Random Entry & Perception

[BSD]

BSD, is your real name Tom Basso??

Haha no Sir !

It fails to capture the imagination for several reasons

1. It does not satisfy human desire to be right, the majority of people are not comfortable with 75 pct losing trades

2. It does not satisfy human desire for excitement. My system runs at 8-10 trades a month. I last traded EUR/USD two months ago, and am still short.

3. It does not fulfill the need to feel clever (selling highs/buying lows). It can never do this, as that would be counter-trend.

4. It requires programming knowledge, C++ is an advantage. It took me a few months to work it out.

5. You will ALWAYS have large, open positions.. some people like to be flat each evening.

For many reasons it doesn't appeal, and there are many times when it's "tough" money. But your emotions are all in the head, the only thing that really matters is positive expectancy (no pictures on a scorecard, as they say in golf).

Excellent post again.

Huge relief to see that at least some people are real traders here who get the success relevant factors !

Like I said, thing with the smoke and mirrors squad is pure compensation for lack of success and ego over a desire to make money, they try to try to appear clever by obfuscating a perfectly simple issue beyond all belief, by trying to imply that there is some secret holy grail behind net profitable trading which only they are privy to but of course never ever divulge, when obviously nothing of the sort exists.

Trading is nothing but a probability game, same principles drive that like running a casino or insurance company.

Not rocket science any of it !

OK good night for now, off to get my beauty sleep.

 

[bevok]

You spin me round . . . great accompaniment!

Good comments encapsulating mechanical trading. Richard Dennis' comment about being able to publish his trading system in the newspaper and it would still work encapsulates it all really. The psychological factor is one thing, most people just can't understand how it could work, let alone mentally hack it. The other is that most investors don't like it. They prefer nice steady 10 - 20% returns and run for the hills if there is a 5% drawdown in a month. They prefer shining lights like John and Carlos in "Fooled by Randomness" who will give them a silky smooth ride.

I'll just take my plain old garden variety trendfollowing system with its ups and downs - at least its worked over 30 years on about 60 markets so is less likely to bomb out when 'market conditions' change again.

 

[DionysusToast]

BSD - there are 2 very important things you do not understand.

1 - What curve fitting actually is
2 - The fact that in most cases, curve fitting is something done unintentionally

Until you understand these 2 basic facts, you will continue to argue against something you think has been said as opposed to what has actually been said.

 

[DionysusToast]

Ah no, I would not call anyone specious twice in a day

It was to DT.. the assertion was that a random system cannot work because if you gave it to 50,000 people and they were all profitable, where would the money come from.

You could apply this to any trader or system.. e.g. Soros/Buffet cannot make money consistently because if they could, they could teach their tricks to 50,000 people and.....

Actuall MR - there is a point. Systems work because they fit market conditions.

Systems will fail if all people use them. Market conditions will change because of the over use of the systems. HFT systems will eventually suffer the same fate.

If you are saying you have developed a system which does not adjust to market conditions, yet works independently of market conditions, then you have just solved perpetual motion. The market CAN stop any system from working if it changes enough.

The best you can hope for in systems development is to exploit something in the market until it can no longer be exploited.

The system in the book would probably consist of less that 20 lines of code. It has no adjustment for various market conditions and it has a tunable parameter (the ATR used) which would undoubtably work better in different markets with different settings.

If you have spent 6 months writing a system and it works - then that's great. I do not see how that relates in any way to the system in this book as it could not possibly take anyone 6 months to work on.

 

[DionysusToast]

OK - so let's attempt another explanation but in Engrish....

The authors are of the opinion that ALL systems will benefit from their money management technique. To prove this, they apply the money management system to a random entry system.

It is a fact, that all systems benefit from the relationship between the various components. If we change one of the components, we can expect the system to fail if the other components are not robust or complimentary.

It is also a fact that all systems require complimentary market conditions or rather, there are market conditions that will cause any system to fail. Of course, more complex systems can adjust to different market conditions but that is not what we have here.

I think we all know agree on the following:
In up trending markets you can buy highs, in downtrending markets you can buy lows.
In choppy markets you must sell highs, in choppy markets you must buy lows.

The Random System
In a trending market, this system will place trades which will fail, until they get on the right side of the market in which case they will ride the trend until it’s over. It has a 50% chance of being on the right side of the market at the initial trade and on the ‘reversal’ trades when the market changes from an uptrend to a downtrend.

In a choppy market, again, the system will be right 50% of the time when the market reverses.

In terms of various market conditions – the number of ATRs stop loss could be tuned to better fit the market conditions. The number ‘3’ is not a magic number, more likely it is the number that fit best during their backtests. Tuning this parameter is very difficult without hindsight.

