my journal 2

This is a discussion on my journal 2 within the Trading Journals forums, part of the Reception category; Originally Posted by travis Thanks for the advice. By the way, read your Larry Hite signature again That's the game ...

Closed Thread
 
LinkBack Thread Tools Search this Thread
Old Apr 18, 2010, 9:50pm   #701
 
dog4's Avatar
Joined Feb 2004
re: my journal 2

Quote:
Originally Posted by travis View Post
Thanks for the advice.
By the way, read your Larry Hite signature again

That's the game we play...
dog4 is offline  
Thanks! The following members like this post: travis
Old Apr 19, 2010, 8:34am   #702
 
travis's Avatar
Joined Mar 2003
Profit Factor vs Return on Account

travis started this thread I will now write a post about Profit Factor vs. Return on Account and reason on which one is best, and if we should use both and when and how and all implications.

I will use this excel sheet for some tests:
performance_according_to_PF_vs_ROA.xls


PROFIT FACTOR, IS IT ENOUGH?

1)
Ok. Let's first try with simple numbers. Let's try to demonstrate that two systems with equal profit factors are different in terms of ROA and this would mean that it makes sense to use also ROA, next to PF (or maybe ROA alone).

Let's say we have a system that fails once every 3 trades, so it goes like this:
1
1
-1
1
1
-1

very regularly, so regularly that its maximum drawdown is 1 and its Return On Account is the Profit of 10 divided by that maximum drawdown of 1, which returns 10.

This system will have a profit factor of 2, because it would go say to 30 trades, and make a gross profit of 20, and a gross loss of 10, so 20/10=2.

2)
In our second example, we hypothesize a system that also reaches a gross profit of 20 and a gross loss of 10, and therefore a profit factor of 2. But in a different way. If first loses ten times in a row:
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
And then it wins 20 times in a row. The profit factor is always 2, but we have a system that can make us so desperate that we will stop trading it. Yet, it seems as good as the other system, in terms of profit factor. How does its performance rate in terms of ROA? The profit is, as for the other system, 10 and we have to divide by the staggering maximum drawdown of 10, for a result of... 10/10=1.

So, from these two simplified but not impossible situations, we know that Profit Factor is not enough, and that Maximum Drawdown describes the whole picture, or at least a much larger picture.


RETURN ON ACCOUNT, IS IT ENOUGH?

Let's try to find examples where we can see that ROA is not enough either and that it can be completed with Profit Factor. If we didn't find such examples, it would mean that we can discard PF and just use ROA, which is what I would think actually. And that is why originally I only had ROA.

Let's use our second example again: a system that makes a profit of 10 and has a maximum drawdown of 10, which is quite typical. This returns a ROA value of 1. But how many more ROA values of 1 can we have for systems that are not as good or not as bad as this one? How blind can the ROA be?

Let's see... if we have a system that made a profit of 1 and had a max dd of 1, we'd still score ROA 1. But this would not be a problem , because if we used 10 contracts, the two systems would be the same. So it's not in terms of size that we have to analyze this, because in that sense ROA doesn't fail.

I can think of one example that might cause it to fail.

The first case is when a system makes 30 trades, as in our earlier example, and fails on 10 straight trades, succeeds on 20 straight trades, we will have that mentioned ROA of 1 (profit of 10 divided by dd of 10).

A second case is if a system goes down 10, goes up 20, and then goes up and down, one point at a time, for 200 trades. The final profit will always be 10, the max drawdown will always be 10, and the ROA will always be 1. Yet this second system would not be nearly as good as the first one, because it keeps your capital busy while producing no results. So ROA alone is not enough either.

But how could we measure this difference in performance? The average trade would produce less money, sure, but then we see that even in this case, Profit Factor would reflect the difference:

Gross Profit 20+100=120
Gross Loss 10+100=110

The Profit Factor would decrease from the original 2 to 120/110=1.09, thereby indicating that the system is screwing around most of the time.

Thefore, just as ROA integrated PF in assessing performance, now PF does the opposite. So the two indicators are quite good if used together.

So I am pretty much done, even though I am not capable of seeing all possible situations and implications of these two methods of assessing performance.

Let us recapitulate the 4 examples used above.

