Recently, I have been noticing a lot of talk about the viability of automated trading systems on the forums and I have been very keen to share my opinions on this subject as it is one of my many passions and main trading style for the past several years. As it was a rather long piece, I decided to write and share it on my own personal blog. Here is an excerpt from that post:
It has come to my attention recently that opinions on automated trading, especially in the retail Forex market, tend to consist of mixed opinions which usually results in heated discussions within the trading community. I think the main cause of these disagreements is due to the fact that everybody seems to have their own definition on what automated trading means. In today’s post, I will attempt to explain the situation as best as I can from my personal and professional perspective. At the end I will offer my own opinions on the matter.
I have been writing and testing ea’s for about 6 years now. Mostly just for fun. trying to make profitable systems and find the holy grail hahaha. I guess it is algorithmic trading
For real money trading i mostly trade long term with commodities. I also have them automated but they might be more along the lines of automated trading … although there is a logical operation for opening/closing the trades … But the automation is more for controlling risk. Commodities are boring and low risk(for me).
ea’s and automated trading can be a great tool! They take away the psychological issues, doubts, fears and risk(maybe). They can make you stick to your rules and make money. Depends on how you trade to whether an automated system will be of benefit to you.
I don’t think anyone could say automated trading is not viable? Maybe only if they have no system in place?
Thank you for sharing your insight and experience! I agree that automated trading, done correctly, can be an amazing tool; I am a 100% automated trader myself.
You may be one of the few individuals on Babypips that believe automated trading is a viable strategy. :54: There are many misconceptions in the industry regarding “EAs” and “trading robots” due to all the “get-rich-quick” products. And because they are seen bunched together with automated/algorithmic trading, many traders have lost the respect or confidence in automated trading as a whole.
Thank you and everybody else for the positive feedback!
To come up with a successful automated system is not something you should take it lightly. Devising the basic strategy for algorithm is relatively easy when compared to the forward testing and fine tuning once it’s up and running. No matter how successful you were with Strategy Tester results, you will run into a plenty of unforeseen problems in the real-time environment. Prepare to spend a lot of time going after the problems because the system will break. For each additional logic and fix, you will have to follow up with regression tests which will take you hours and hours as the test data grows. I spent 4 years to come to this far. Yet, I am not comfortable to risk more 10% of my account balance.
From the start, I wanted to build an MT4 system that can handle all market conditions with no blackout periods. A system that covers trending, ranging, and often suddenly spiking after the news. The most difficult part of it all has been when the market ranges. Most of my 25% losses come from that.
How are you trying to accomplish this? I would assume you are using a sequence of filters to determine the current market condition and then using a set of asymmetric entry/exit logic? Would love to hear your take!
I would love to understand what you wrote Clark:) What do you mean with a sequence of filters They have to filter the market behaviour? But a " a set of asymmetric entry/exit logic "? Or is it very simple
TP>SL
So I was assuming that a sequence of filters, using various measurements, would first have to determine what the current market condition is (assuming he has also categorized price action into pre-determined conditions). Each market condition would also be assigned different entry/exit conditions.
What I meant by asymmetric entry/exit logic means different rules for long and short orders.
BTW, I also replied to your comment on the blog, not sure if you saw it.
I still have no clue how to make an EA with your description or that I use something like that in an EA that I made and never saw it in others. Maybe that is why they never work I will take a look at the blog
Yeah, its a big ask to make an EA to work in all market conditions… Just as a human wouldnt trade all market conditions. (of course you could… but trading profitably?)
All EA’s use filters… That’s what determines the trading. and they almost always use different rules for long and short orders.
Maybe that is why they never work I will take a look at the blog
hahaha A non believer! Its not that they dont work… They work perfectly! It’s just the trading rules they are given are more (greatly more) then often very flawed!
Maybe you should turn it around and look at it the opposite way… If you are trading manually and think your system is good… Make it into an EA! An EA will follow your rules to the letter! Then you can see if your manual system has any merit.
*** Ok, and before you say, "my system cant be put into an EA… as “it has no rules” or “uses price action” or “I use the vibe” or “I just eyeball a chart” or “I trade using COT data” or “I trade news,etc” What you are really saying is either “My system has no rules” or “I dont understand the rules”. If your system has no rules, you are going to fail! If you dont understand the rules, thats’ fine but you can work out the rules if you try.
