MA Cross Optimization EA (very cool) | Page 4 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

New very cool CandleTime and session time Clock indicators 189 replies

Rsi cross indicator? (NOT cross 50, etc) 44 replies

very cool Trading Panel EA needs some adds 19 replies

Royal Cross/Fresh Cross 7 replies

World Market Hours - Very Cool 8 replies

Attachments: MA Cross Optimization EA (very cool)
Exit Attachments

MA Cross Optimization EA (very cool)

  • #61
  • Edited 11:04pm Jul 16, 2008 10:29pm | Edited 11:04pm
okay so 25,000 pips over 30 years might not seem like much.... hell, not even 1000 pips a year. but if you can find 3 or 4 systems as consistent and profitable as this one, then you have enough components to create a full trading system.

just thought i'd express where this is all going.... not leaving it up to the imagination.

suppose you combine this strategy with the one found here:
http://www.forexfactory.com/showthread.php?t=95895

between the two strategies, you would have made 28,000 pips over the last 19 years.... which is a pretty good start to a system. unfortunately, both are heavily dependent on dollar volatility, but its a start.
  • #62
  • Jul 17, 2008 12:38am Jul 17, 2008 12:38am
HOW CAN I GET EA'S THAT WORK VERY WELL THAT ARE 100 PERCENT GOOD PLZ SEND TO ME

Quoting tdion
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.
  • #63
  • Jul 17, 2008 12:41am Jul 17, 2008 12:41am
HOW CAN I GET GOOD EA'S THAT WORK VERY WELL PLZ SEND TO ME OK

Quoting tdion
okay so 25,000 pips over 30 years might not seem like much.... hell, not even 1000 pips a year. but if you can find 3 or 4 systems as consistent and profitable as this one, then you have enough components to create a full trading system.

just thought i'd express where this is all going.... not leaving it up to the imagination.

suppose you combine this strategy with the one found here:
http://www.forexfactory.com/showthread.php?t=95895

between the two strategies, you would have made 28,000 pips over the last 19 years.... which is a pretty good start to a system. unfortunately, both are heavily dependent on dollar volatility, but its a start.
  • #64
  • Jul 17, 2008 1:04am Jul 17, 2008 1:04am
Quoting Jtmoney
HOW CAN I GET GOOD EA'S THAT WORK VERY WELL PLZ SEND TO ME OK
http://www.forexfactory.com/showthread.php?t=96322
  • #65
  • Jul 17, 2008 3:41am Jul 17, 2008 3:41am
Quoting tdion
the 18/16 one more time, with a 975 TP and 300 SL.... (thus, not always in the market 100% of the time)

profit increased 3,000 pips (to 25,000) w/ the TP and SL....

maximum drawdown only 1,700 pips
Hi Tom,

Interesting result. The low drawdown is impressive. I guess if you can also attain a similarly improved result using minor variations on the [300, 975] numbers, then you can satisfy yourself of the lack of curve-fitting. The fact that the TP > 3 x SL would please all those who advocate high RR. Must admit I'm slowly being persuaded to join their ranks (), so I too find it comforting.
_______________________________________

A little off-topic, perhaps, but a thought about non-linear MAs like Hull, Tillson, Jurik etc. If they performed significantly better (profit-wise) than SMA, EMA etc, then I guess everybody would use them.

Consider the following: A coin toss will produce random trends of heads and tails. Suppose we toss repeatedly, assign +1 to heads, –1 to tails, and plot the running total: we will end up with a graph of rises and falls, just like a price chart. If we apply the non-linear MAs to this chart, we’ll get snugger fitting curves than with the traditional MAs, but will they (or their crossovers) offer greater predictive value? The answer must be no, because our data generator is a random one, hence no prediction is possible, no matter what method of analysis is used.

So what I’m saying is that all of the problem characteristics of the linear MAs – lag, overshoot, lack of smoothness – are not in actuality the problem. In fact, they are only a smokescreen, because the real problem is the extent of randomness in the data itself. An indicator has predictive value only to the degree that the underlying data has – on balance – some recurrent, non-random pattern that the indicator aligns itself with.

As I read somewhere else, indicators are a symptom, not a cause.

Just a thought.

David
  • #66
  • Jul 17, 2008 7:33am Jul 17, 2008 7:33am
Y DID YOU DO THIS I USED YOUR EA AND IT CLOSED A TRADE WITH A LOSS OF 20PIPS...........WHAT IS TH PROBLEM?HOW WAS IT PROGRAMMED

Quoting tdion
I came up with idea a
bout 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.
  • #67
  • Jul 17, 2008 7:36am Jul 17, 2008 7:36am
Quoting Jtmoney
Y DID YOU DO THIS I USED YOUR EA AND IT CLOSED A TRADE WITH A LOSS OF 20PIPS...........WHAT IS TH PROBLEM?HOW WAS IT PROGRAMMED

Have you been reading the thread? Or did you blindly attach this EA to a LIVE account?
  • #68
  • Jul 17, 2008 9:44am Jul 17, 2008 9:44am
Quoting Jtmoney
Y DID YOU DO THIS I USED YOUR EA AND IT CLOSED A TRADE WITH A LOSS OF 20PIPS...........WHAT IS TH PROBLEM?HOW WAS IT PROGRAMMED
They say you never get a second chance to make a first impression. Unfortunately, you've gotten off to a bad start.

