The fallacy of Metatrader EA trading contests

Automated entries don't work because during a specific time period of say 3months a specific pattern emerged that just happened to generate good returns on this 1 out of 1000 EA's you are running at the same time. This is why there is always a few people making good returns during the Metatrader EA contest, out of hundreds of EAs , the probability due to random luck that a few of them will just so happen to make money during a specific trading pattern that won't repeat itself(months 1,2), is very high. Keep on running the same competition over a year period and losing EA during the first period could actually be the winning EA during months 11,12. During any short period of time and taking a large random hodge-podge of EAs, some of them based on random luck will generate good returns, but during another arbitrary period these same EAs will fail.

For example an EA could align its buy instructions with phases of the moon during the EA trading contest for that specific period, but keep on trading on the phases of the moon and over a 12 month period the account will be deep in the red.

The market follows the 60/30/10% rule: 60% counter trend, 30% trend and 10% breakout. If the market just so happens to be in a trend during the EA competition then an EA tweaked for trends will do well but during counter trends will fail. The trading contents doesn't prove anything because the period is to short.

It's called survivorship bias, it's hard to differentiate the lucky from the successful in the short term (or even the long term if you believe Taleb).

Be careful with commercial EA's... take a gander at this, and tell me if Forex Funnel still looks good.

http://www.forexpeacearmy.com/metatr...ex_funnel/demo

A Mathematician Plays the Stock Market, book by Paulos

The accounting scandals involving ?WorldCom, Enron, and others derived from the data being selected, spun, and filtered. A scam I first discussed in my book Innumeracy derives instead from the recipients of the data being selected, spun, and filtered. It goes like this. Someone claiming to be the publisher of a stock newsletter rents a mailbox in a fancy neighborhood, has expensive stationery made up, and sends out letters to 68 John Allen Paulos potential subscribers boasting of his sophisticated stockpicking software, financial acumen, and Wall Street connections. He writes also of his amazing track record, but notes that the recipients of his letters needn't take his word for it. Assume you are one of these recipients and for the next six weeks you receive correct predictions about a certain common stock index. Would you subscribe to the newsletter? What if you received ten consecutive correct predictions? Here's the scam. The newsletter publisher sends out 64,000 letters to potential subscribers. (Using email would save postage, but might appear to be a "spam scam" and hence be less credible.) To 32,000 of the recipients, he predicts the index in question will rise the following week and to the other 32,000, he predicts it will decline. No matter what happens to the index the next week, he will have made a correct prediction to 32,000 people. To 16,000 of them he sends another letter predicting a rise in the index for the following week, and to the other 16,000 he predicts a decline. Again, no matter what happens to the index the next week, he will have made correct predictions for two consecutive weeks to 16,000 people. To 8,000 of them he sends a third letter predicting a rise for the third week and to the other 8,000 he predicts a decline. Focusing at each stage on the people to whom he's made only correct predictions and winnowing out the rest, he iterates this procedure a few more times until there are 1,000 people left to whom he's made six straight correct "predictions." To these he sends a different sort of follow-up letter, pointing out his successes and saying that they can continue to receive these oracular pronouncements if they pay the $1,000 subscription price to the newsletter. If they all pay, that's a million dollars for someone who need know nothing about stock, indices, trends, or dividends. If this is done knowingly, it is illegal. But what if it's done unknowingly by earnest, confident, and ignorant newsletter publishers? (Compare the faithhealer who takes credit for any accidental improvements.)

Similar to the newsletter scam, but with a slightly different twist, is a story related to me by an acquaintance who described his father's business and its sad demise. He claimed that his father, years before, had run a large college-preparation service in a South American country whose identity I've forgotten. My friend's father advertised that he knew how to drastically improve applicants' chances of getting into the elite national university. Hinting at inside contacts and claiming knowledge of the various forms, deadlines, and procedures, he charged an exorbitant fee for his service, which he justified by offering a money-back guarantee to students who were not accepted. One day, the secret of his business model came to light. All the material that prospective students had sent him over the years was found unopened in a trash dump. Upon investigation it turned out that he had simply been collecting the students' money (or rather their parents' money) and doing nothing for it. The trick was that his fees were so high and his marketing so focused that only the children of affluent parents subscribed to his service, and almost all of them were admitted to the university anyway. He refunded the fees of those few who were not admitted. He was also sent to prison for his efforts. Are stock brokers in the same business as my acquaintance's father? Are stock analysts in the same business as the newsletter publisher? Not exactly, but there is scant evidence that they possess any unusual predictive powers. That's why I thought news stories in November 2002 recounting New York Attorney General Eliot Spitzer's criticism of Institutional Investor magazine's analyst awards were a tad superfluous. Spitzer noted that the stock-picking performances of

most of the winning analysts were, in fact, quite mediocre. Maybe Donald Trump will hold a press conference pointing out that the country's top gamblers don't do particularly well at roulette.

here, here!

