why curve fitting doesn't work? | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Optimization, statistics, and avoidance of curve fitting 21 replies

Curve Fitting: The Ultimate Conundrum 17 replies

EA Backtesting and Curve Fitting Concern 2 replies

(curve)-fitting and overfitting, good examples wanted 0 replies

"curve fitting" or just improving?? 5 replies

Attachments: why curve fitting doesn't work?
Exit Attachments

why curve fitting doesn't work?

  • #21
  • Sep 14, 2007 7:59am Sep 14, 2007 7:59am
Quoting bluefox
I still don't understand how could a system that produced such a realistic equity curve (attached) over more than a couple of yrs go all wrong.
bluefox, there's nothing wrong with optimization if done correctly. Here is the first problem as I see it:

That is not a realistic curve. People don't make $111K in 80+ trades and you don't stand a chance of surviving if that is what you are shooting for. Anytime I see a juiced up curve with a PF of 2.5 or more I shrug and walk away. That's not real life.

As Mr. Trend mentioned, you need to perform a walk forward anaylsis. Take a data set, reserve 30%, test on the known data and then expose the new data. If the walk forward is valid you should see around 50% of the optimised results on an annual adjusted basis.
  • #22
  • Sep 14, 2007 9:37am Sep 14, 2007 9:37am
Quoting jimbil

My plan is to backtest and optimise the past 6 months H4 data, trade these inputs live for the following month, and then backtest/optimise & possibly change the inputs at the end of each month. How does this plan sound?
it sure sounds interesting. I once had a similiar idea and tried it out but the result wasn't very promising. however it might be just because of my EA. it could very well work out for you. please keep me posted about your experiment if you will.

good luck
  • #23
  • Sep 14, 2007 9:54am Sep 14, 2007 9:54am
Quoting philmcgrew
That is not a realistic curve. People don't make $111K in 80+ trades and you don't stand a chance of surviving if that is what you are shooting for. Anytime I see a juiced up curve with a PF of 2.5 or more I shrug and walk away. That's not real life..
I know there are realatively not many trades at all. but my system is a trend-following system. and the 80+ trades are generated over a 5 yr period. the curve is even more impressive when I turn on fixed ration MM for the EA. see attached picture http://www.forexfactory.com/images/icons/icon10.gif.

Quote
As Mr. Trend mentioned, you need to perform a walk forward anaylsis. Take a data set, reserve 30%, test on the known data and then expose the new data. If the walk forward is valid you should see around 50% of the optimised results on an annual adjusted basis.

this is exactly what I'll be doing in the near future. if not for profitability, then for fun and the kick of producing a nice equity curve http://www.forexfactory.com/images/icons/icon10.gif. I'm not sure about the meaning of your last sentence - "If the walk forward is valid you should see around 50% of the optimised results on an annual adjusted basis", could you please clarify a bit more?

thanks a lot!
Attached Image (click to enlarge)
Click to Enlarge Name: TesterGraph.gif Size: 22 KB
  • #24
  • Last Post: Sep 14, 2007 10:12am Sep 14, 2007 10:12am
Quoting bluefox
I know there are realatively not many trades at all. but my system is a trend-following system. and the 80+ trades are generated over a 5 yr period. the curve is even more impressive when I turn on fixed ration MM for the EA. see attached picture http://www.forexfactory.com/images/icons/icon10.gif.



this is exactly what I'll be doing in the near future. if not for profitability, then for fun and the kick of producing a nice equity curve http://www.forexfactory.com/images/icons/icon10.gif. I'm not sure about the meaning of your last sentence - "If the walk forward is valid you should see around 50% of the optimised results on an annual adjusted basis", could you please clarify a bit more?

thanks a lot!
While the curve might be impressive, if it doesn't represent real life then it remains a pleasant fantasy. I would also suggest that you optimize without money management first and then bring that part in after you have the best possible curve.

Regarding my last statement - You optimize a curve. Let's say you do it for 1 year and it makes $10,000. You then walk forward for 3 months and those 3 months yield $1000. On an annualized basis (comparing apples to apples) $1000 (per 3 months) is $4000. So the optimized makes $10,000 per year and the unseen makes $4000 per year or 40%. My point was that you will rarely exceed 50% of the optimized amount. If you approach 50% you should be very impressed with your efforts.
0 traders viewing now