You should be able to write your system into a concise set of rules! and you can write those rules as code! EA’s are more then just automated bots… They will without any bias tell you where your system is flawed.
I am not a non believer I think automated trading works but only not my creations so far. With testing I am many times on the right site but not that it will work live.
I start programming EA before I knew how to trade. After that I tried to learn manual trading and slowly getting better. Programming PA is not so easy and yes I don’t have a good set of rules. But candles you can program S&R also and trend can be a SMA I think all ingredients for an PA-EA
Right now I am testing an EA with something called VSAX patterns but I am running at the limits of my computer.
What do you think about these test results? (based on MA cross)
I have extensive filters for entries, identical for both long and short. For the processing efficiencies, I have the pre-filters to eliminate unnecessary entry evaluations. For the entry evaluations, the qualifiers are sequenced. Once qualified, the post filters are applied for confirmations. For exceptions, especially for spikes, optimum movement heights over historical data are used for entries.
The exits follow completely different logics independent of the entry rules. Sometimes, they collide each other, and they must be handled exceptionally… A long story…
The trailing stops play integral part of my trading system. Many different variations of it depending on situations.
I have built this trading system from the ground up using a few basic (non-optimized) indicators, not influenced by anybody’s theories or readily available information on internet.
haha Yeah, PA is very hard to define. RSI or CCI, etc might give an indication of PA?
What do i think? ahhh… The graph is headed in the right direction
Optimising with MT4 you have to be careful. It only looks at the bid price, not the ask price. The problem with this is signals wont be accurate. and if you have a small take profit, it wont be accurate.
Spreads at low volume times will give very false results.
When optimising always accept the most average results. Not the most fantastic… Average results are more realistic of system performance.
Optimise with a higher spread then you expect… (Stress the system).
Try to use time filters. Even if you want it to trade 24hours, break the EA in to hourly trading.
And very important… Watch the trades in visual mode… If the trades look too random or unpredictable, then they probably are. Change settings to trade the way you want, even if it is not optimal.
Then demo forward test… Test with multiple currencies with different settings… Run lots of demos… Use myfxbook or similar to analyse results… also try on different broker platforms. The same system can give very different results with different accounts.
I treat it more like a game… or a challenge. I know ppl say you should not be in search of the holy grail of trading systems but I am still searching haha… Making and testing bots, you find out a lot about patterns. I think you have to find patterns that are predictable and repeat every day. I have made and tested 100’s of systems. Only a couple have i ever run on a live account. I really dont like to lose Of those, only 1 has performed long term with very high profits consistently and it works under the most stressful conditions and scalps for about 5 pips with about 5 trades per day.
Anyway… Just one observation with your graphs… Why the change in lot size? try not to use any lot multiplier as it very seldom works in your favor long term.
I used this EA to test MM to scale in a new position after the first position is trailed to BE The first position is the biggest and, second is smaller as well as the third and 4th. When this EA catch the trend it makes good profit. Opening is price crossing 3 different MA Problem is how to filter out ranging markets when the 3 MA almost on top of each other. This EA has a time filter, different for every pair.
I dont trust the test results, they look to good to be true. On some pairs it is only losing money and I can not find a setting to get it right. What market behaviour can do something like that? and can that happen tomorrow on the other pairs that tested good?
Once I tested a program with the spread set to current and I had a good result Then I tried it again in the weekend and it made a big loss I checked everything and all setting were the same and it took some time time to discover that it was caused by the spread. Testing with MT4 is not so reliable.
I look at the stats, sometimes I find errors because the EA is only making Long trades or Long and short trades are out of balance. Also when you zoom in on a trade on the 1Min chart you see trades that are completely outside the price range
Hi,
this usually is a sign for overfitting- when your system did good on backtest data, but completely fails if you change a parameter slightly or apply it to a different market. thats why my backtesting/optimisation strategy doesn’t go for the highest return of all parameter variations, but more something like “good enough” for this market and every other I want to apply the strategy to.
or to make it short: robustness wins over highest return
regards
denis.
Its not that they dont work… They work perfectly! It’s just the trading rules they are given are more (greatly more) then often very flawed!
