Fratelli EA by Dragos and magft

[johnjwong]

Well said, backtesting does not consider the change in spread. In the weekend it got lock up to sometimes 20 pips for some brokers. However, good thing we have the mt4i spread changer that we can change the spread for ST purposes to an amount we see reasonable, then perform the backtest.

Not good enough! That will NEVER give accurate results.

You MUST use real spread that changes with every tick...the difference between Bid and Ask. If you want to additionally pad that spread, so be it. And as commissions per trade are very common these days, you must be able to account for this, too, or your results will not be very accurate. Empty4 does not allow for this OOTB...hence the use of tools to give Empty4 a kick in the butt!

You are well knowledged Kenfen

[garyfritz]

I wanted to contribute something to the thread, so... I implemented a Tradestation strategy ("EA") for Fratelli.

I'm not sure it's working 100% correctly on all cases yet, but it seems to be very close. If someone has a complete list of correct Fratelli trades, I would like a copy so I could compare against my strategy. Dean's results aren't valid because he uses his trailing EA. As far as I can tell my results are right. I may have some misunderstandings of some details of the rules.

I couldn't remember the correct trading hours so I approximated. By default it uses 07:10 - 17:00 GMT.

Since everyone's just been discussing this: this TS strategy does not calculate fills properly using a bid/ask basis. TS is using only bid for its price quote, so all the inaccuracies you would expect from that will be present in these tests. I don't try to deal with varying spreads in my testing; I find over the long haul I can just add an estimated slippage figure and that will reflect actual fills pretty accurately. In several hundred real-money trades, I average less than half a pip slippage on stops, and that includes all spikes & other unusual fills.

So, that said:

Here's Fratelli (or something very close to it) running on AUDUSD:



The indicator and the system have no "repainting" or "changing its mind" behavior. What you see in the backtests (visual or automated) should reflect what you'd see in realtime. The Supertrend candles don't match the Empty4 version (prior to Mickey's recent non-repainting changes) exactly but I think that's because Empty4 repaints.

I backtested the system for the past year. AUDUSD has won over 400 pips in the last 3 months -- BUT it was a steady loser for 9 months before that:



EURUSD was similar:



EURJPY has had ups and downs but has won over 1000 pips in the last year:



Then I did a quick analysis of EURUSD trades by time. I opened it up to 24hr trading and looked at the results of trades entered in each hour:



You can see that the Frankfurt & London opens (7:00 - 9:00 GMT) are by far the most active. In my testing, at least, they are also the least profitable. So I changed the trading hours to 09:00 - 17:00 GMT, and EURUSD looked a lot better:



Obviously this is "in-sample" testing -- looking at the same period that I used to select the trading hours. I don't know how well it would carry forward. Looks like it might be worth studying, though.

Gary

[kenfen]

Gary...what a fantastic effort and contribution!!! Standing ovation, please!

OK...so where to start. First of all...all of my tick data is still in the process of being converted to a usable format for Empty4. I'm probably still about two hours away from being able to roll. Once I can roll, the results will be in quick.

As for how close to 100% Fratelli you're running in TS, I'll have to leave that to you, Peter (Fratelli), Micky, Dragos and magft. I can say this...TS just "looks" so much nicer/crisper than that broken down old goat...Empty4!

I guess from here, we really need to see some results from my end. That is probably going to be more of an apples to apples comparison, given the Empty4 origins of the system, indicator and EA. Tell me this...is your strategy running as close to "Fratelli" as possible? I know the EA here has "JumpStops" or other things going, which will allow it to follow the runners, but I'm going to try and eliminate that at first to just backtest by the book. Once we get close (90-95%) accuracy in trades taken, then we can explore those options. From your AU chart posted, I suspect you're "by the book." <--good

As to your Trading hours/Time of Day observations: I'm not remotely surprised that skipping the Frankfurt hours improves things. When I'm manually trading those hours (which I'm not right now...my shop is pretty much closed until Jan. 15th or so), I'm in "uber-scalp" mode. It is a great time to grab 10 pips here and there, but is a very dangerous time to try and catch a trend.

What I see often...and I'm not looking at any charts right now...is: 15-20 mins before Frankfurt opens, price will start to mini-trend (Dangerous for Fratelli.). Frank's bell rings, and price will push through Supply/Demand/STOPS (Fratelli's Pending Orders) in that direction. 5-10 pips above/below those levels, it will violently turn in the other direction (Fratelli Stop Loss). Now up/down to the other extreme of the last few hour's quiet Asian session range. Stops get done again...looks like another trend forming (probably hasn't been enough to trigger Fratelli in that opposite direction, though.) Then stable out for 20mins or so, then start the same cycle again coming into London's open...and another potential "fakeout" is on the cards. Anyway...blah, blah, blah. I have a feeling that 10-20mins after London open will provide the most stable results. Bit of a prediction...we'll see.

Last bit: Spread, slippage etc. One element I didn't mention that I can add to the backtests is slippage. I will do. Although Fratelli uses PO's which will certainly keep slippage at a minimum...and if your broker is any way decent...it will keep it at ZERO! I'll run a few tests using different settings, but I normally go with:

Use Real Spread: True
Slippage: 0.5pip (in and out)
Commission: 0.5...maybe 0.7

Those are VERY "real world" settings. I'll pad the spread (to make it worse), up the slippage and commission, too, for comparison sake, although...I don't think that is where the profitability...success/failure...is going to lie here.

