my journal 2

[Yamato]

Yes, the edge is safe, also because my posts are quite long so few will bother using my system. Besides, these systems only make 3 trades per year, so few people will want to trade them.

Regarding everything else you said, you are right on target.

 

[Yamato]

support and resistance system cannot be automated

Continuing from here:
http://www.trade2win.com/boards/trading-journals/85510-my-journal-2-a-206.html#post1424298

This is the last time I will work on this system because I have met an unsolvable problem: I cannot automate it due to a history request problem.

But first let me summarize the system:

The system is done. It works perfectly (LONG and SHORT) on all currencies except AUD and it works also on GC and CL.

If you add some tolerance (0.1% of price) to the support and resistance being exceeded or matched, then trades double and profit increases by 50%. So this system would produce about 5 trades per future traded per year, so that is good enough, bringing my trades to about 35 trades per year, which is 3 trades per month. Just perfect so far. And it has 80% of wins.

One last thing: this system cannot absolutely be traded in discretionary mode, because it would require constant monitoring and placing of limit orders. Automation would solve everything.

But, and here i get to the problem, automation cannot be implemented, or at least not on all futures, and it's only worth doing this if I'll trade it on all 7 futures.

First of all: support and resistance work best if you use the forex value rather than the futures. Also because the futures will give you a lot of problems with contracts expirations.

So it is out of the question to measure those values on the futures.

But then, whereas EUR and GBP are perfectly fine and I should be able to download the needed 3 months of history for them, JPY, CAD and CHF are reversed for forex (it's not USD/JPY but JPY/USD).

This could still be solved by monitoring the reversed forex values and automating the system for the forex while still executing it on the futures, but it would involve a lot of work, maybe too much right now. The main question is if TWS will easily let me download 3 months of history, every day, for 5 currencies.

On top of this, I need to find a way to set up on TWS a continuous contract for gold and oil (GC and CL), since GC and CL contracts actually last at the most 2 months, which is less than the history I need.

I might or I might not be able to solve these problems efficiently. Having said this, the back-tested system is perfectly profitable and worth trading (still have not checked the out-of-sample, but I am very confident: I am just saving it in case I need to make more changes).

You see, for one thing, here they say that IB does not offer "continuous contracts", so this would mean that I cannot implement my systems on GC and CL:
http://www.elitetrader.com/vb/showthread.php?s=&postid=2293155&highlight=tws+continuous#post2293155

...it seems that IB TWS does not offer continuous contracts. So it can show you any contract you want (including expired) but cannot link them...

---------

[...]
Written later

Solved it!

I have discarded CL which didn't work well enough.

I have tweaked some other settings, slightly.

I have decreased the time span during which I poll the lowest low (viceversa for short strategy) to just 1000 fifteen minute bars. I also decreased the Pre-period to just 400 bars. The total history I will need to download from TWS is 21 days, which is good enough because even on the worst future, GC, I can find decent history from TWS, which goes back long enough to detect the highest high and lowest low.

Everything is perfect to start trading it soon. I will wait anothe week to see if I come up with something better and then I will run the whole thing on the out-of-sample.

Overall, each of the six systems trades 6 times per year, which means I get about 36 trades per year from all systems, which is 3 trades per month. Very very good.

Considering each trade makes about 100 dollars on average, I can make 300 dollars per month, without any worries.

These systems have a profit factor above 5 on average, and a percentage of wins around 80%.

It's not much, but it's about 3000 dollars a year with a margin needed of about 3000. So it's like another one of these systems that return 100% which I can add to my portfolio.

Now I have a portfolio of 77 systems (provided these last six systems all pass the out-of-sample).

------------

I am not sure what this type of trading is called. I need a short name for my "system type" tags. These systems are based on bounces off support and resistance.

If anyone has any advice, let me know.

The categories I have right now are:

ON bounce
Opening Gap
overstretched
Range Breakout
Volat.Breakout
WeekDay Bias
WITH ID trend

Considering the categories I have, I might call this new one (unless anyone has better advice) the "ID SR bounce systems" or the "ID HHLL bounce", which stands for "IntraDay Support and Resistance bounce" and the "IntraDay Highest High and Lowest Low bounce".