The 2 examples for profit factor show that PF fails at measuring how regular a performance is: if a system makes 20 and loses 10, it will be rated the same regardless of the distribution of losses. PF rates my systems in the same way it rates the rolling of a die. If they win 2 times out of 3, it would be like losing only when you get 5 or 6 on a roll. But I wish my systems lost that randomly. Losses instead tend to stick together, and that is what screws me, so PF is definitely not enough.

The 2 examples for ROA show that ROA fails at measuring how much a system screws around, which is instead measure by PF. I am sure ROA fails at many other things, and so does PF, but I can't see them right now.

Those two guys are definitely better than just focusing on absolute profit or even on profit per investment, which in the case of my systems is quite meaningless, since they trade futures and they trade them frequently enough to not be concerned with margin. Besides, margin generally mirrors ROA values. If a system causes me a max loss or a max dd of X, that X is usually closely related to what the margin required is. So I don't want to use an indicator which will measure twice something.
__________________
Read: E.P. Chan, Cogneau - Hubner, Sewell, Tverberg. Search: expected shortfall, Monte Carlo VaR, extreme value theory. Trade.

Last edited by travis; Apr 19, 2010 at 11:58am.
travis is offline  
Old Apr 19, 2010, 12:32pm   #703
 
travis's Avatar
Joined Mar 2003
10 time zones and their profitability

travis started this thread Ok, what was the purpose of my reasoning above? Because my systems are so bad and I have such little capital, that I was looking for a new way to improve them, and I think I have found and will use it starting tomorrow.

My systems' signals are divided into 10 timezones, 5 for the intraday signals of each day of the week, and 5 for the overnight signals.

These signals are not all as good, and I used to appraise their performance in terms of absolute profit per timezone.

But then came Weighbridge or I don't remember who, and he told me that it wasn't good, because systems might not trade very much on a given day of the week and still make a lot of money in relative terms, and viceversa.

So I started measuring this in terms of average profit per trade.

But this didn't matter much, because I wasn't still using them as a filter.

Today I've realized a need an extra filter, really badly, because I don't want to blow out another account. With this realization, I also thought I should really make sure that average profit per trade is a good assessment of performance.

Then I thought of using Profit Factor, because it's not as hard as ROA to compute, and here it is:
time_zones_performance_by_profit_factor.xls

The results are similar but not exactly the same as the average profit per trade.

Here's a description of similarities and discrepancies.

Tuesday OverNight signals are unprofitable and therefore have Profit Ratio of less than 1, and this is a similarity, as Tuesday ON also has an unprofitable average trade.

Monday sucks badly on both parameters, and Tuesday IntraDay is the best on both.

Thursday ID/ON vs Friday ID/ON time zones are almost the same for Profit Factor, but in terms of average profit per trade they're quite different.

On Wednesday, for average profit there's a huge difference between ID and ON, but not for Profit Factor.

Why are there these differences?

Well, average profit measures net profit divided by trades. And profit factor is gross profit over gross loss. But should I divide this value by the number of trades? It definitely would make a difference, yet I am not sure why I should do it.

I am kind of tired. I am at work, my brain is overheating, I didn't sleep enough. I am tired from all these sacrifices, years of work... I won't divide it by the trades.

I am basically measuring the Profit Factor of the time zone. I don't see why I should divide it by the number of trades. Profit Factor is already computed according to the regular formula, and it's therefore complete, so there's no need to divide it by the trades.

So what do we exclude? We will use profit factor for sure.

First I will subtract a 1 from all of them, since anything below 1 is useless. This will give me better proportions in their profitability. Indeed a PF of 1.5 seems close to a PF 1.5, but it's 5 times as good.

Ok, I am done. I have reattached the file linked above, and now i highlighted in BLUE and BOLD all the time zones I am planning on trading.

No, wait. I will not trade wednesday ON because it's like just thanks to the WeekDay Bias systems, which I don't trade anyway.

So basically I will just allow the ID systems on all weekdays except Mondays. It sucks but I have to do it, in order to protect my capital. Especially now. It really sucks to keep my systems from trading more than half of the time.

-----

[...]

