MA Cross Optimization EA (very cool)

I came up with idea about 10 minutes ago.... it makes too much sense....

Have you ever wondered what MA cross works best historically on a pair? NOW YOU CAN FIND OUT!

The EA attached incorporates the MT4 optimizer, and cycles through all permutations of MA crosses, and gives you a list of all profitable crosses in a range over a given period of time.

YOU decide the timeframe and pair, and the max and min ranges to test.

I must pat myself on the back for this one.

Instructions: Run w/ strategy tester.... select timeframe and pair.... In Expert Properties set ma1 to 1, and ma2 to 2 .... set "Step" to 1 for both... set "Stop" to a high number for ma2 (say 10,) and ma1 to ma2 - 1 (9 in this example)

Check Optimization, and click Start. Click the Optimization Results tab and wait for it to cycle through all permutations.... if the cross was profitable, it will appear on the chart...... when it finishes cycling through all permutations (this could take awhile,) click the "Profit" heading repeatedly until the profit is sorted in ascending order.... then look at the values for ma1 and ma2 for the first item in the list...

Now uncheck Optimization, and replace ma1 and ma2 with the numbers you just retrieved.... run the backtest, and then look at the Graph and Report.

This EA works on the closed bar of the timeframe you select.

EDIT) If your chart system truly has a quantifiable edge, there should be no reason why optimal historical parameters can't be found with this type of EA.

I just found the best cross for GBPUSD on the daily timeframe between 1/1/01 and today (checking between 1 and 10 day MAs).... it took like, 20 seconds! This is some powerful sh#t....

6 for ma1 and 10 for ma2.... good for about 6,000 pips.....

Here's the graph....


By the way, for those that are skeptical, click on Open Chart and you can verify that the trade was placed right after the cross and the close of the bar (I attached what you should see).... You can use this build REAL systems on the closed bar.... not those fake graphs the scammers try to sell you....

Now expand this idea into other arenas.... Price Action optimizations.... Breakout optimizations.... systems composed of combinations of TA indicators!

PS) Spreads are taken into consideration. Graph are raw, uncompounded performance with 0.1 lots per trade....

If you interested in having the parameters of your strategy optimized, PM me and we can talk about it. This EA should give you enough evidence of my work (see my About Me section in my profile for credentials.) I'm pretty busy w/ my career, but will try to respond if I think the idea has merit.

Just a few things that are doable:

1. Price Action
2. Breakouts (range hour optimization)
3. MA envelopes
4. Bollinger Band squeezes or breakouts
5. Triple/quadruple MA crosses
6. Combinations of several TA indicators
7. Divergences (these require A LOT of work but can be done)
8. No carry baskets

Not only parameters of the entry/exit criteria can be explored, but ranges of TPs and SLs too.... and avoiding certain months, trading particular sessions, avoiding certain days of the week, etc...

I think strategies that use an indicator can be done too, but I don't have hands on experience to say definitely.

this is a joke right?
no way has Tdion just discovered optimization of past data and believes its going to provide useful parameters for trading the future
either this is a joke or you are tdion

my past has come back to haunt me!

just because i believe the majority of TA based systems are random curve-fits, it doesn't mean that this can't be educational!

and.... if i keep an open mind, i might be convinced otherwise (the sample size would need to be HUGE though)

besides, if i can provide a service to people, it doesn't really matter what i believe, does it?

the educational aspect will be learning that optimisation is curve fitting of past data and provides you with trading parameters that would have been perfect for the past but entirely useless for the next trading day. Your perfect ma cross will underperform from now on I guarantee it. You of all people know why this is so. If it were not then the market is the opposite of random and number crunching machines far more powerful than metatrader would have shut the market down decades ago. Come back into the light tdion

and if it were a cross on the hourly chart over the past 8 years with 5,000 trades, would that be curve-fitting?

this thread is to open people's eyes to getting the best performance they can out of their current systems. it doesn't have to be limited to MA crosses....

i mean, if you have a system you currently trade, doesn't it make sense to find the best parameters? i threw the EA out there for people to noodle over...

8 years and 5000 trades - useless tomorrow, try it you'll see. I have 24 pages of code developed over quite a few years. I have optimised ma crosses and slopes incorporating pivot points, candle stick patterns and time of day and many other parameters. I have produced systems that have worked beautifully for years tested on 15 min and hourly data obtained by subscription. I had one system that provided the exact level to enter just prior to the london open with the perfect stop and profit levels which were then snapped to the perfectly optimised distance from the current days pivot points. I could even tell which days worked better than others and a bunch of other stats. Foolproof or so I thought. It performed nothing like it had in the past and lost. I ask again, dont you think someone with far more resources than you or I would have done this years ago if it were possible to discover ideal parameters from market behavior in the past?

there is no way to prove or disprove the an edge exists with TA alone, but a large sample size can give you statistical confidence. the reason 5,000 trades over 8 years is convincing is that it is:

 

1. enough of a population size to consider better than lucky
2. a wide variety of market conditions (sideways, trending, etc)

so if the results are linear over 8 years with a large population, it does say SOMETHING....


the bottom line is that there is probably an edge (cause/effect) that is invisible to us, but is there.

you would think so, If it worked for 8 years and 5000 trades you would think that warranted alot of confidence. why not just go live with everything you have got? what could possibly go wrong?

nothing is guaranteed.... but you knew that before you put your first dollar on the table....

ps) this is not the right thread to have a philosophical debate over TA's lack of predictability... please.

ok lets assume the next 8 years will be the same as the previous and produce the same results due to permanent hidden inefficiencies that your meta trader optimiser has exposed. What was the max drawdown from the 6/10 cross? what was the biggest losing trade? what is the average losing trade? these stats will tell us what the risk level is so we can determine position size and profitability of the system

exactly....

first determine your strategy for green pips, and then move on to money management (a different thread than this one)

Very neat work. Although I don't believe that price moves based on MA crossovers. Seems like you are contradicting what you've been saying all along.

M.A.C.Doug is 100% right.

genetic algorithyms create endless # of 'WOW!!!!' systems, perfect traders of yesterday. but all lose $$ in today. try if don't believe. you will learning hard way.

it is a fun game paint picture of past, but it is not real trading.

p.s. true profiting systems have no parameterization numbers ... only truth about markit, not numeraligy.

for example ...

here is one of my many excelent ea systems ...
NO params, NO optimizing, only pure be one w/ market always...
p.s. ... NOT Mr.Martingale (loser!)

Do not make me pull out my 3 billion dollar 6 month EA test.


Tdion, I understand--partially--your intentions. But here are my two cents..


You'll need to develop some sort of rolling optimization to make sure that the cross is appropriate for the market. I personally would use the previous 6 months to predict the next month. I don't expect to be accurate, but as long as I'm within one standard deviation, I can say I successfully predicted the market.

if i increasing risk, i can make same or more. but not smart for real trading.
most of mine = 0.01 lots. you?

This is with a 50 losses to broke MM. The size of the lot is as if trading 50% risk, but ultimately goes to 2% when the stoploss + 2times the spread is hit. It has the all-important block trading feature built in so that it can open multiple orders with the broker once it reaches the broker's upper limits.

With 0.01 lots fixed,it still goes to about 700,000 in 6 months. The equity curve however is inverted (the slope gradually decreases to 0).

When trading forward, do I expect my EA to do as well as the backtest? Of course not. But I do give myself 20% leeway in either direction as my "target"