When you post messages in all capitalized letters, you're seen as YELLING! It's considered very rude in most public internet forums.

EAs, scripts, indicators, etc., are freely posted by members for the benefit of the community at large. It is entirely up to you, the user, to understand how these programs work before implementing them. I hope you tried it on a demo account first. If not, consider your loss as a tuition fee. Learn from this experience and move on. No one owes you anything, least of all an explanation.

Welcome to the forum.
  • #69
  • Jul 17, 2008 12:20pm Jul 17, 2008 12:20pm
Quoting opie999
http://www.forexfactory.com/showthread.php?t=96322
  • #70
  • Jul 17, 2008 12:25pm Jul 17, 2008 12:25pm
Quoting Jtmoney
Y DID YOU DO THIS I USED YOUR EA AND IT CLOSED A TRADE WITH A LOSS OF 20PIPS...........WHAT IS TH PROBLEM?HOW WAS IT PROGRAMMED
try this:
http://br.youtube.com/watch?v=sk0ARmuGvjY
  • #71
  • Jul 18, 2008 5:08am Jul 18, 2008 5:08am
Quoting tdion
the 18/16 one more time, with a 975 TP and 300 SL.... (thus, not always in the market 100% of the time)

profit increased 3,000 pips (to 25,000) w/ the TP and SL....

maximum drawdown only 1,700 pips
To those we were sceptical/negative: come on, guys, give Tdion a break (and the credit he deserves)!

He didn't invent MAs, he didn't invent MA crosses - those instruments have been around for ages; some use them often, some a lot, some never. But an MA is an MA - there is not mistery and no uncertainty about it.

If MA crosses at certain conditions for 30+ years, will then stop doing that tmrw? Of course not.

Tdion is not offering a ready-to-use trading system, but just interesting food for thoughts. We do know that some MAs, when crossing, indicate some trend change. Tdion's analysis helps to see which MAs produce stronger (and more consisent - not 100% right) effect. That's all. It's then up to each one of use to do (or not to do) smth with that knowledge.

Tdion, as David "Hanover" said, low drawdown is impressive. But what if that happens at the very beginning of the trade? Just a hypothetical question - is there any correlation between MAs crosses and other "strong" indicators which you may have noticed which would allow to apply additional criteria - i.e. choose only the trades with high expectancy of high RR? I guess that's what you meant by combining MA cross with other strategies, so I thought maybe you have some specific systems/indicators in mind (apart from the one you already mentioned).

Thank you for sharing your findings with us, keep it coming!
One Chart is Worth Ten Thousand Words.
  • #72
  • Jul 19, 2008 10:21pm Jul 19, 2008 10:21pm
forex007,

if i cared what other people said, i wouldn't get anything done. the way i look at it, it's my money, not the naysayers....

okay, rewrote the optimizer to find optimal entries based on RSI values (instead of MA crosses)

here's an H1 EURUSD RSI backtest from 2001-2008 that made 5000 pips in 500 trades...
Attached Image (click to enlarge)
Click to Enlarge Name: TesterGraph.gif Size: 17 KB
  • #73
  • Edited Jul 20, 2008 4:35pm Jul 19, 2008 10:51pm | Edited Jul 20, 2008 4:35pm
one more....

same strategy, optimized for the H1 pound.... 4,000 pips (360 trades) over the last 3 years....
Attached Image (click to enlarge)
Click to Enlarge Name: TesterGraph.gif Size: 17 KB
  • #74
  • Jul 19, 2008 11:13pm Jul 19, 2008 11:13pm
Ok, now that we have this data, can we try combining them?

Main indicator combinations used are:

MA cross
RSI
Stoch

what else?
  • #75
  • Jul 19, 2008 11:16pm Jul 19, 2008 11:16pm
let me think on it, ronald

i'm not sure if it will help or not.... and i don't want to waste my time coding a complex strategy if i can see a flaw in the logic before hand....

ps) thanks for stopping by
  • #76
  • Edited Jul 20, 2008 12:40am Jul 19, 2008 11:51pm | Edited Jul 20, 2008 12:40am
just for fun

(same RSI strategy) GBPJPY H4 2001-2008.... 4000 pips, but only 125 trades
Attached Image (click to enlarge)
Click to Enlarge Name: TesterGraph.gif Size: 16 KB
  • #77
  • Edited 12:40am Jul 20, 2008 12:21am | Edited 12:40am
i am reaching a crossroads in my forex journey.

m.a.c.doug, your points are clearly noted. the fact that curve fit strategies are possibly purely coincidence (vs. an invisible edge) is a definite possibility and puts my stomach in knots.

however, i am proving that optimized mechanical rules have the POTENTIAL to produce huge pip totals in appropriate market conditions. do i expect them to always be profitable? NO. i realize that a large population size is my best chance with this approach.


let's face it people: we don't have insider information into what the central bankers will do. that is the only true guarantee i can ever see to taking profits... so we are left to speculate.

at the end of the day, whether you trade james16, or elliott, or fib, or gann.... you have enough faith the put your money on the table based on that type of gamble.

i am showing hard evidence that certain TA patterns can be converted into mechanical rules, and produce profitability (which is more than i can say for 99% of the systems i see on FF) no offense... it's just that i require some form of a track record before i trade real money, and a 6 year back test with 500 trades is about the best i can ask for.
  • #78
  • Jul 20, 2008 2:58am Jul 20, 2008 2:58am
The old saying goes... "Past performance may not be indicative of future results." Or as someone once said "We trade futures, not pasts."