I am home and I made some more calculations:
10_timezone_performance_by_3_parameters.xls

By all parameters, the first used, of absolute profit by timezone, the second of average profit per trade in each timezone, and the latest of profit factor by timezone, the top 5 timezones are always the same ones:

All intraday signals from TUE to FRI, and just one Overnight signal on Wednesday (mostly because of the WeekDay Bias systems).

In general ON systems suck, and Monday entirely sucks as well.

These trends are clearly defined after 1500 trades, with luck obviously cancelling each other out... the trend is quite clear. My ON systems suck and should be kept from trading except the best ones maybe.

All other systems can be allowed to trade, but not on Mondays.

So, I will now work on excel to create timezone filters for every system, allowing exceptions for the best ON systems which would otherwise almost never trade.
__________________
Read: E.P. Chan, Cogneau - Hubner, Sewell, Tverberg. Search: expected shortfall, Monte Carlo VaR, extreme value theory. Trade.

Last edited by travis; Apr 19, 2010 at 5:32pm.
travis is offline  
Old Apr 19, 2010, 6:42pm   #704
 
travis's Avatar
Joined Mar 2003
analogies for finding an edge in system trading

travis started this thread ok so I was thinking of this.

In trading we have... in automated trading we are trying to predict an event: going up or down. Considering our trades are pretty much long term, but we still have to account for commissions and spread costs, which counts in total for about 2 ticks, and considering we make about 100 dollars per trade, which on average is about 10 ticks, so it's like we're trying to predict an event which is already... you get my point an event as rare as 40%, to cover all costs.

So let's look for an event which is as rare as 40% and we have to come up with a system that bets on that event.

Calls from my friend Laura.

If she sends a text message I make my win, and if it's someone else I have a loss. Equal losses and wins.

On average she sends me text messages 40% out of the total amount of text messages.

The system will simply be based on the fact that she sends me more text messages in the weekends. I would back-test a hypothetical phone transcript and verify that every weekend more than 60% of all text messages come from Laura. Then I would proceed to invest based on this statistical study, because my wins would be enough to cover commissions and to produce a profit.

But then if one day Laura gets tired of me not replying, or Laura goes on vacation... all these things could happen, and that is why we need money management. We need to be able to keep on betting for a long time.

Anyway, this is just because I was trying to figure out why it is so hard to find a new trading system that works.

Because, yes initially you would think that predicting the direction of markets has already a 50% probability on your side, from the start. But that's not the case. Due to fixed costs, you are betting on predicting something rare, like rain (in Rome at least). You're not betting on something that happens half of the time, but only 40% of the time, and this is due to commissions and spread costs.

So it's like you can't bet on something almost certain like the sun rising. Actually that would be equivalent to betting that in 10 years the Dow Jones will rise, but we don't care about that kind of returns.

To have more returns, we have to build a system on shorter term events.

Not equivalent to the sun rising, nor equivalent to "it will rain within 12 months". It has to be an event that, all other things being equal, for the average knowledge we have (without studying statistical data), happens less than half of the time (due to costs, as I said more than once).

On top of this, the statistical data has to say that this prediction, on a rare event, has worked for the past 10 years.

That's what makes it hard.

Let's look at some... rather than "rare"... "improbable" events, which doesn't specify how improbable they are. "Slightly improbable".

Let's look for events in our daily lives for events that are slightly improbable and try to devise systems that make a prediction of them happening a profitable prediction, in other words, let's find situations in which a generally improbable event becomes probable.

This will provide us with a good analogy for what we have to do in system trading. It will help us clarify why it is so hard and how we can go about it, to make it easier.

Improbable events in my daily life:
the neighbours not slamming their door
my dad saying something nice and uplifting to me
...

So far no good because those events are more improbable than finding an edge in the markets.

right, the next text message being from laura!

Let's stop here, because I find this hard to do... and I want to finish the post.

So here's what I would think of, before betting that the next text message is from laura:
1) in the weekends she writes more, but even betting on it in the weekends would only bring my probability, after costs, to a 50% chance of being profitable...

Thinking... thinking...

Got it:

2) the fact of her not calling for a while would, in the short term... wait. If she didn't write any messages for a while, and then she writes one, asking me to go to dinner, then, once I reply, I can bet that the next message I'll get is from her rather than from someone else. And this system most likely would be profitable on all securities (people's text messages).