On one final note...LIVE tests is where it is at! Once I see that an EA, Strategy etc. is coded in a decent manner...as in...it isn't going to open and close trades on each tick until my account is gone...I run live...smallest lots possible. My only hope with the backtests is to save Peter, or anyone else, serious hours of F12 heaven!!!

Trading Fratelli manually is always going to be different than running an EA/Strategy. Once we lock our fundamentals to a 90/95% tolerance of the actual system, then we can start Jumping stops etc. Or, put another way...trying to "humanize" it. If I enter a Fratelli trade and price takes off in my favour...I'm not going to close the whole trade at +20...let your winners run...at least a portion of them!

Wow...what a book I've written! Sorry.

My data is nearly ready to roll...should have some results in the not too distant future.

Nice work, Gary!

[blahn]

I consider myself a very novice forex trader, but for what it's worth had an idea. We've all seen fratelli's system remain pretty consistent as long as it's traded under moderate volatility to create the right size waves. It seems to be the ranging and the high spiking volatility sessions that hamper it a bit. Is it too rash an idea to suggest adding a ranging/volatility filter to the EA to act as an on/off switch? That way none of the rules are changed, but trades during weak spots are avoided whenever possible.

[kenfen]

Is it too rash an idea to suggest adding a ranging/volatility filter to the EA to act as an on/off switch? That way none of the rules are changed, but trades during weak spots are avoided whenever possible.

I don't think its rash at all.

Do you know of a "ranging/volatility filter" that might assist? Chart it out. Wouldn't take long to implement...just need the tool to use.

[garyfritz]

Tell me this...is your strategy running as close to "Fratelli" as possible?
From your AU chart posted, I suspect you're "by the book." <--good

That's what I'm trying for. Once I have the "by the book" behavior, I can start testing and improving. And backtesting -- one major reason I wanted to do this is because I don't trust the Empty4 backtester, and I want a measure of Fratelli's long-term performance that I can trust.

If I decide to trade it, I would eventually need an EA of some form. Fratelli seems to only behave between about 1am and 9am my time. I'm not coherent enough to trade at those hours!

On one final note...LIVE tests is where it is at! Once I see that an EA, Strategy etc. is coded in a decent manner...as in...it isn't going to open and close trades on each tick until my account is gone...I run live...smallest lots possible.

Live testing is of course the ultimate acid test. But backtesting gives you an idea how the system performed in the past, over many more market conditions than you're likely to see in the next few days or months. So if I look at what it would have done for the last year, and if I find it lost money pretty consistently, I'm not going to waste time or money live-testing it.

*If* my implementation is correct -- which is NOT certain -- then I am seriously concerned about Fratelli's stability. Looks to me like it would have lost lots of money before September.

[MickyMouse]

Good morning everybody .. nice work ....
need to have a presso and catch up with things

[blahn]

Is it too rash an idea to suggest adding a ranging/volatility filter to the EA to act as an on/off switch? That way none of the rules are changed, but trades during weak spots are avoided whenever possible.

I don't think its rash at all.

Do you know of a "ranging/volatility filter" that might assist? Chart it out. Wouldn't take long to implement...just need the tool to use.

Might be easiest to just use current ATR values to turn trades on or off. low ATR is ranging and high ATR is too much volatility. we'd just need to find the right ATR range for the sweet spot of our timeframe.

[kenfen]

Might be easiest to just use current ATR values to turn trades on or off. low ATR is ranging and high ATR is too much volatility. we'd just need to find the right ATR range for the sweet spot of our timeframe.

Hi,

Don't forget...ALL indicators lag. I wouldn't call Fratelli an "Uber" fast scalp system, but its setups happen pretty quick. If you're going to use an Indy for a "yes/no" trade filter...well...we all want to hear which one works!

I'm not going to have time to test this (and I'm more trying to make a point), but why don't you test ATR? Suggestions:

Given that indicators lag, I don't think you can use the given M10 TF, or anything above. Why don't you open the M10 chart with template...and open a second M5 chart with ATR attached and chart it out? See what you find, and report back.

[MickyMouse]

me and another 2 team of traders did an ea for this system .. Actually 3 ea`s in total ...

We figured out how to run backtests on 10 min tf and we backtested the system.

I feel is my duty , as a trader, and member of this community to warn everybody out!!!!!

1 . What we came to see is that the last 2 - 3 months the system has been profitable .
2. Before of that , the system has been a constant looser .
3. The system tested on dukascopy tick data only on the eur/usd , blew the account !!!!!!!!

At the time beeing we are experiencing some strange market conditions.. the system seems to like them and make profits. There is no guarantee that the system will continue to work this way . Actually i am 100% positive that if markets calm down, the system will loose big time ...

Please be aware of this when you make a live trade on a live trading account with REAL money inside!!!!
I strongly suggest to everybody until things will be clearer , to only DEMO TRADE the system !!!!


As a trader , when you click buy...selll it becomes you`re business ... still i felt the need to tell everybody where we are now .

Thank you and many green pips