Result of Backtest with 99.9% model quality. - page 2
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Do you mean in backtest no slippage or spread?
How to test with slippage or spread?
Please advise me.
Thank you in advance.
Do you mean in backtest no slippage or spread?
How to test with slippage or spread?
Below is the result of fix lot =15 and the other remain the same.
I use pepperstone demo razor account properties to generate hst and fxt files.
So spread of pepperstone as low as 0.2 and I use CNS VPS to faster connect to pepperstone server.
You didn't answer my questions . . . slippage is not implemented/simulated in the Strategy Tester, so if your trades are slipped by a pip or two how will that affect your performance ?
99.9% is a meaningless figure, it just means that you used tick data, it is simply the figure used to show you that you used tick data, it could just have easily been 1245%
Your performance looks good, what date range is this over ? how many pairs have you repeated this on ? is this optimized at all ?
I use only EURUSD in year of 2011 for testing.
Previously I try to optimize to select suitable period of indicator and other 3 parameter.
after that I choose suitable parameter setting for previous backtest result.
So what cause of the backtest result after select suitable parameter.?
Please tell me.
If you use non-fixed, it will choose parameter sets that favor the earlier portion of the test and aren't bad in the later portion. That's why you must optimize on part and test on a different part (year.) Set could be terrible on another year.
If you use fixed, it will choose parameter sets that favor the best average performance over the entire portion.
I wrote the multi-pass optimization so I could find get a parameter set that was positive each year over multiple years.
If you use non-fixed, it will choose parameter sets that favor the earlier portion of the test and aren't bad in the later portion. That's why you must optimize on part and test on a different part (year.) Set could be terrible on another year.
If you use fixed, it will choose parameter sets that favor the best average performance over the entire portion.
I wrote the multi-pass optimization so I could find get a parameter set that was positive over multiple years.
Only one way to find out, Live test it. You should have a pretty good idea after about 3-weeks. Or 100 Independent trades.
You should also try other currency pairs (example:EURJPY) and test additional years (example:2012). Faster to backtest than live test. I mean it simply cannot hurt to produce additional tests.
Don't put too much faith in Optimizers, if the system wasn't profitable before optimization, chances are that it's curve fitted.
Parameter hunting is_not the answer. Focus more on market_characteristics like trending vs ranging etc. Imo, its much more dynamic and simpler to adjust system to market_characteristics than Optimizing, Indicator|Stoploss|TakeProfit|Trailing & Time_Frames every other week.
In my Opinion, realistic will depend upon allot of factors. Only people who really understand Statistics could make bold clams within this area. I suck at math, however, I've read and taught about this allot since my black-jack days. I cannot say if you're correct or in-correct however I believe whats more realistic would come down to the type of system someone is developing; their personal take on risk and how-much investment-capital they have. All of the previously listed are Subjective and make this question less black & white. Just some points to consider.
1> How many people trades with fixed lots for the rest of their lives?
2> How many people trades 1-trade at a time. (Independent trades)?
3> How many people can afford the Broker's Minimum Lot?
4> How many people trades for capital-growth vs capital-preservation?
I'll try to give my interpretation of the above questions.
1. Very few people would trade with fixed lots. As their accounts moves from 1k to 2k, most people would tend to double the fixed lot. If the system does well doing the 1k_Period but not_so_well during the 2k_Period, that 9% Relative DrawDown now becomes nearly 20%. But they're trading thinking the system still has a RelativeDD of 9%.
2. Very few people again would trade 1_Trade@Time. Achieving this with Fixed_Lots is obviously the Holy_Grail of all testing. I'm going to make a Bold statement here "It simply cannot be done". My 1st years of trying to develop systems like this failed. Now sure I did make some using the Strategy_Optimizer, but never works live. If you've succeeded in creating a system which does this without depending on Strategy_Optimizer then I would love to hear. Just a simple I've created it would do :)
3. Now allot of people can afford this (or think they can afford it). And this kinda relates to question#1. Even someone who has no plans of increasing the lotsize can mis-calculate their risk by testing with 10k and then depositing 100$ into an account thinking they have the same risk. If the system's maximum_drawdown is greater than 500$, this person might be setting themselves up.
4. I believe allot of people trades for capital-growth but still wants to make constant with-drawers. In this example again, their relative risk could be higher than expected because the tester produced low relative-dd because it kept all the profit within the account.