What's Your Opinion on EAs and How to Approach Them!

So guys as the title describes what are your views on algo trading with EA's and managing potential risk.

My Thoughts...

You can optimize an Expert Adviser all you want on past data but you will never get the same result in the future I have been playing with EAs for a while now on fx and testing different situations and settings.

I would not let an EA trade all the time in each market I personally think you should ideally create different set files for each market condition such as trending and ranging and be able to watch the market and apply the correct settings for the market condition.

When optimizing I think its also best to optimize 1 or 2 settings at a time and have a look at the optimization graph to see which ones perform best and if there is a tight group of parameters working well then work with that range and narrow it down then move onto the next settings then back to the first settings a few times I also think its a mistake to chase profit instead you should be aiming for the best profit factor, expected payoff and drawdown.

Passes compared to losses when optimizing your EA you should also be keeping an eye on the tests that don't pass and seeing what your max drawdown could have been or did it blow the account don't ignore this data after all data is all we have so make use of it.

For example, you have a system you think will work then run it you get 90 fails 10 passes and in the 10 passes your best was 5% gain but on the fails, the worst was 10% loss would you trade this! I really hope not for a start why just why when 90% of the time it failed and made double the loss than what the top 10% result would have given you. Stay away from these and look for a move evenly balanced fail to pass ratio and try aim for something that gains more on the passes than it loses on the fails.

You should always be managing your risk when trading EAs and not oversizing lots and keep them in proportion to the account size. And even if your expected drawdown is let's say 10% it could easily surpass this so it's a good idea to deploy some kind of equity and balance protection if it's not built into the EA your using already have it built-in and modified or apply a second ea to manage this process so your positions will be closed and EAs removed to prevent blowing your whole account if things take a turn for the worst.

Use good systems and be logical and be realistic with returns don't curve fit the ea or use any risky strategies like martingale this is a disaster waiting to happen.

I recently created a system that survived a 10 year backtest with little optimization and produced a nice profit gain after I optimized a few things it got a lot better but more important it passed more times than it failed in the optimization phase with potential profits exceeded the potential losses the drawdown was about 12% with a gain on account of 170% not much for 10 years right but this is 1 pair yes I could have got more but I want a stable ea, not a risky one and hey manage the risk right nice and slowly wins the race your not getting rich overnight and if you do you have taken some great risks and it could have gone the other way.

Slow and steady wins this game create some good systems and build out a portfolio with several systems and pairs depending on market conditions and spread the risk out always remember to keep your capital safe and manage the risk correctly.

Sorry for the long post guys went on a bit of a rant but this is my thoughts on robots and what I have learned so far I would love to hear some feedback on what you guys think I know we all have our opinions but this is how I think they should be approached we will never know what the future holds but one thing I do know is I'm not losing my whole account to a badly configured robot that's taking a huge risk for no reason before stepping in.

I think the main problem of all the EAs around is lack of mathematics and real research. What we have? Averaging against the trend? Stop losses? RRR? Martingale? Most of EAs even don't have dynamic lot sizes, tps and sls. Hope and pray. We don't have multiEA & multiAccount tests. Only 1 acc, multipair. And I'm even not talking about volume. Everyone trade without information about volume. Magic!

For example I was trading this way: split my account to 100 and run EA on two. Then one of them "stuck" just move money from any remaining account. And so on. The system was very complicated and it was very hard to test it. But it was extremely profitable. The problem is that it is very expensive to make it. I have a dream to hire 1-2 coders full time and have about a year or two. I have tons of ideas to try out. But if 2 coders is 5000/mo then I need at least 60000$/year. And who said it will take 1 year only, may be 2 or more. But since I don't have 200-300k$ for that first I have to earn them. But I can't because it is impossible to earn serious money with current approach... vicious circle

My Thoughts.

At first you need a strategy, which is working manually for you. Then you can search for a coder, who actually makes an EA for that.

