Is it really impossible to write a safe EA?

I don't want to trade manually anymore, so I was thinking to code again my strategy. Simply it can't or even is to hard to try it. So I would run an Expert Advisor from a VPS and while I sleep: let the money come to me


Probably the first trading system, is the Cross EMA, who everybody have heard and there are derivations of those.

My main problem it is the Market is switching between Trending and Ranging and Flat.

So if I take any system it is optimized for one of the type: or for Trending or for Ranging.

Let be a TradeSytem1 which is good for Trending.
You can choose a Currency pair and a Time frame and you can optimize to make super high profit. For that known of data
set!
As soon if you change the time frame you need to change the parameters!
As soon if you change the currency pair there s a high chance the system will loose your money.

Low time frames : M1, M5 has many Flat portions + the spread becomes an important distance to SL.
Higher time frame: weekly, monthly to few trades.
Middle: H1, H4, D1. They are ranging and are trending and the different broker timezone offsets shows different daily, maybe H4 candles to. The H1 is the same only.

I can't really distinguish when is ranging and when is trending grammatically.
I have combined an SMA -applied to Low and another to High and the period is 30 -so it is 1 Month or 1.3 Month average Low/High
and I have added a Stochastic, which is good for ranging.
Sadly one of the indications almost every time shows fake signals.



Can be adjusted they parameters to find the perfect values for a know period of the time, but other period needs other optimal values.

How I could write an EA, which it isn't optimized for a period of time, but it is at a currency pair to a timeframe?

The MT4 has to few data, only from 2015 I can optimize my EA. Is it possible somehow to download 1-10GB tick data from somewhere and have it from 2010 and in this case I could have a lot larger period? - and make EA to survive over that time.

I don't know how is written exactly the MT4, maybe he wants to load all content of the file onto memory, and for that reason the number of candles are limited, or maybe want to keep for the chart all the candles and don't swap out of memory, hence limited again.

The only way ea it's gonna be safe if you monitor market. While the ea is running you will need to watch the price and you''ll have to judge if in this condition your ea will be profitable or not. If not , you probably what to make an adjustment to your ea. So it's not very different from manual trading, except you will be able to trade more currency pairs faster. At the end forex is all about finding the best setups. When you find one trad it. Believe me, from 12 pairs you will find one. It like when you are in the store and store has 10 different types and chocolates. Yo don't buy them all. You buy the one you like the most.

How could I do programatically?

Lets say I will monitor the market by checking which currency pair is moving.
Oanda has a good tool, but can't set to lower time frame, anyway here is with higher time frame: monthly to weekly/daily

1 month:


1 week / current month:


last few days:



It gives a reason to open the Aud/JPY chart and search for Long entry.


The reason why I didn't traded manually this currency is because "I don't know" the JPY.
I live in EU, and JPY is moving while I sleep, so for me it was a bit safer to scalp the trash EUR/USD in 1m, 5m chart in direction of 4h, while I watch the market with his news.

I have another problem too:
The EA can be tested with only 1 currency pair, not with those 8-18 main.
And exactly that I want t test with more and with tick data:
I would look to closing candle of daily-4h-1h and do my math, analyze for trend whatever.
But if is an unclosed candle at m5 and is already 100 pip, then wait for close the 1h candle or the daily and do not trade until that? Manually it is not how I am doing.

static string pairs[] = {"AUDCAD", "AUDCHF", "EURAUD", "GBPAUD", "AUDJPY", "AUDNZD", "AUDNZD",
"CADCHF", "EURCAD", "GBPCAD", "CADJPY", "NZDCAD", "USDCAD",
"EURCHF", "GBPCHF", "CHFJPY", "NZDCHF", "USDCHF",
"EURGBP", "EURJPY", "EURNZD", "EURUSD",
"GBPJPY", "GBPNZD", "GBPUSD",
"NZDJPY", "USDJPY",
"NZDUSD"
};

It is 28 pair to monitor
I am wondering how much data trafic has this, if I start monitor in 1min or ticks..

I have started to make an indicator like that.
The first thing on what I am interested it is the spread, because maybe it is moving but has a huge spread.
Some brokers widen they spread at news time and then is moving, others claim they are ECN and they say the market's conditions are this.
Now it is weekend:


Of course I need the spread in my deposit currency not on whatever else currency.
It is made for 1 Lot size, maybe I need to make it custom...

Then various sorting.

The MT4 programming API has no key-value pair.
I needed to write a special sort function which takes 2 array instead of 1 to sort and in this way are done the sorting.
So primitive. It is like coding in '80 ages

I wrote an indicator, which simulates the trading: sell and buy signals and exit trade

The result is good : 100 trades and 20-30 pip profit in 1 year, it is just the spread, at least not negative

I wrote the 15 min implementation of my idea:



And what a surprise:

It is 20 pip / trade average and he did 44 trades!

Hi, any update?

Not much.

I have found a different version of http://www.forexfactory.com/showthread.php?p=8491204 (Quantum London Trading Custom EA)may work in ranging market, but soon as you got a news, trending something you can wipe your account. So or find a pair which do nothing - for a period - and trade it, because at that period it isn't traded my big money movers.
At Tokyo session the GBP/USD may be an option

Or find a trend following system ( something related to cross ema) and use it when ranging or you believe it will (ex CHOROS System , but can't find they clear rules, if they has it at all ).

Something related to this heat map / currency strength is here: http://www.forexfactory.com/showthread.php?t=446654 (Basket Trading EA) but the author want to keep the code for himself and I don't want to risk my money on a back box, not even in demo account(not worth the time).

I have found this:
Instead of starting the EA to code make an indicator, which has all the logic for entry exit signals, which are time or price or volume based ( do not bother with opened positions simulations for now)

Play the conditions of signaling entry and exit while you like what you see.

Only after that rewrite logic of the indicator into the EA and start optimizing if you want.

If the indicator is fairly complex calling the iCustom from EA maybe it will be able to run , but it will be so slow as it won't be able to do optimization with "every tick" signals

If you think you have found the holly rail make an indicator, which will try to emulate much as possible the EA.
Try to emulate the -2% equity SL - it is worth the effort:

Maybe just had a feeling the buy the low and sell the high it works or grid, martingale works, but just try to calculate subtracting from points/pips the spread+commission+swap.

Every system has a weakness, which can blow any account.
Do not start coding EA until haven't found that one and a protection against that!

1000 usd account, minimal 0.01 trade size, the spread+commission usually it is 1 pip ( 10 points) at least.
Having -2000 points it is -2%
Having -10k points it is -10% from Equity.

I think is worth the effort to make a detailed indicator, because it is a lot easier to filter out the unwanted cases...and fall into trap of over fitting or just cancel the implementation

I have the statistical exceptions from behavior.
And a kind of "normalized" bars too, with HH-LL "band"
For sure can be written a few systems.