Optimizing "The Sauce" | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

Broker Challenge "NDD/STP" vs. True "ECN" and "Mkt Mkr" broker talk 46 replies

Dealing with "Invalid Price", "Requote", "Server Busy" in MT4 3 replies

Thoughts on backtesting and optimizing 6 replies

Twoblink and "The Secret Sauce" 79 replies

Attachments: Optimizing "The Sauce"
Exit Attachments

Optimizing "The Sauce"

  • #1
  • First Post: Edited 10:42pm Jun 18, 2006 12:11pm | Edited 10:42pm
http://www.forexfactory.com/forexfor...0&page=1&pp=15

Now that I've found the sauce.. I now need to "optimize" the sauce.

So I have closed my old journal of looking for the sauce, and now have opened up another journal to document the optimization of the sauce.

PROBABLY, this will be done in python..

I will be using historical data from 4 pairs:

Eur, Yen, Cable, and CHF. The Dailys.. I will run them back as far as probably Jan of '04.

I will tweak them, and post simulated results.. I will probably test worst case scenerios of rulesets.

Worst case means, I always assume that if I'm going long, I buy at the highest, and if I'm shorting, I shorted at the lowest. And when I close, visa versa.. I take highs and lows against me. If I do that and it still produces a healthy profit, I know the system is very very good.

So I probably will post some observations and frustrations of data testing.
google:
  • #2
  • Jun 18, 2006 12:19pm Jun 18, 2006 12:19pm
Fast testing..

I assume the programming can be done in a day, and optimizations plotted. I will probably try to do a cool graph or at least a nice spreadsheet.

The reason the programming will be fast is because...

1) Python is a fast language to program
2) "The Sauce" doesn't require indicators!! That means I'm just using raw data (Open,High Close, Low)!!
3) The ruleset for it will be fairly simple.

Probably, the most amazing thing about the sauce is, no indicators needed!!

I'm going to reward myself by going to bed at 1:30am tonight instead of 3:30..
google:
  • #3
  • Jun 18, 2006 12:20pm Jun 18, 2006 12:20pm
The best way to optimize the sauce is by adding a little more paprika... Hungarian paprika if you can get it

Alright - I'll leave your journal alone now
  • #4
  • Jun 18, 2006 7:42pm Jun 18, 2006 7:42pm
Quoting twoblink
http://www.forexfactory.com/forexfor...0&page=1&pp=15

Worst case means, I always assume that if I'm going long, I buy at the lowest, and if I'm shorting, I shorted at the highest. And when I close, visa versa..
Hi twoblink,

Isn't this the best case?

Shouldn't worst case be buying at the highest and closing at the lowest?


Regards,

Matsu
  • #5
  • Jun 18, 2006 10:05pm Jun 18, 2006 10:05pm
Quoting Matsu
Hi twoblink,

Isn't this the best case?

Shouldn't worst case be buying at the highest and closing at the lowest?


Regards,

Matsu
Yeah... that might make a difference twoblink.
  • #6
  • Jun 18, 2006 10:16pm Jun 18, 2006 10:16pm
Hey guys,

I think it's safe to say this was a typo by twoblink. Based on his journal and OCC thread he seems like a smart guy.

  • #7
  • Jun 18, 2006 10:28pm Jun 18, 2006 10:28pm
Quoting lonespruce
Hey guys,

I think it's safe to say this was a typo by twoblink. Based on his journal and OCC thread he seems like a smart guy.

I think we both knew that but... just in the off chance... Why let him waste his time.
  • #8
  • Jun 18, 2006 10:42pm Jun 18, 2006 10:42pm
Well.. that's what happens when you type this stuff at 2am isn't it??

Yep, made the corrections!! Thanks for catching them.
google:
  • #9
  • Jun 18, 2006 10:46pm Jun 18, 2006 10:46pm
I have to say, I am constantly amazed that you guys actually read my ramblings : and with a fine enough toothed comb to catch my mistakes!