No matter whose messages you're betting on, this is an edge. If you get a message from someone and you immediately reply (except if it says "happy birthday" and similar), probability is on your side if, after replying, you bet that the next message will be from the same person. Unless of course you're extremely social and receive dozens of text messages per day.

We just found how an otherwise improbable event becomes probable in certain situations.

A similar edge in trading can be found with "volatility breakouts" or similar. If you have a security that's not going anywhere (no text messages from a person) and all of a sudden it starts going somewhere (you start exchanging messages), you bet that it will continue (you bet that the messages will continue).

A similar thing, but that doesn't have to do with a volatility breakout is simply a trend reversal. If you see price just going down and down, and all of a sudden it starts going up, chances might be on your side if you bet that it will continue for a while. But, interestingly, sometimes I found the contrary to be true. It depends. Anyway, I still have not found a system where your % of wins (with equal wins and losses) is 90%. I hear people talk about such systems, but I never got higher than 75% of wins, with the Opening Gap systems, but they trade very rarely unfortunately, so they're outperformed by systems with much lower edges.
__________________
Read: E.P. Chan, Cogneau - Hubner, Sewell, Tverberg. Search: expected shortfall, Monte Carlo VaR, extreme value theory. Trade.

Last edited by travis; Apr 19, 2010 at 7:04pm.
travis is offline  
Old Apr 19, 2010, 7:38pm   #705
 
travis's Avatar
Joined Mar 2003
bad memories about people I resent

travis started this thread I never got into a fight in my life. I remember I pushed a person once, because he was behind me and he tried to trip me. Not much more than that. I never punched anyone and was never punched.

This was not really a choice but the way I was brought up. If I was in control of myself, I would never attack anyone and avoid physical fights at all costs, unless of course I lost control, which happened as I said a couple of times. Another time I jumped a bully, but it was a ridiculous situation and luckily they separated us, or else I would have gotten hurt.

Anyway, what I am getting at is that I never got into fights but I have put up with a lot of humiliations, because once people realize that you're peaceful, some people, a small percentage will take advantage of that.

Now, these people are not doing anything smart, because overall their behavior is not convenient. Sooner or later they all get in trouble for abusing other people. It's not a convenient attitude to be a bully.

But I don't know how convenient it was for me to be a peaceful "show the other cheek" person.

Most people bluff and pretend they're willing to get into a fight any time, and this way they get a lot more respect than I do. Instead I either fight or I don't pretend that I would fight, and this lack of bluffing has caused me a lot of abuse.

Yeah, because the situations in which I am willing to fight are very rare, and even then, it would not be good to lose control and to be willing to fight until the end, because in that case you might really end up getting killed. So I am tough at the wrong times, and only when I lose control and stop reasoning.

I don't like walking with a threatening posture like most people do. I don't like this kind of bluffing.

Because you might meet, actually you do meet a majority of friendly people and it doesn't make sense to show them a hostile face or posture.

I don't know... I don't like this world, at least the people who have been around me most of my life. I liked some countries in Northern Europe, where people are generally polite.

Here in Italy, especially in Rome and the South, the norm is aggression and unfriendliness. But even in the States, it kind of sucks.

My type of people is the computer science people, the people in Boston for example, the vegetarian people, the scandinavian people, the intellectuals. But then in the US you have the "frat" brothers, who are college students but whom I have nothing in common with. Those confrontational *******s, always looking to be rude to someone.

Well, as I was saying, I remember once I was in college in Boston.

And I was with my friend, another philosopher like me, we were going to a "party", I hate this word.

So we went and until then and after then I never had a similar problem, but there was this one "brother" in front the house, who lived there, and he said to me "you are not getting in". It was many years ago, but the memory stuck with me all these years, and I still resent that guy very much. I think about him several times a year, and how he told me I couldn't get into his party, and what I should have replied. I replied nothing, because I don't bluff. Should I have pretended that I was going to get into a fight because he said that to me? No, because it wouldn't have been fair.

But probably he told me that, precisely because he could tell he would have gotten away with telling me that. One time, in Boston, one girl in the street approached me and asked me if I'd be her friend and take her out to lunch. It turned out she was stripper from miami. I asked her why she approached me in particular. And she said I looked like a nice guy.