An MA Crossover system
In a trending market, this system will place trades on the right side of the market. It has a 100% chance of being on the right side of the market when the market changes from an uptrend to a downtrend.

In a choppy market, the system will be wrong almost 100% of the time as the entry portion of the system is geared towards trending markets.

In terms of various market conditions, the MA length could be tuned to better fit the market. The longer MAs would keep you out of some choppy markets but give you a smaller bite of the trend. Tuning this parameter is very difficult without hindsight.

Combination of MA Crossover and the 3 ATR trailing stop
How would the 2 systems above work together?

The MA crossover will tolerate a pullback as long as the MA’s don’t crossover. In a trending market, the only thing the trailing stop will do is to get you out of the market BEFORE the MA’s cross over. In effect, in a trending market the trailing stop will actually be detrimental to this system.

In a choppy market, we would still be selling the lows and buying the highs and have as many losing trades. W

So – when using the trailing stop in this system. We can see that our winners would more than likely be cut short but we would not have any less losers because we don’t have randomness to give us better entries 50% of the time during the chop .

So – what the authors have tried to assert is :

A specific money management technique will make all systems make more money all of the time in all markets.

What they actually proved is :

A system can be designed where the combination of random entry, money management and prevailing market conditions can be show to yield positive results on past data.

Their money management method would fail with many non-random entry techniques.

 

[pssonice]

"this system will place trades which will fail"

what kind of random function do u propose?

 

[bevok]

I agree with a lot of what you've said, although I think you need to read that passage in context - the 'money management' was icing on the cake so that they could say 'we made money by entering randomly with this small enhancement'. The main thesis was that with a 'good' exit on its own a system could be profitable. You're right, that exit would never be profitable in all market conditions and Van Tharp consistently underplays the entry factor because Tom Basso, someone he looks up to, didn't go and test the opposite hypothesis of good entry / random exit.

I think we're getting to one of the nubs of systematic trading here. Every factor we introduce to a trading system potentially adds to the results, but fits the system more to the period it is based upon. In statistics these are called 'degrees of freedom' and the more we add the more our method is based on our data. We will never WANT a system that performs optimally for a short time period, however I believe it is possible to build systems that do perform in the long term. They need to have a minimum of degrees of freedom, be built on good principles (e.g. trade with the trend) and test well on a very large sample of data encompassing a large variety of markets and market conditions. Your example above is good - adding an ATR stop reduces the profits of the 2MA system, however it will also probably greatly improve the profit:drawdown ratio. Then again, it is another degree of freedom. These decisions are what mean that giving away trading systems doesn't work, and one thing that stops every mechanical trader from winning (aside from the more common mental aspects).

The trader has to make a decision - do I add another parameter, or not. Testing can help - looking at whether a setting on a parameter introduces a spike in performance, monte carlo simulation (not predictive but another way seeing impact ona large sample) and out of sample testing. The big thing to avoid is 'back and forth' testing - add another idea, test etc.

My preference is for very simple ideas. Note the plural, as well as diversifying instruments I believe you should diversify systems or parameters in one system. You'll find an intermediate term system will do well for a while then tail off a bit, but a longer term system will come into its own then. The great thing is the drawdowns are concurrent, but the end profits are cumulative.

Finally look at a system like Aberration for something based on good market principles, very simple. Introduced in 1993, its had some dreadful drawdowns over the years but in 2008 it made 134000 + in a medium sized portfolio (single contract basis). Now I personally couldn't trade that system - the drawdowns are just too large - but I bet in another ten years its still up and has still made some big profits in that time.

 

[meanreversion]

Actuall MR - there is a point. Systems work because they fit market conditions.

Systems will fail if all people use them. Market conditions will change because of the over use of the systems. HFT systems will eventually suffer the same fate.

If you are saying you have developed a system which does not adjust to market conditions, yet works independently of market conditions, then you have just solved perpetual motion. The market CAN stop any system from working if it changes enough.

The best you can hope for in systems development is to exploit something in the market until it can no longer be exploited.

The system in the book would probably consist of less that 20 lines of code. It has no adjustment for various market conditions and it has a tunable parameter (the ATR used) which would undoubtably work better in different markets with different settings.

If you have spent 6 months writing a system and it works - then that's great. I do not see how that relates in any way to the system in this book as it could not possibly take anyone 6 months to work on.

I think I was pretty clear about how I coded the random system "along the way" - it certainly does not take six months to write. You're not easily satisfied tho are you? If I write it quickly, I am "pooping" it out, but if I take six months from start to finish, the implication is that it is far too long..!