Profit.


Problem is, that many coders are not able to code things like trendlines, support/resistance zones and all these things, which are basic stuff for almost every strategy. I use EAs for NZD bank meeting trading, London breakout trading e.g.


EA's also need to be flexible(no fixed lotsize, no fixed SL/TP). Martingale will drive you directly to the bankruptcy.


My thoughts.

Test well before using.

what you describe will never exist
hope dies last

This is the most correct answer. because if you already know how to trade whatever technical patterns it is you're trading. the EA has to be able to see that and the coder has to be pretty damn talented to pull it off. if not, stay trading without one

most people do it the other way around and program the indicators into the EA but don't program the indicators to quantify a basic technical pattern that repeats itself frequent enough in the market.

Well, it already exists in my folder

I am coding for MT4 for almost 6 years now and I finally decided to open up an account here and offer my thoughts and coding skills.


In most cases the real problem is not the coder(there are so many talented coders out there!), but the trader. It happened so damn often in the past that someone asked me to create an EA out of his strategy. "No problem" I told them. But as soon as I got into detail they often said:"Oh.. I do not have a rule for that" or "I can not define that for myself".

How should I then code it? If you want to code an EA, then you need a rule for EVERYTHING! No wishy-washy, but real facts, rules, numbers! That is what you need for coding.


Or another great example, why coding an existing, working strategy often fails:
Imagine you, as an European, are trading a breakout strategy in EURUSD/GBPUSD/NZDUSD/GBPNZD (many people actually do that) and you get your best friend to code it for you. You have rules for everything but the 99.9% backtest(which is useless in most cases, but that is another, very complicated topic...but for this example we expect, that the backtest shows the truth) fails. It is not profitable!
Why?
Super easy: You, as a human, need sleep. So you will mostly trade during the London session. The EA is now trading the asia session in these pairs too.
But the asian session is known for low volatility in forex pairs, which do not contain asian currencies(e.g. USDJPY is volatile, EURUSD isn't) and trading breakouts without volume is just garbage.
So obviously your EA won't be profitable, because at least 1/3 of the day your whole strategy sucks because of missing volume. This means, that you have to code your sleeping time inside your EA(in some cases, like this one). Some coders do not think about stuff like this but it is acutally super important!

Just a quick outlook at a small example to show you, that you realy have to think about EVERYTHING, as soon as you want to code an EA.

EA's are working, but people mostly aren't able to code/define it correctly


Would I descripe myself as a good coder? No, there are many out there, who are much better!
Would I descripe myself as experienced? Yes, absolutely!
Do I have a holy grail? No.

If I've ever had an strategy programmed, i always have them program start/stop times as a true/false option. try both and see what works with your setup

Ive used quite a few EAs with not much success if I'm honest, I've tried ones that were martingale and then ones that went in with a specific set of rules. I just think once a pair trends with no SL you're in trouble. It is a difficult one for me to do now because I manually trade as my primary trading activity now

Don't waste time on EA's and learn how to trade manually. You will never make a smarter EA then your brain. EA's might work for few weeks/months, then they have to be readjusted...And that's even if you make a decent one which 99.9% of people here wont. If you can trade manually that's something you can count on for the rest of your life. Letting go of fantasy that ill make a highly profitable EA is the best decision I made few years back, I still mess around with them as hobby but that's about it...there fun. This is of course just my point of view.

I have worked with a team of 8 traders to develop Algorithm's over a four year period where we tested well over 50 different trading strategy's. We also had access to professional coders who maintain a Algorithm only prop firm in our city so we had a good learning base from which to build profitable EA's.

I would say around 6-8 EA's proved profitable during this period but can I lay it all on the table and be honest.

Best EA's were price action setups with the best being inside bar breakouts.
The most popular indicator in EA's seems to be the Bollinger Bands
Best method on EA's is to have no more than two external functions (that's what the pro's do, very simple clean EA's)
The demo data is manipulated so forget useful testing data
Allow 500-1500 hours to develop each EA with testing.