So in the same way, this guy told me "you're not getting in", because i looked like a nice guy, and also because he didn't like me. Maybe he felt I was a "euro-trash", which is... like "the rich europeans, stealing our women"... or something similar. I think if he saw me with blond hair, freckles, a beer belly... a baseball cap... he would have let me in. I was too refined, too peaceful... i didn't look tough and rough enough... so he told me "you're not getting in".

This and other sad memories stuck with me. Because my mom taught me not to beat up other children, and it all began from there. And now I have a big resentment towards all these people who took advantage of it.

I have my file with a list of about 30 events of abuse I took in my life, which resurface in my mind from time to time, several times per week, at least lately.

Not that I want to have these people killed, but i feel I am not willing to take any more of this abuse. I would like to have a way... but the most abusive person, emotionally, has been my dad, who's threatened me verbally and humiliated me in several ways and situations, mostly uncalled for.

He's the person I have the most resentment against. My mom taught me to put up with abuse without defending myself, and my dad was the first person who took advantage of it, and on top of it, he got me used to the idea that I was not worthy of respect, got me used to be humbled (by him)... I have a lot of resentment for all this. He's on top of my "people I resent" list.

These people are basically all people who never played a fair game with me. People who don't play fair in life. They make fun of someone in a difficult situation, they attack someone in a weaker position, they don't help you when you need them, but they're expecting to not behave like them, know you won't, when you will be in the same situations. They rely on you playing by the rules, when they don't. All bullies do not play by the rules. The main rule is "don't to others what you wouldn't want them to do to you, regardless of the fact that you might be able to get away with it". Behave fairly to people. Don't slam your door, don't laugh at someone who's having a difficult time, don't gang up on someone, don't talk loudly, don't wake up people in the middle of the night, don't steal, don't ruin people's property... you get my point. I am upset about injustices against me. Verbal abuse mostly. I have a feeling that the biggest injustices tend to be done against the nicest people. The more you play by the rules, the more people who don't play by the rules will come to play a game with you. "Foul play" i think is called, at least in Italian it translates literally to that. The concept exists everywhere, but the percentage of people who play fairly changes from country to country.

Like that neighbour now... he just slammed his door, the one below me. And yet he came up to here complaining we were moving chairs, and I replied politely "I am sorry" we will try to stop, even though it was before midnight and it only happened once a year, and he said he was going to call the police. Now that ******* slammed his door. How would he react if I went to him and asked him not to slam his door because it bothers me? Not well. And that is why I don't go to him telling him that. But if he comes to me... you see? I am playing by the rules, and I think it is the most convenient thing to do, because I never get in trouble, whereas this guy risks getting in trouble continously by abusing others and by attacking them if they slightly bother him. So I am doing the most convenient thing, but I am left with a sad feeling of being abused. It would feel much better to just reply "**** you". But I don't, because it's my second nature by now to not say "**** you" to strangers, nor to anyone.

So I guess I've lived a peaceful but humbling life until now, and I am not happy with it, but maybe i'd be unhappier if I got punched or similar. Or if I went to jail for punching someone.

Anyway, my dad has always been a fault-finding control freak, and he criticized me for everything i did, treated me with contempt, tried to shape my behaviour according to his desires in every little detail, such as my haircut, my shaving habits, how i walked... everything. Never any encouragement, just criticism and sarcasm. No wonder i never felt any affection for him. The best I can say about him is that he is sick. His deficiencies outweigh his qualities, which could be intelligence, precision, efficiency. I don't think the fact of being an efficient person should entitle you to treat everyone else like ****.

I came back to add something important that I forgot to say. The most violent place i found in my life was the States, but maybe not because of the place itself but because of the age I lived there. It makes no sense to generalize. However, I would generalize and say that northern europe in general is more peaceful.

Anyway, the two violent things that happened, the two negative things I didn't like in terms of aggression in the states were these:

1) "you are not getting in" from that guy i mentioned at the frat house
2) this is the worst offence: a pepsi half a liter plastic bottle thrown at me from a running car, while the guy said "have a pepsi!". It hit me in the head and this memory resurfaces in my mind even more often than "you are not getting in".

(TO BE CONTINUED... in future posts)

__________________
Read: E.P. Chan, Cogneau - Hubner, Sewell, Tverberg. Search: expected shortfall, Monte Carlo VaR, extreme value theory. Trade.