 

[Yamato]

new systems on Natural Gas and possibly other futures

This is not normal:
http://futuresource.quote.com/quote...50&chartDensity=MEDIUM&userStudies=&x=25&y=16



A future that goes straight down for several weeks, non-stop, is not normal. A system can be built on this. This is similar to the bond futures. Currencies and stock indexes futures move quite differently.

Now that I am pretty much done with the Support and Resistance systems, I will focus on trying to exploit this peculiarity of NG, and also try on all my futures, as usual.

 

[Adamus]

Re: support and resistance system cannot be automated

Considering the categories I have, I might call this new one (unless anyone has better advice) the "ID SR bounce systems" or the "ID HHLL bounce", which stands for "IntraDay Support and Resistance bounce" and the "IntraDay Highest High and Lowest Low bounce".

I think it's called swing trading. See what you think of the definition at wikipedia

 

[dog4]

Re: new systems on Natural Gas and possibly other futures

This is not normal:
http://futuresource.quote.com/quote...50&chartDensity=MEDIUM&userStudies=&x=25&y=16

View attachment 102822

A future that goes straight down for several weeks, non-stop, is not normal. A system can be built on this. This is similar to the bond futures. Currencies and stock indexes futures move quite differently.

Now that I am pretty much done with the Support and Resistance systems, I will focus on trying to exploit this peculiarity of NG, and also try on all my futures, as usual.

I don't think there is more mystery to this than a trend

 

[Yamato]

Re: support and resistance system cannot be automated

I think it's called swing trading. See what you think of the definition at wikipedia

Thank you for the detailed advice, but I am positive that here the term has to include these concepts: bouncing and support/resistance. As you know, I am talking about a specific method of trading, which triggers a trade once the previous high or low gets exceeded or touched.

 

[Yamato]

Re: new systems on Natural Gas and possibly other futures

I don't think there is more mystery to this than a trend

I am pretty positive about this one, too, and I am pretty sure you will agree that the way currencies and stock indexes move is quite different from bond futures and natural gas. Much less zigzagging, and much more up up up up and down down down down. Whereas the other guys do this more or less: up up down up up down...

And this can be exploited. I will be able to build another system on this concept. I am pretty sure.

 

[Yamato]

sure enough

Sure enough this one was even easier than the previous system, and more profitable overall. Not only that: besides NG, it also works on 4 more similar futures (GBP, CL, GC, ZN), that behave "up up up up" like. I had always noticed that ZN, CL and even GC moved much in the up up up up / down down down down fashion. I had never noticed it on GBP, but if it works I'll use it.

In less than two weeks, I've just created another 11 systems (6 on S/R bounces, 5 on up up up up trends), so this brings my total to... 71 plus 11 = 82.

This was just the creation part. Now I'll need to ponder things for a few days, and then finally run all the out-of-sample tests, after which the process is finished and comes the automation part.

Now, you see... there's no way I could have ever traded manually all these edges, let alone remember what the edges were, and there's no way I would have ever been happy with any one of them. But if you put them together, you come up with an amazing product, even when, like in my case, only one third of the systems work well enough to be worth trading.

 

[Yamato]

taking a break

Maybe now I'll post some music, like I used to do in the past, to reward myself for all this work.

I need to take a break and let these systems rest, and wait a while, to see if I can think of any changes, before running the out-of-sample.

http://www.moviewatch.in/watch-403-Eternal-Sunshine-of-the-Spotless-Mind

What I did find out is what I'll call my two new series of systems.

They will both belong to the 13th wave of February 2011.

The present categories are:
ON bounce
Opening Gap
overstretched
Range Breakout
Volat.Breakout
WeekDay Bias
WITH ID trend

The 11 new systems (presumably, if none fails the out-of-sample) will be grouped as:
"WITH ON trend", because that's what they do
"S/R bounce", because that's what they do: they bounce on Support/Resistance.