The star stock analyst Elaine Garzarelli, who had predicted the Black Monday one week ahead of the crash has a poor performance in the years following the crash. This is the best example for survivorship bias.

The stock analysts has no difference than newsletter publishers. They get paid by selling you an article.

Hi,

Agree it's like an EA that may get great result in trending market like Euro fall recently then get big losses in summer low volume.

I hate using black boxes. First step is before even using any is understanding the inner & strategie used.

Regards
13

Nothing to do with volume but the 60/30/10% rule. 60% of the time a market is in a counter trend like we have now, then a trend, then a break-out. The euro's plunge was a historical event that won't repeat anytime soon.

There is no statistical significance to volume in trading, it should be ignored.

Here is another good book: "Fooled by Randomness"
http://www.amazon.com/Fooled-Randomn.../dp/1587990717

I would beg to differ though one can actually win the markets, the trick is in proving it and not just asserting it can be done. My track record at
http://stocktwits.com/stephanusR is 28 winning trades out of 30, documented , time logged. My timezone is Cyprys time. This is far better than a real trading statement because one could have hundreds of real statements and only tell us about those that actually worked.

Yeah, some EAs can have runs of wild profitability, but excellent all-terrain strategies are not the types that usually win trading contests.

I think if you have a good EA and you know it works, you should keep it for yourself other wise eventually it will stop working. If you had a secret formula to win the lottery and you told the hall world, then nobody wins, do the math. Also you don't see Bill Gate or Steve Job or Google giving away their secrets, that’s why they are successful

Hi everyone,

There is my opinion...the most important to update your EA with market and have high class programmer with master degree...I run private funds with my own EA and have no problem with that...yes, I follow with market and updating all the time...and I agree when you have a good EA you will never sell it even for big bucks...RG

don't know my success secret? You didn't read a lot, many books and magazines have revealed it

I still think that gibing away your secret formula for wining in any business or sport or anything is not a good idea. It is OK to help, witch I will do when I become successful, but I will never give away my secret formula. Are you the real Bill Gate?. Because if you are, I have to say that you are one of my great heroes, together with Thomas Edison, Albert Einstein, Newton and of course Warren just to name a few. I admire smart people, I'm not as smart but one day you will know my name.

The key it to understand that 100% returns and becoming millionaire in a day is not possible. 30+ return a year is really good. After I got that developing EAs became easier.

Hand on my heart, I cannot understand or fathom, how come, clever, highly intelligent and studied people would dedicate hours of time, capital and hardship to waste on Retail Algos? In my most humble of humble opinion, for the Retail industry the real edge it has is discretionary manual trading, with skill, patience, endurance, money-management and good capitalization.

Algos (I would like to use the term 'Algorithmic trading'* as a generic term here) are the realm of locals (MM on Exchanges), Price making banks and hedge funds.

(just a recent example I saw here on FF, someone stating clearly that he had coded his manual strategy to rather trade via an EA, turned out that it could not generate the desired results - the only way to trade that pure PA-driven strategy successfully was to be done via manual trading - so back to mouse clicking it was!)


* just thought I'll add some additional info about Algorithmic trading:

Algo trading is a generic term that describes a style of trading where algorithms are used to generate the orders. It means there is a computer model rather than human interaction in terms of artificial intelligence/sophisticated matrix (aka ''EA's'') that will generate trading signals.

People talk about 'Algo' trading but there is a lot of crossover between the use of the term Algo Trader and the use of High Frequency Trader. By definition, a high frequency trader is an algorithmic trader but not the reverse.

But also, you could have a system that is algorithmic that generates signals without necessarily generating a trade and it doesn’t have to be a high frequency trade or a high frequency signal. So you could have an algorithm that generates the signals, you can have a separate algorithm that would actually execute the orders and send them into the market. (ie. VWAP algos, Time slice algos etc. etc.)

Further,

Looking at exchanges, they have a process where they queue your order on the time angle, but also, have a 'pro-rata share-out' where it depends on the size of the placed bid/offer compared to others.

With regards to 'flash trading', flash trading is also a form of high frequency trading. But the controversy about this particular type of trading is where the exchanges created a club of favored high frequency traders so to speak, who were getting to see orders in a very short period of time – but still a period of time – before the rest of the market would get to see them. And because high frequency traders have algorithms that can make a thousand different decisions in a split second, by giving them even a fraction of a second to see an order before the rest of the market could see them, it gives them an unfair advantage. (Thus 'flash trading' got great attention lately in the media, though imo, it is no different then in the olden days when locals standing in the pit had also this 'unfair' advantage, the difference being to today, that, these days we are talking about 'co-location', but in reality, it's the same principal: some people are getting valuable information before others!).