Last edited by travis; Apr 19, 2010 at 9:41pm.
travis is offline  
Old Apr 19, 2010, 11:41pm   #706
 
travis's Avatar
Joined Mar 2003
natural selection (continuing from evolution posts)

travis started this thread Quite interesting and scary stuff at the same time. Science is scary. Astronomy and all that crap scares me more than anything else. It makes you raise your eyes to the sky... and see emptiness and all that... scary stuff.

http://www.youtube.com/results?searc...cumentary&aq=0

I am watching this one right now. I like it:



__________________
Read: E.P. Chan, Cogneau - Hubner, Sewell, Tverberg. Search: expected shortfall, Monte Carlo VaR, extreme value theory. Trade.

Last edited by travis; Apr 20, 2010 at 12:12am.
travis is offline  
Old Apr 20, 2010, 4:43pm   #707
 
travis's Avatar
Joined Mar 2003
Getting rid of Sugar as well... SB (So Bad)

travis started this thread Sugar, SB at ICE (aka NYBOT).

It only trades half of the time. Did you know that? Forget building a system on it. It's got all its weird expirations, and it only trades half of the time, it has lower volume than the others... forget about it.

So now, besides my 9 futures, I will only have to add systems on AUD, CAD, CHF, NG (Natural Gas).

But let us study in depth what SB's schedule is and why it had escaped me that it only traded half of the time.

https://www.theice.com/productguide/...arketId=860080
Quote:
NEW YORK
3:30a-2:00p
03:30-14:00
http://www.interactivebrokers.co.uk/...369115&lang=en

Damn, for some reason IB doesn't have the schedule, as it usually has.

Anyway, sugar only trades for 10 and a half hours per day, look even here on the IB chart:

snap1.jpg

On the chart above, you see 11 candles per day, of which one is only half an hour, the first one.

How on earth did I think to test and automated a system on a future that trades less than half a day? Simple. With all the different and complex schedules they have, different timezones, different exchanges... I thought I'd first buy the data and then find out what was the deal, expecting the schedule to be good enough, since also the volume was good enough. And yet I was wrong. I was wrong because these agricultural mother ****ers... are different in everything.

Out of all these ****ers I started with, 3 of them are no longer with us. Sugar is the latest to have departed. Soybeans was too expensive and he had to go. Same for Corn.

The only good data with good volumes with good hours is the three extra currencies, which are definitely good by all standards. And the Natural Gas, which has great volume and the advantages of:

1) not being correlated at all with OIL
2) having the same exchange, schedule and expiration as OIL
3) using the metric system, unlike the ZN mother ****er

Yeah, natural gas and CL are just good.

With these 13 futures I am done. If anything I will add systems to all markets, but not markets to the systems at least for a long time.

So we are going to be trading 13 futures markets: 6 currencies, 2 interest rates, 2 energies, 2 stock indexes, 1 metal.

Given that I have 40 systems on the existing 9 markets, I will have about 60 systems on 13 markets. This is a good deal. Now I can't do it though, because I am too busy with implementing a good money management on the existing systems. I will worry about the rest once I am done with these. Also I have been busy with testing the new server and it will be great progress if I can get it to run smoothly from now on.
__________________
Read: E.P. Chan, Cogneau - Hubner, Sewell, Tverberg. Search: expected shortfall, Monte Carlo VaR, extreme value theory. Trade.

Last edited by travis; Apr 20, 2010 at 4:53pm.
travis is offline  
Closed Thread

Thread Tools Search this Thread
Search this Thread:

Advanced Search

Similar Threads
Thread Thread Starter Forum Replies Last Post
say hello to my little journal gozinsky Trading Journals 13 Sep 27, 2011 9:37pm
my journal travis Trading Journals 2334 Jan 18, 2010 11:57am
The EA Journal! Victor90 Trading Journals 1 Oct 28, 2008 12:47am

LinkBacks (?)
LinkBack to this Thread: http://www.trade2win.com/boards/trading-journals/85510-my-journal-2-a.html
Posted By For Type Date
Science Forum - why Standard deviation is better than Avarage Deviation This thread Refback May 22, 2010 2:34am

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)