 

[Yamato]

Since so many people are reading, I am going to ask questions about the first group of systems, the S/R bounce systems. The WITH ON trend systems are just perfect so they don't need any fixing nor observations.

If anyone has any suggestions, let me know.

Here's what the systems do, their qualities and their problems.

What they do:
the systems check the last month of history, and trigger trades when they see that a monthly low/high gets reached or exceeded after 5 PM CET. Then that trade stays in place until the evening. There's other minor details (such as the exact hours, and the fact that the high has to be in place for a few days before and after - swinghigh/swinglow concept) that I won't disclose in order to keep some secrecy, but the concept is simply what I said: when the monthly high or low get reached a trade gets triggered in the opposite direction and it lasts a few hours.



The reasoning behind the timing is this: after 5 PM CET the markets are tired and tend to reverse, so if anything happening at around that time, such as big move breaking resistance, it's not likely to last.

The rationale behind the days before and days after, besides that it works (without it, the system does not work), is that a high and a low, even if it is a monthly high/low, is not meaningful unless it's been sitting there for a few days after it happened and it was not preceded by another lower low or higher high just a few days earlier. Just how long should be the optimal period is still a mystery to me, both in terms of "whole sample", pre-sample, and post-sample, but the reason i kept the three samples short is that I cannot practically download more than 1 month of history (futures expire often).

Qualities:
1. A high % wins (with wins bigger than losses). high profit factor.
2.The system works across a basket of futures almost with the same exact parameters.

Problems:
1. I have the feeling that i am only exploiting a small part of the potential of this concept (the famous concept of "support and resistance").
2. The system only makes one trade every 3 months on average (one every 3 months per every system/future).

 

[Yamato]

drunken insomniac trader

Yeah, that's right, I am back.

Woke up at 5 AM, turned the systems on, drank some beers to try and go back asleep. Just like travis, I can't sleep at night. Waiting to retire to the island. Life will be beautiful when I won't have to get up at the same time every day, to go to work. All I'll have to do is turn the systems on.

Yesterday I came up with more SR bounce systems, on an intraday basis. If they work, they will bring my total to about 90 systems, after which I will rest for a while. Maybe two months.

I am tired. This new boss is giving me more work than expected.

Finished my second bottle of Heineken. Will try to go back to sleep now.

 

[Adamus]

Re: drunken insomniac trader

Finished my second bottle of Heineken. Will try to go back to sleep now.

Heineken? You're using the wrong medecine. That's only one step removed from dishwater. Try a bottle or two of Belgian beer - fruit beer or chocolate or whatever your taste says.

Regarding the number of trades from the monthly S/R bounce system. Did you try the weekly S/Rs? That would give you theoretically 4 times as many trades.

You say you're looking at the intra-day systems now - do you mean in terms of the time frame for the S/R levels?

I found one system on the NinjaTrader forum that used session pivot points - Asian pivots, London pivots, New York pivots. Maybe the S/R levels can also be used intra-day from different sessions.

 

[Yamato]

Thanks for the advice regarding the beer, but the beer you're talking about is not easy to find. Especially since I am not willing to go very far to get it.

Thanks for the advice regarding shortening the timeframe, but I did it already, and the least period it works with is one month.

Yes, today I will test S/R bouncing on an intraday level.

Your last advice is too complex for me. I have to keep things simple, as always.

 

[Yamato]

I tested the S/R bouncing system on an intraday level and it does not work well enough to be traded. There's something out there, some sort of an edge, but I can't exactly exploit it for now.

I mean, there's things that I've seen happening, such CL reversing right after touching support at around 19.30 CET. There's a tendency for CL to reverse at 19.30 CET and there's a tendency to do it even more when it touches S/R, but some this doesn't seem to work too well, at least for the years 2000 to 2005. So I can't use it, because it would be "illegal" to use the out-of-sample to optimize a system that I can't get to work on the in-sample.

Ok, the creation phase is over. Now I will only allow myself a few days to rest and fix a few details in those two strategies and, then I'll run the out-of-sample, and then I'll have to automate them, which is heaviest work (simple but heavy, vs creation, which requires a lot of thinking but little work).