What I have attempted to do is keep the degrees of freedom to a minimum. I have six or seven parameters in the system which I can vary, plus I get to choose the markets as well. I think even the most hard-core discretionary trader will use three parameters (moving average crossover, amount to bet) plus he will choose markets, so I don't want replies about the evils of optimisation. I understand the difference between that and curve fitting, so I use the same parameters across all markets, even when it's sub-optimal on a backtested basis (which it always is). Simplicity is the key.

Two additional points - Bill Dunn re-optimises his parameters from time to time, to adjust to changing market conditions.

JW Henry has stated that his researchers looked back to the mid-19th century for evidence of trends, and were satisfied that they existed back then. If all markets cease trending in any meaningful way for the next few years, I'm in trouble. However, this is why most trend systems will look at commodities as well, where the trends are real and powerful (look at natural gas and sugar, for example).

If the "95 pct of retail traders fail" line is true, it implies that most of the discretionary styles are not that productive, yet trend trading has been around for decades. As I mentioned in a prior post, the biggest trend of all, real estate, made large sums for tens or hundreds of millions of people. Which kind of makes your "50,000 people" line of argument look redundant..

 

[DionysusToast]

Good grief MR - I think we agree !

Of course developing a system is possible and I would never imply that it isn't true.

The 50,000 people argument is just there to show that any system is at the mercy of the market. If a system is purported to work outside the market or in all markets unconditionally, then self-delusion is at play.

Of course, a system that adapts to different market conditions, either automatically or by being tweaked by an experienced owner is also not an issue.

The line in the book that says "It made money 100 percent of the time when a simple 1 percent risk money management system was added. That’s pretty impressive." is very misleading as it implies that 4 or 5 simple rules can beat the market in all conditions.

Hence the word 'perception' in the title of the thread.

Also of interest is that we also perceive each others posts quite differently to how they were intended. I find people have the same issue with email.

The oddity is that when it comes to the written work in books (and newspapers), we give those words some sort of exalted status as if being in print in a book is somehow validation that the authors words are infallable, despite the fact that what we read may not be what they wrote.

 

[meanreversion]

Oh, I just spotted the 95 pct thing in your tag line.. I knew I got it from somewhere lol

That is subconscious imprinting at its best. I now accept it as a fact, and it's cos I've seen it scores of times (from reading your posts).

The other guy (bevok?) makes a good point though.... we have not tested random EXIT. In his Turtle book, Curtis Faith surprises himself when his tests show that a time-based exit performed better than other systems, this is a semi-random exit in that it doesn't take into account any aspect of market behaviour when it gets out.

Maybe I'll code up a random exit system.. will be back to report in 6 months!!

 

[DionysusToast]

Hmmm - seems I could get up to some mischief with my subliminal tag lines. Perhaps "transfer $1000 dollars to IB account number...." would be good.

I use time exits, although it's not a hard exit per se but it is 'almost' a random time. Basically, if price does not move with me in a certain amount of time, I will close the trade. This is day trading and the certain amount of time could be 5 seconds or 30. The theory is that you will exit if price action does not prove you right, as opposed to waiting until the price action proves you wrong.

If you code random exit and random entry - just make sure you don't get out before you get in...

See you in half a year.

 

[BSD]

BSD - there are 2 very important things you do not understand.

1 - What curve fitting actually is
2 - The fact that in most cases, curve fitting is something done unintentionally

Until you understand these 2 basic facts, you will continue to argue against something you think has been said as opposed to what has actually been said.

BS buddy.

As it is, there are 2 major flaws in this study :

1 - The approach taken to the study is flawed
2 - The test itself is NOT proof of random entry, rather they admit to curve fitting as well as omitting key data

The system made money 80% of the time over the data set.

Then they changed the system

Then the system made money 100% of the time over the same data set.

This is curve fitting.

They did not CHANGE THE SYSTEM lol, they just added a 1% mm stop !

Anyone who calls that curve fitting is totally, entirely and utterly clueless, it's a normal part of optimizing performance which is TOTALLY valid as long as one does not OVER-optimize !

AND it still does not explain your slanderous accusation that they ADMITTED curve fitting or omitted key data, they did absolutely NOTHING of the kind, they merely explained how they backtested !

Don't you understand the word ADMIT in the ciontext you posted it ?

What part of ADMIT do you not understand ?

Your lie that they ADMITTED to fiddling data is nothing more than your dishonest attempt to smear them so that you can fit your facts to your fiction, which is all your doing in this pointless thread.

AND it still doesn't explain the MAIN logical fallacy behind this entire thread, why on earth someone on nothing more than a PERSONAL fact finding mission merely for themselves would con themselves.

That premise is nothing short of idiotic.

