feel free to reply or to not reply forever).

Are you kidding me? How is it harder to pick profitable systems than to choose, out of the blue, a trade (entry and exit) to make? It's very easy to detect and identify systems that have been profitable. I can expand on this if necessary, but i don't think it's necessary. This "discretionary part of automated trading" does not really leave much to discretion.

Your subsequent example of how "simple" it is to make a discretionary trade is quite eloquent: it is not simple at all. It took you twenty lines to explain how simple it is for a discretionary trader to make his choices. Can I make my "discretionary" choice appear simpler? I think so. Did such and such system make money? Yes? Then, let's enable it.

How may parameters? Maybe just one is needed: sharpe ratio. Back-tested and forward-tested sharpe ratio above 2? Definitely enable it. Conflicting sharpe ratios? If the sum is above 4, then still enable it. This could be, among many others, one of the automation formulas of the process, if I really had to do it. But, as i said, things work better if I still keep this part discretionary.

Yes, I am clear what these parameters are (total money made in back and forward testing, and sharpe ratio achieved during the same two periods). It is all "hard and fast rules", way harder and faster than anything that has to do with discretionary trading. The discretionary input is so small that i could automate the whole process tomorrow and it would still be profitable, maybe even more. Can I say the same about discretionary trading? Not at all.

Were you going as far as implying that discretionary trading is more automated/mechanical/univocal/unequivocal than automated trading? Impossible to do so.

Bring it on.
Click to expand...
 
bbmac said:
I found it interesting that your current period of drawdown matches the pattern (thus far) of the last 'significant/prolonged' period, ie

a-b = decline
b-c = recovery
c-d decline lower than point b
d-e recovery higher than point c
e-f decline but point f higher than point d (deepest part of the drawdown)
f-g recovery exceeding point a.

I find it particularly interesting because it is reminiscant of the principle of one the 3 repeating price action patterns I look for that combine with certain indicator patterns to complete an entry set-up (should the correct combo of potential supp/res/sbr/rbs factors be present.) etc.

Have you noticed any repeating patterns in the drawdown or equity advance ?

jfj4gl.jpg
Click to expand...

Yes, but not in as much detail as your drawing above. These are the 5 periods in which I would divide the equity line, also considering that it is a consequence of different combinations of contracts/systems:

1) period from June 2010 to late December 2010: very little capital allocated plus some mistakes in choosing the systems to trade and the equity curve just goes up and down endlessly for 6 months.

2) same capital, but better systems and more luck: long draw-up from late december 2010 to mid april 2011.

3) shorter drawdown, shorter drawdown ever, since start

4) long draw-up from early may to late June

5) long drawdown ever since June 16th. It is starting to frustrate me.

Snap1.gif

If we forget about the first six months of tests, I would say the time it goes up is more than the time it goes down.

Also I would say (all guesstimates) that it goes up as fast as it goes down.
 
bbmac said:
Thanks for the reply...it answers my central question perfectly, and your reply is further illuminating.

Semantics, probably, but to be clear .. '.....Were you going as far as implying that discretionary trading is more automated/mechanical/univocal/unequivocal than automated trading?...' my answer to of course, No !

G/L
Click to expand...

Yes, and by the way I am just about ready to automate the choice of systems. If I had to come up with a formula it would be this. Back-tested sharpe ratio + forward-tested sharpe ratio > 4. E.g.: They're both 2, one is 3 and the other one is 1... one is 4 and the other one is zero.

Something like this:

comparison.gif

The average sum of the enabled ones is 6.50.

The average sum of the non-enabled ones is 2.92. So far this principle I thought I have applied (unintentionally) still holds.

Except for 4 exceptions of enabled systems where the sum is below 4.

It is almost automated, but you see, we're doing this without even realizing it. But I can see this is the principle that almost unconsciously we've been following.

I have to admit that you opened my eyes with your questions.

Oh, wait. But then we cannot forget about the fact that we also keep in mind how well the combination sticks together in terms of combined drawdown, so once again we're not ready to automate this. But the rule above of sum > 4 could be a requirement: not enough but necessary.

[...]

Writing from work.

I have figured it out and implement it as soon as I'll get home:

HOMEWORK

I will do a scatter plot that will represent the number of trades on the x-axis and the average sharpe ratio (between back-tested and forward-tested). The scatter plot should represent all the systems, from the best to the worst. Enough with not showing bad systems on it.

Any system thas has made more than 10 trades and has an average sharpe ratio above 2 should be trading. But the exact moment they'll be enabled will depend on how well they fit together in a combination, in terms of combined profit and drawdown.

It would be ideal if I could mark in red font all those systems that are being traded. I'll need to do a web search for this. Maybe I could simply manually change the font and maybe it will stick to it even after they get updated with different sharpe ratio values.

[...]

Ok. Back home.

The scale doesn't allow me to keep them all on one chart and see the labels clearly. So I split them in two charts, one with the traded systems:

traded.gif

And one with the untraded systems:

untraded.gif

If you download them and switch back and forth from one chart to another, you can see the tendency of the traded ones to be higher points (better sharpe ratios) and more to the right (more trades). That's indeed what makes them more reliable and why I chose them.

However, both charts are only showing the systems that have good performances, in that they appear in the right area of the chart as far as sharpe ratio. There are dozens more that appear below the sharpe ratio value of 1.

So this also means that all the untraded systems, with blue font on the second chart, should be considered for trading. Right now there's only 27 systems being traded.

Oh, and needless to say, now the charts above are all automated.