 

[Yamato]

ng on bottom, but can't do anything about it

I wish I had the guts and the capital to just jump in, LONG, with 1 big contract. This is easy money. You could make 10k in just a few days. It really seems to have touched bottom. But I also know that, no matter how positive you are, you're never positive nonetheless that it will go where you think it will go, so you can't bet the farm even with this beautiful chart and opportunity:



So I'll be watching it as a spectator, with a mini NG contract.

But my advice to anyone willing to risk losing 4k is to go LONG one big contract and make 4k. The probabilities are pretty big in favor of a long trade.

 

[Yamato]

weekly bragging

Weekly boasting about my systems. Here's my track record and real money non-tampered automated trading (up to Thursday, because Friday is not in yet):



This week they haven't disappointed once again. What can I say... no more boasting needed.

Here's the forward-tested performance of my top 18 systems (on the real account only 7.25 out of 18 good systems are being traded):



Damn. Too bad I didn't know exactly what the best systems were at the start. But they behaved well in the past few months, since I've identified the good ones, so it's not like I am just fooling myself and selecting only the ones that worked (however, to a degree, that is always the case).

A similar struggle, to identify and trade the good systems, can be seen on the first chart, which didn't really take off until, one by one, I got rid of the four bad systems I had mistakenly selected (which caused us losses for about 6000 dollars). It took me until the end of 2010 to get rid of the bad ones, also because the investors weren't eager to just drop systems that quickly once I had selected them for trading, so they resisted my urges in both directions (adding and removing). This is good of course, because my nature is compulsive gambling, and otherwise I would not have stuck with my choices long enough to see if they were good or not.

Here's the scatter plot that clearly singles out my very good 18 systems (sharpe ratio on y, forward-tested trades on x):



It took me 8 months of work with the investors, who made me stick to my choices and realize that out of dozens of systems, only a few selected are really worth trading.

Until recently my selection process was more guided by wishful thinking than by actual performance. Now I just look at the systems that show on my selective scatter plot, so there can be very little wishful thinking involved, because if a system doesn't show on it, I don't even consider it for trading.

And to think that just six months ago I still hoped that 100% of my systems worked, one way or another (the famous "reversing" theory, which implicated reversing their signals depending on their previous weekly performance).

But here's what my 71 systems look like, if I traded each and every one of them:



Pretty disastrous. You see, once I was willing to accept my limits (that only 25% of my systems are excellent), I also realized how valuable was what I did have. And I turned a disastrous situation into an excellent one.

And here's the future systems that I want to expand to:



I used green circles for traded ones and red circles for new entries.The CL ones will all trade the mini contract (half leverage), because their leverage is too high otherwise. The NG has to go from the one-quarter mini contract to the full contract.

 

[Yamato]

for that last half hour of sleep...

Damn. Sleep is so important to me.

I've slept about 7 to 7 and a half hours, but for some reason I woke up at around 7.30, have missed my 8 hours of sleep, and since then I haven't been able to go back to sleep for the following reasons:

1) worries and anxiety (my fault)
2) traffic in the street below
3) bitch with two year old child who screams all the time and she talks to him like to an adult ("I told you to stop crying...!"), and he'll be a psycho and a serial killer one day.
4) other neighbour bitch who gets in and out of her house, every half an hour, and each time she slams her door, because she can't close it normally, because no one raised her properly. She then takes the elevator, which is next to my room, so I can't sleep because of that noise as well. I wish death upon her. May she die in an elevator plunge.

For that half an hour of sleep that I am missing my whole day will suck today.

The only way I could get it back and sleep is to get a bit drunk, but I can't each time open a bottle of wine. I'd need beer, which actually is even more expensive, but anyway, I just don't like to keep on opening bottles of wine when I need to sleep. Even though the financial and health damage is the same or worse with beer. It's impractical, because I only need half a bottle, and then I won't need the other half a bottle for a week, and I'll have to empty it in the sink.

The only way this problem will be solved for good and in a healthy way is to move to the island, where most likely there'll be close to no noise, except for the birds singing.