I had 8 computers running 24/7 testing 28 pairs on 22 strategies - hope your rooms air conditioned.
Be prepared for arguments with your parnter - you will become obsessed in all things 0 and 1

END OUTCOME

1 Guy still traders EA's full time but in stocks now, not Forex / indices.
6 Guys quit EA trading and got obsessed with Crypo manual trading (don't get me started)
Me - Still run some EA's for a few dollars but they EMOTIONALLY DESTROYED me - Returned all my energies back to Scalping where I belong and have always been successful.

DON'T DO IT -

Played around with EAs, but not a single one of them worked long term.
The ones that did work initially were pyramiding and hedging EAs, and with those there's always a risk of ruin. In short they will always explode accounts - one day. That's still useless because you end up with wasted money AND time.
The best EA that I coded was a breakout EA. I just bracketed a timeframe range, triggered a trade on the breakout, and then optimised take profit, stop loss, and trailing SL on pair data. I still wouldn't use it, and that because fake breakouts occur frequently in trading, and detecting THAT is not very simple.

I can trade successfully, and even though I can code an EA, I could not get Metatrader to code the way I trade.... so? I gave up on EAs.

My thoughts on robots are:
I'm not even talking about the fact that it is quite difficult to create them. At least for me
EA can only use technical analysis when trading. I understand that most traders use technical analysis, but the benefits of understanding how to trade the news at least at a basic level can be helpful.
Also, an expert cannot make decisions in non-standard situations. It doesn't contribute to making a profit. And most often it is the cause of large losses.
Well, the ea has no intuition. And as we know, luck and intuition is 50% of the success of our trade.

I dunno. this guy seems to have some good ideas seeing is how hes a champion at this kind o thing.

Algo Trading With Kevin Davey - YouTube

I have worked on one of his EA’s and talked to him two years ago at a Algorithm traders conference. He is a very knowledgeable trader.

I did have one of his models running for six months (with a few changes) and it was a nice strategy.

I don’t run it anymore - for a reason which it’s best I comment no further.

Cheers

the dude said hes done over 500,000 backtest over a decade and a half.

yeah i doubt any of us have done that

Impossible to do 500,000 back tests in my opinion.

Each computer if they were testing tick data would only get through 2 pairs a day. On bar close of course could be through 50 a day but the results would be rubbish.
So I had 8 computers running 24/7 testing tick data - therefore in a year I achieved about 5000 back tests of good quality.
That would take 100 years to get to Kevin’s figures, by which time I would have shot myself in the head.

Hang on a minute, why didn’t the F...k I have then mining Bitcoin?

Would have made more money.

may be an opinion but we don't know HOW much data he used and how it was employed etc.

the guy is a champion algo trader so he must know HOW to do that. and he said around 567k actually.

He’s a World champion and two time runner up championship trader and a nice guy to boot.
He also develops some nice indicators, the best in my opinion is a breakout indicator.

Talking about trading competitions, the average winning return over a year is just over 100%. The odd year can be 300% return but normally very rare.
If you do the sums, we should just accept a fair target is only 5% return a month and a good return is 10%.
I know there are traders claiming here they are making 30% plus a month but that’s not attainable over a twelve month period.


so who’s entered comp’s?

I think there if someone fail to create good ea is not equal that eas are not working.
Next thing I'm thinking about... most fail because they are creating robots that are working like human. Just let's say lazy human.
But I think here is the main and the biggest mistake. Robot must do what human could not.
For example what is the purpose to run ea on 1h tf and higher? No purpose. You can trade all the 28 pairs manually, on 1h tf there is no hurry if you will enter your trade 5-10 min later. But running it on 28 pairs, hedging with futures, arbitrage if needed, catching huge market imperfectness - this is very interesting. No human can catch few seconds situation on 28 pairs. But robot can.