The way I'd use the second chart, with untraded futures, would be to monitor it periodically so to be alerted as to good systems I'm not trading, and for example quickly ask myself: why am i not trading these circled in red?

untraded2.gif
 
Last edited:
The Swingle Singers Music Video Piazzolla 'Libertango'

http://www.youtube.com/watch?v=-uiG5jJavTU

Masterpiece as usual. These guys deserve to start making movies, musicals, and become famous. I mean much more famous than they already are. I mean fashionable, so successful that even idiots who don't know them start appreciating them like the sheep do. Because idiots only appreciate things if everyone else around them is already appreciating them.

But maybe it's better like this. If the idiots start appreciating them, the swingle singers will start making compromises and their stuff will become like a movie with tom cruise in it: cheap laughs, stupid, low... good for almost everyone, which means crap for me.
 
I am really starting to wonder if this is a good idea to have a server far away. The problems are increasing lately. Power outages, once every 4 months (that can be taken care of by buying a more powerful UPS). Connection down, used to never happen, but now it's happening about once a month.

I am starting to get stressed out and paranoid.

I was very satisfied for over a year and now suddenly I find myself wondering if there can be a better solution, such as running things from home. I don't think so. I can't run things from here, because on the laptop is heavy, it breathes heavy, the connection is even less reliable... the only advantage is that power outages will never be an issue.

Damn. I have to accept this, with whatever problems it will bring. I don't trust anyone else to run a server with my systems on it. No way. Just this one friend I know from many years ago, and who helped me with building trading systems among the other things.

Still, this doesn't keep me from being frustrated. It's as if there was a plot to keep me from making money, now that everything seems just perfect. People are plotting to take success away from me.

I mean: we've had more problems in one month than 1 year. And now we're much more organized than we used to be (computer never failed, excel never failed, and so on).

Power outage two days ago. Today there was a connection failure in the Twin Cities area. A month ago there was a problem with the ISP company, doing business with my friend's company.

Will my life as automated trader ever get "easy"? I just want to run things smoothly as it was meant to be from the start.
 
Last edited:
major tcp-ip counter-attack from me

Here's what I am going to devise with the help of my friend.

We'll run two servers. Neither from my home. One from his house, as now, and another one, a back-up server, we'll run from someone else's house. Maybe the backers or something like that.

Both servers will need to upload the file to one another all the time. When one fails, the other one takes over: it starts TWS and then it runs excel.

I need to do this if I want to stop worrying about power failures and disconnections. Goddamn. It's going to cost twice as much though.

[...]

What the **** do i do now?

The problem with the server is still all there.

Basically a couple of closing trades tonight were screwed. All openings of overnight trades were screwed (problem started at 21.40 CET).

I am starting to get pissed off. But as usual, once I have problems, I solve them.

I think I am going to devise something that sends me an email when the server works, but then... too many emails. That sucks. Wait. I have an idea. No. Just more problems.

But not only that. I need something to upload to me, once an hour, the latest workbook saved, so I can run it from here whenever there's a problem there.

But I can't use dropbox (which would work) or similar software -- not safe. I need something that only involves me and the main server. Problems seem to never end.

With five trades going astray, or five stray trades... or... you get my point. With things out of control, what do I do now? Do i try to go to sleep?

Do I stay up until things are ok again? What do I do?

I know I am pissed off. I don't know how to solve this problem. Not in any convenient way. I could set up an email to be sent to me with the file every hour, but that's a messy solution, which will cause more bugs than solve.

What the ****...! I was running everything fine with the goddamn maid here, causing power outages, with my crappy pc... and now with my buddy ISP owner... top server at his house... in one month more problems than in the last 2 years.

And guess what. Now it's even too late to connect from here and close the trades manually from this computer.

Now I have to wait until after midnight, because this is the one time markets are closed.

I hope at least something good and final will come out of this. A solution which will ensure more security. Let me know if you have any advice. Anyway I have asked advice to my friend who's the ISP owner himself, so if he doesn't give me the right advice to fix this problem for good, I don't think anyone else will.

I hope you're getting my point. Unlike what one might imagine:

1) we're not talking about a windows problem: nope, never had one.
2) we're not talking about a programming bug: nope.
3) we're talking about power outages
4) we're talking about disconnection within the range of 5 miles, then 50 miles, then 100 miles.

Each time it has not been a problem with anything we did wrong, but with the whole area.

So the only solution is that we need 2 servers, constantly backing each other up. But tws can only run on one server at a time. So this needs to be fixed as well.

I need a software like dropbox that constantly keeps me connected to the server and checks if the connection is fine, and if it's not fine, it warns me and then I run things from here. Damn. Does that software exist? I think so.
 
Last edited:
Ok, all solved. I talked to my friend: we'll get a secondary connection and a bigger UPS.
 
Thanks for the feedback.

Anyway, after hours of discussion, I was wrong again: we're buying a simple laptop. 400 dollars, no UPS concerns, no external wifi routers needed. No screen needed. No mouse... all set, ready to go. And just 400 dollars.
 
Last edited:
Status
Not open for further replies.

Similar threads

JTrader
  • Poll
Ignore Function functionality suggestion
Replies
10
Views
2K
shadowninja
shadowninja
J
Jay's option trading journal
2
Replies
33
Views
5K
wuming79
W
Yamato
my journal 3
255 256 257
Replies
5K
Views
582K
new_trader
new_trader
HowardCohodas
  • Locked
Reprise - Watch HowardCohodas Trade Index Options Credit Spreads
2
Replies
27
Views
6K
barjon
barjon
N
Ninja's Journal
Replies
5
Views
2K
Ninja377
N
Top