Trader's self-deception: distrust of forwards. - page 8
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
How forward checking is done - we start from the basics.
You take a sample, divide it in a certain proportion, e.g. 80/20, into back and forward.
And how do you think forward checking is done - we take the entire sample up to the present moment and accept it as back? And then real time results are tested forward as the quotes come in? And how long does it take you to forward??????
You can take any segments with any breakdown on back and forward, but one thing you cannot change - forward is always a run on an UNOPTIMIZED section. You can do whatever you want on backwards - turn everything upside down, change, modify, try, but result of forward check will be one forward. If you spin some more, you get one forward again, etc. But sustainability is (as a rule) the repeatability of the result over time. While practicing your methods on one and the same area you will not know if there is a repeatability on different areas. So you will need forwards not only of different variants of code (not settings), but also forwards from different parts of history.
The history part is not optimized, the indicator settings are optimized exactly for this part of the history.
Again in order - we take the entire history, split 80/20 (20% of the forward will be the last observations).
We take the indicator, set the parameters, and run it on the backward section, obtain the MO.
Then we change the parameters and run it again. And so on.
We pass it 100 times with different parameters and select, for example, 20 variants of parameters that give different POSITIVE MO at the back stretch.
Let's try all 20 of them on the forward - some of them, for example, 5 give positive values on the forward position too. We selected the parameters of the indicator that give the maximum forward MO.
So - all this is the same fit, but only ON BACK AND FORWARD.
The history part is not optimized, the indicator settings are optimized exactly for this part of the history.
Again in order - we take the entire history, split 80/20 (20% of the forward will be the last observations).
We take the indicator, set the parameters, and run it on the backward section, obtain the MO.
Then we change the parameters and run it again. And so on.
We pass it 100 times with different parameters and select, for example, 20 variants of parameters that give different POSITIVE MO at the back stretch.
Let's try all 20 of them on the forward - some of them, for example, 5 give positive values on the forward position too. We selected the parameters of the indicator that give the maximum forward MO.
So, all this is the same fit, but only both on the BACK and on the FORWARD.
You have already described it. And I told you that it is not a forward check. You are still cheating by "peeking" into the future when you choose from some forward options. Forward is an honest thing, not allowing "peeking" at answers. You should make only one decision on backtest and get only one answer - good or not. Because when you set a set of variables on a really working EA, you will set one set, not 20, and there will be nothing to choose from, because the future hasn't come yet.
You have already described it. And I told you that it is not a forward check. You're still being sly by "looking" into the future when you choose from some forward options. Forward is an honest thing, not allowing "peeking" at answers. You should make only one decision on backtest and get only one answer - good or not. Because when you put a set of variables on the really working EA, you will put one set, not 20, and there will be nothing to choose from, because the future hasn't come yet.
I don't understand.
You have the AUDUSD history for the year M15. You split it up - 11 months back and the last month forward.
You have just learned that there is such a cool indicator as a moving average. This indicator allows you to make adjustments - you select the period of the moving average.
You build the owl and set the sliding period to 3. Run it on the back - negative MO.
You set the period to 4 and again run it through - positive MO. Roll owl on forward - negative MO.
Put on period 5 and repeat.
And so on.
Exactly the same thing I wrote above.
I don't get it.
You have the AUDUSD history for the year M15. You split it up - 11 months back and the last month forward.
You have just learned that there is such a cool indicator as a moving average. This indicator allows you to make adjustments - you select the period of the moving average.
You build the owl and set the sliding period to 3. Run it on the back - negative MO.
Set the period to 4 and run it again - positive MO. Roll owl on forward - negative MO.
Put on period 5 and repeat.
And so on.
Exactly the same as I wrote above.
OK, make it 11+1. What should the forward look like. Run all options on the 11-month back, get the optimum value. Put it on forward - got the answer. That's it. Whatever the answer is, it is not a set-up, it is a true forward. Any further manipulation with the results of different forwards during the last month is a fitting. If you need to study sustainability you do a consecutive series of back-to-back forwards using the 11+1 formula. I.e. 6*(11+1) with month shift and you get 6 forwards that can be compared with each other to get repeatable results. Ideal result is when all 6 forwards from different sites will satisfy you, which will be an indicator of stability.
))) And if there were three variants with positive OR, you chose the one with maximal OR as the optimal one, ran it on forward and it showed negative OR on forward?
What will you do, will you check the rest two variants on the forward or not?
Or with a cry of "That's not fair!" You'll delete them without checking?
))) And if there were three variants with positive OR, you chose the one with maximal OR as optimal, ran it on forward and it showed negative OR on forward?
What will you do, will you check the rest two variants on the forward or not?
Or with a cry of "That's not fair!" You'll delete them without checking them?