Bitch, bitch, bitches... damn. I am going to be upset all day long, because of that last half hour of sleep that I needed. And June is not helping out this time, even though she almost did it again:

But probably soon we will scale up, and make about 1000 per week, which would be enough to quit my job, and I am also making money on my own account... maybe soon I can move to the island again. Within a few months. First I need to reach 20k on my account, and we need to scale up on the other account. A question of a few months. With 20k on my account, I'll be making 4000 per month, of which 80% is mine. Then I can start withdrawing it and changing my life.

There she goes again, she just slammed her door again. Psychologically it is very disturbing to hear people slam doors. It's like hearing... the sound of violence. A car doesn't bother me, a plane doesn't bother me, even if they made more noise. But a person screaming or slamming the door is very disturbing to me. And this is exactly the two things I've been hearing all morning.

I always hate one enemy at a time and this is happening because Vito moved out of my room. I am now focusing all my hate on the neighbouring door-slamming bitch. When Vito was a daily annoyance, the bitch was in second place. Now she's become a huge problem because Vito is totally gone from my black list, since he's moved into another room, to bring pleasure to other misfortunate people.

When I'll have the money, the first two things I will buy are space and time. Free time and free space, to isolate myself from the people I don't like.

Such as a big aparment, or an apartment with a big field around it, and free time, so I don't have to work with people I don't like. My dream is first of all to stay away from people in general.

Why do I have to hear these two bitches all the time even when I am not deciding to meet them? Why do I have to hear the screaming baby even when he's not my baby? This is not ok.

 

[meanreversion]

Re: for that last half hour of sleep...

I hadn't looked at this thread before, but it's interesting.

With regards nat gas, it's one of the markets I trade.. by and large it's been not too bad over the last year, but my trend filter has always been on the short side. Most recent trade was to sell a 2 lot at 4.41, which came within a whisker of being stopped out, but is now looking a lot healthier.

There was an interesting article in the FT about nat gas the other day.. the reason it's so low right now (in the US) is due to all the shale producers aggressively hedging their output. When you find gas in shale, you tend to get a large portion of it upfront and quickly. These producers are hedging like crazy, driving down the futures. In Asia, the effective price for gas is around 4 times as much, based on some of the very long term deals signed recently.

Anyhow, back to systems.. do you look at correlation between monthly returns? Or perhaps Sharpe of combined system vs that of individuals (is that the graph you sometimes display). Apologies for these basic questions, I've not read thru all the thread.

 

[Yamato]

That is interesting what you say about natural gas, and thank you for the information you gave me and for your contribution to this journal. On the other hand, I must also add that I try to focus merely on technical analysis. And to me it looks like it will bounce. I completely ignore fundamental analysis, because I feel that I don't know enough to do it right (on any market), so I prefer not to do it all, and that's been my thinking ever since I started trading, in 1997.

I don't know exactly what you mean by "correlation between monthly returns", but maybe I can answer your question by telling you what these charts are.

They represent the individual systems as far as the scatter plot (see above post for details). They represent the best 18 as far as the equity curve. Actually, that was intuitive but I don't know what else to explain about them. Oh, and the weekly chart, as I said, represents a real account that I've been trading for the past few months.

 

[meanreversion]

I don't trade fundamentals either, but it's always interesting to see what drives the underlying. Part of the reason for the recent divergence between WTI and brent is the oversupply in the States in Kushing, Oklahoma, where new pipelines have come on stream and more oil than before is now arriving from Canada. But I digress.

My main system is a medium term trend model, alongside which I was also trading a shorter term breakout model in FX. My starting point for any system is that it should have positive expectancy, but on doing a bit more analysis of these two systems, the Sharpe & Sortino of the combined over the last ten years is lower than that of the better system.

In other words, by combining these systems, my risk adjusted returns might actually be worse. The correlation of monthly returns was >40%, which suggests they are too similar.

I've done a little further testing with combining a mean reversion system with the trend system, and that seems to be additive in terms of risk adjusted returns, so I'm looking to switch to that fairly soon.

That's just my take, I know others look at it differently.