Of course optimal mechanical rules can produce great potential results on past data, take a look at Ron's billion dollar EA thread. It's possible to create an EA that trades exactly to past data, but that doesn't mean anything for future results. I read a book on mechanical trading and the author made a clear point... optimization should be limited to no more than a few robust points. The more you optimize for past results the worse things perform on new data. That's the point of a Monte Carlo simulation, you run a bunch of random data inputs and see how a system performs given a purely random sample over large periods of time. Some systems are profitable even given purely random data.

Quote
i am showing hard evidence that certain TA patterns can be converted into mechanical rules, and produce profitability (which is more than i can say for 99% of the systems i see on FF) no offense... it's just that i require some form of a track record before i trade real money, and a 6 year back test with 500 trades is about the best i can ask for.

Except if you take one set of rules and apply it to another pair... does it show the same profitability? Of course not. You have to optimize to past data, should things change... you're in trouble. Unless of course you find a way to generate optimal parameters based on past price data. Pair behavior does not tend to change overnight, so if you could automatically adjust you might be able to stay ahead of the curve. Of course that's basically a neural network at that point, which is way beyond the scope of traditional mechanical traders.

Quote
at the end of the day, whether you trade james16, or elliott, or fib, or gann.... you have enough faith the put your money on the table based on that type of gamble.

But yet people are profitable doing all of those. Why? Why can you be profitable with what is essentially a random entry? Perhaps entries don't matter, or matter a lot less than we think. I've been saying it for a long time now... a good exit and good risk management can turn a losing system into a winning system. If you only win 1 out of 4 trades, but have a 5:1 average return ratio... you're good to go. So instead of focusing so much on entries why not focus on ways to shrink your losses?
  • #79
  • Edited 4:38am Jul 20, 2008 4:22am | Edited 4:38am
"some systems are profitable with purely random data" - i'm calling BS. you'll need to show it, not just make the claim.

as for optimization... i suppose i could just trade parameters that are the worst losers. that's the answer, right? of course not....

i have repeatedly explained how a large sample size is the best, and how i recognize the curve-fitting issue. if you truly have a winning strategy, it will have performed well in the past, and will in the future. therefore, it is a blanket statement to say that curvefit systems are garbage. within the entire set of all curve-fit systems are some legitimate profitable strategies.

FYI the past 3 graphs are all the same strategy, but parameters unique to the pair. i would never expect to trade EURUSD with the identical trading rules as GBPJPY, because the ATRs are different. of course the fibonacci lemmings will tell you to measure bars and trade at the levels (which reminds me.... isn't fibonacci the ultimate curve-fit? i mean, how anal are those guys to trade at EXACTLY the 23.6% level? )


This is scary:
Quoting rabid
I've been saying it for a long time now... a good exit and good risk management can turn a losing system into a winning system.
  • #80
  • Jul 20, 2008 4:42am Jul 20, 2008 4:42am
Quote
"some systems are profitable with purely random data" - i'm calling bullshit. you'll need to show it, not just make the claim.

Depends on what you want. I have a system that works w/ purely random data, but I'm under an NDA that limits what I can disclose. I can say that it's a trend following approach that looks for certain variation with a very tight risk control. I must admit I don't particularly feel a need to convince anyone of the possibilities tho. Still, if you take a look at some of my recent posts you may be able to deduce some of the basic concepts. It is a purely mechanical method.

Quote
as for optimization... i suppose i could just trade parameters that are the worst losers. that's the answer, right?

That's why I said some of the best mechanical systems are those that use a few robust points. You pick, say, 3 parameters and optimize those. Example: 1st moving average, 2nd moving average, time frame.

Quote
i have repeatedly explained how a large sample size is the best, and how i recognize the curve-fitting issue. if you truly have a winning strategy, it will have performed well in the past, and will in the future. therefore, it is a blanket statement to say that curvefit systems are garbage. within the entire set of all curve-fit systems are some legitimate profitable strategies.

If you have a system that works well in the past there is no guarantee that it'll continue to work in the future. The turtle's system is a perfect example of this. Try it, as documented, on the EURUSD and see how it performs. Now take the same system and run it against the GBPUSD and compare the results. There are a number of systems that have become so well traded that it makes more sense to fade against them than to trade with them. This doesn't happen overnight, clearly, but does happen. Which is why a self-optimizing system could certainly work as well as anything else, maybe better.

If people can be profitable w/ any variety of entry systems, why do entries matter at all? I agree that fibs are a curve-fit voodoo, yet they work. How's that possible if the entry is, for all intents and purposes, random?
0 traders viewing now