One of the problems I see, is that how long of a period you use, will so drastically influence your results. Back test on a good year, and you get incredible results. Backtest on a bad year, and you get horrible results..

All systems will have drawdowns.. So I'm not sure what is the "fairest" way to backtest it in the simulator..

I should test, 6 months, 12 months, 18 months, and 24 months; as different groups... or that's my current initial thinking...
google:
  • #10
  • Jun 19, 2006 4:49am Jun 19, 2006 4:49am
Preliminary tests seems to show that with the sauce on a daily chart, you will eventually end up making serious money.

I don't know if I should giggle like a 6th grade girl over this; or cry over the fact that my last 19 years of learning about the markets have all been a big farse..

Even with worst case scenerios of buying high and selling low, it still makes money... I pretty much have no idea what else I can ask for from a system. The pathetic thing is, it's not even a system.. It's just sauce..
google:
  • #11
  • Jun 19, 2006 5:31am Jun 19, 2006 5:31am
Quoting twoblink
I have to say, I am constantly amazed that you guys actually read my ramblings : and with a fine enough toothed comb to catch my mistakes!

One of the problems I see, is that how long of a period you use, will so drastically influence your results. Back test on a good year, and you get incredible results. Backtest on a bad year, and you get horrible results..

All systems will have drawdowns.. So I'm not sure what is the "fairest" way to backtest it in the simulator..

I should test, 6 months, 12 months, 18 months, and 24 months; as different groups... or that's my current initial thinking...
I tested my system on any 3 month period during last 24 month and it always came in smaller or bigger profit.

my suggestion is to optimize on the worst period of time in history and then test the same settings on the max time you can get your data on and then look on the overall picture.
  • #12
  • Jun 19, 2006 8:20am Jun 19, 2006 8:20am
Hi Twoblink,

Regarding the backtesting, attached is a very thorough backtesting routine written by Merlin.

Kind Regards,
Matt.

Quoting twoblink
I have to say, I am constantly amazed that you guys actually read my ramblings : and with a fine enough toothed comb to catch my mistakes!

One of the problems I see, is that how long of a period you use, will so drastically influence your results. Back test on a good year, and you get incredible results. Backtest on a bad year, and you get horrible results..

All systems will have drawdowns.. So I'm not sure what is the "fairest" way to backtest it in the simulator..

I should test, 6 months, 12 months, 18 months, and 24 months; as different groups... or that's my current initial thinking...
Attached File(s)
File Type: doc The System Behind the System-Very Thorough Backtesting Procedure.doc   36 KB | 715 downloads
  • #13
  • Jun 20, 2006 9:12pm Jun 20, 2006 9:12pm
So I went ahead and wrote a sim for all this..

Interesting results. What I am seeing in reality is, random entries on random sides of the FX market seem to do much better than most systems presented on this forum

Now the scary news.. Random entries on random sides of the FX market seems to do much better than most systems presented here.

Yes, I repeated that twice..

It also seems to show, something about "staying power". Staying power has to be priority one.

Over 1 million iterations were run at a time.. The results were fairly consistent.

It shows that what "The Sauce" does, is reduce profitability by about 10% over maximum profit equations; BUT reduces risk by about 50%.

I think that's a great trade.. 5:1 risk reduction vs profit reduction.
google:
  • #14
  • Jun 23, 2006 11:43am Jun 23, 2006 11:43am
Conclusion:

If you are at 50/50 in terms of odds, you need a 2:1 profit/loss ratio.

If you are at 66/33 odds, then you can be 1:1 in terms of profit/loss ratio.

Basically, any of the two, one needs to be 2:1.
google:
  • #15
  • Last Post: Mar 10, 2011 6:13am Mar 10, 2011 6:13am
How does the secret sauce tastes after 5 years ? Are you retired ?
I came randomly to your "Searching for the sauce" thread because of some keyword. Now that i read these 2 journals, it stops abruptly. What about end 2006, 2007, 2008, 2009, 2010 turned out for your sauce (based on probability calculations to provide you an edge, isn't it ?)
0 traders viewing now