An interesting topic of discussion.

I'm currently building several EAs, using iExpertAdvisor's Virtual Traders Studio - the latest version. But, I did not just start building trading systems yesterday afternoon. I've developed a proprietary trading system several years ago and that system involves many different indicators. It was a system that signaled one trade per day and it was based exclusively on daily, weekly and monthly data.

Now, I'm planning on taking advantage of bars that are smaller than one day, which I was not able to do before. I spent years developing the knowledge of how to create trade logic, as opposed to using that time learning how to code a programming language. The result is that I lack the programming skills, but have mastered the trade logic skills.

Many people have done the opposite. They have taken the time to master their coding ability, while they lack the ability to code trade logic that works in every market phase (bullish, bearish and channeling). Between MT3 and MT5, many have developed the ability to write very eloquent MetaTrader Indicators, but have not yet developed the same level of acuity when it comes to writing trading systems that endure.

Just look at the language we use when talking about Traders:

Gee. That Mr. Tibs, is the best MQL coder I know.
Gee. That Mr. Paul, is the best MQL developer I've ever seen.
Well, Mr. Hotshot, is probably the most widely respected MQL programmer on this forum.
Maybe, you should as Mr. CodeAlot, she has the best MQL skills by far around here.
--------------------

But, how often do we here this:

That Mr. Capable, is the best trade logic developer I've seen around here.
Hey, check with Mrs. Profit, I hear she can write trade logic better than anyone.
The best trade logic engineer that I've seen was Ms. ChecknTheMail.
That trade logic designer called Mr. PivotOnTime, is the best there is
----------------------


We never talk about what it takes to Write Good Trade Logic. We somehow assume that if one is capable of Writing Good Code, then those same skills must be transferable to Writing Good Trade Logic. The reality is that nothing could be further from the truth. Being a code software developer, script writer and trading platform coder, does not necessarily translate into being a good Trading Systems Designer. The two types of logic are not one and the same - not even close. They are predicated on solving completely different problems, in completely different environments.

So, when we say that an EA failed, we must be very careful. An EA, is nothing more than someone's Trade Logic. That individual might have glorious computer programming skills, but be the lousiest trading systems engineer on planet earth. So, we cannot simply conclude that no EA can work across multiple market conditions. There are some excellent traders out there, who will never encode their systems - so, you will never see those equity curve results.

I'm developing smaller trading systems right now, that will be based on some core components of a much larger trading system that was designed in something other than MT4. I'm using Visual IDE MQL code generator because I lack computer programming skills sufficient to convert my system. As I finish the systems, I'll post a link to the site that you can visit to see some of the equity curve results, as well as some screen shots of the IDE used to build the Expert Advisors.

Writing Trade Logic <> Writing Computer Programming Code. Never the twain shall they meet. Find somebody good at doing both at the same time, and you've found yourself a real Special Forces Trader. A serious Jedi Knight Trader. A Master of the Realm Trader.

So, when you are out there reviewing these EAs, ask yourself, "Was this person an MQL programmer, or were they a Trade Logic engineer. The two are definitely not the same.

Regards,
iDouble

I need to correct my post.

I am no longer using iExpertAdvisor's Visual Trading Studio, because the company refuses to offer genuine product (technical) support, and the platform is riddled with too many bugs.

I now use Molanis Strategy Builder, and have found that it runs smoothly, does what the company says it will do, and equally as important, I get timely responses to tech support questions without having to beg for help online.

Anyway, I just wanted to clear that up and correct my earlier post.

When you produce a product for traders, it needs to work and it needs to be supported. It really is just that simple.

Agreed for the most part, but there are ways to still make price action strategies work in ATS. I've not come across one discretionary system that uses diagonal line studies (or any other discretionary method for that matter) that works when automated, but horizontal S/R, breakout, trend following, reversal based systems do work. You can use diagonal trend breakouts, trend following or trend reversal methods if done correctly. Don't use any indicators at all!

Automated Systems
They are as good as the designer of the system made them, they have to be made to NOT be discretionary, but mechanical or they will always fail. Code can only do what it is programmed to do or if AI NN then can do what it is programmed to learn, this can be price action, but it will need quite a bit of data for it to not be over fitted to one segment of a market condition. Further optimising as time progresses is necessary as the markets dynamics change. It has to have good RR ratio and has to be a decent win rate, but not curve fitted, as I said.

It is good to have dynamic entries and exits but they must be mechanical when it comes to the rules, otherwise the automated strategy will fail. I know this seems like a paradox in itself, but look closely and you will see it is not.

MT trading contests are for fun, not usually much more, unless you mean more professional ones, then it is not MT at all as a rule. Out of all the Metaquotes champions I've spoken to over the years, none actually traded their winning systems on live accounts.