Besides, their findings were corroborated so many times that your going round in circles endlessly really shows this up as a typical thread your great mentor the expert was so fond of polluting this forum with.

"As Anthony suggested I run 100 tests of the random entries-trailing stops at 10 ATR from 1982 until September 2007 on a portfolio of 69 futures with $100 for commission and slippage and the system made money 100% of the time as you can see in the following figure."

http://www.tradingblox.com/forum/viewtopic.php?t=3637&postdays=0&postorder=asc&start=0


100 runs in 69 markets from 1987 till 2007 (the year he posted that), profitable 100% of the time.

Now, as per your usual drivel, WHAT PART OF THAT DO YOU NOT PRETEND TO COMPREHEND ?

Any idiot knows that trend following systems - which in essence this is, albeit with random entries - work the better the more diversified they are, the more markets they trade, just to enable them to catch those trends that will almost always be available for capture in some markets.

And oh, yeah, calling the addition of a 1% money management stop as per Tharp curve fitting really beggars belief, did I mention that, lol, every single thing I have traded in the course of my career has had a MAX stop of 1 %.

Jeez according to the complete non-logic and tbh plain lying of toast accusing a public CTA of curve fitting I gotta belong in that camp too what with my addition of 1% mm stops.

Like I said, this thread wins the prize, after Socrates and the experts threads, of most idiotic, irrelevant, misleading and nonsensical currently available here at T2W.

But someones gotta play garbage man to prevent newbies from falling for the crap of the smoke and mirrors squad.

All these guys ever have on offer is negative whingeing and whining about how simple stuff don't work yo bro, yet it works for oh so many as I keep proving that it's proof enough of the validity, all that floats their boat is to try and impress some gullible newbies into believing that the posing charlatans spin spinning you around have it cracked, if only you'll do your homework and solve their endless riddles when in truth they aren't even wearing any clothes, your majesty !

 

[meanreversion]

If you code random exit and random entry - just make sure you don't get out before you get in...

See you in half a year.

This is one issue with OHLC data that I'm not able to overcome, so I work round it by only allowing one trade per market per bar. I can never know from OHLC at what points DURING the bar those points occurred, so I don't apply a stop until the following bar. Otherwise, it's possible to get stopped out before the trade even happens - I'm sure you know what I'm talking about.

Goodness, BSD is on the warpath again. I think we need to call the UN to calm things down.

 

[BSD]

lol yeah haha.

Hey Mean I just lost a symbol in Ninja, do you know anything about that ?

FGBL (Eurex Bund) has gone awol, I was updating sthg and pressed a wrong button somewhere and now it's gone awol ...

Any idea where to get that back from ?

Never happened before lol.

 

[meanreversion]

Don't use Ninja sorry. Just spent a few minutes on the phone with Interactive this morning, asking about sugar. It doesn't have "regular" trading hours like most other contracts, it just has one session.

I have to say, I find the help desk at IB very useful, despite the many negative things I've read on the web. But that is also an issue for another thread.

 

[meanreversion]

They need a different title for >3,000 posts. "Iconic member"? "Guru"? Where's the incentive to keep posting otherwise.

 

[BSD]

Yeah thanks anyway I like Ninja but it's not 100% but then what is eh, some stuff there is pretty clunky, like updating new futures contracts, gotta do that manually, and then this happened lol.

 

[BSD]

They used to have stars besides the names of often repped members now that was a sight to behold.

 

[DionysusToast]

This is one issue with OHLC data that I'm not able to overcome, so I work round it by only allowing one trade per market per bar. I can never know from OHLC at what points DURING the bar those points occurred, so I don't apply a stop until the following bar. Otherwise, it's possible to get stopped out before the trade even happens - I'm sure you know what I'm talking about.

Goodness, BSD is on the warpath again. I think we need to call the UN to calm things down.

This is a problem with retail trade automation systems. It's the same with Tradestation - if you want to code a system that works on daily bars data (e.g. daily ATRs), you need to drop down a level to execute the trades.

Ideally, these systems would let you enter a ruleset based on daily data but would execute at a lower level. After all - they have all the tick data.

I had a guy send me a system that ran on Excel - I moved it to Tradestation for him and it worked terribly - for exactly the reason you outline. He was terribly upset. Mind you - not as upset as the guy who brought a system of eBay and got me to... errr... crack the password & look at it. It just exploited a weakness of Tradestation to place impossible trades and pump out a fabulous looking performance report. When I told him this, he was very, very, very upset with ME !

Since then, I tend not to help people with such things. It's not worth the ear bashing.

 

[meanreversion]

It's incredibly easy to mistakenly program impossible rules, we've all been there. That's why the only automated system you should use is one you have written yourself, and one that you understand inside out.