Counsellors to whom. Lots of them and for free! - page 14
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
Yuri, if the rating of strategies in PRS will be assigned not by the period, on which it is generated, but by forward testing, then the strategy survivorship rating will be better? Or is it not feasible?
SSB does not run individual forward tests (And it will not, as this requires additional computing resources. But we do run basic tests for strategies with high ranking, and hence for the oldest TS. So strategies should not only give positive results at the moment they are generated and put in the repository, but also for a long time in order to keep them in the tops. If a strategy starts giving negative results after some time, its ranking will automatically be downgraded.
The sorting in the repository database is arranged so that if the ratings of strategies are different, then the strategies with the highest rating go first in the list. If any strategies have the same rating, then the sorting is based on time, i.e. the oldest (and therefore most tested) strategies have a higher priority.
If you understood the whole nature of these strategies, you wouldn't ask stupid questions, because that's why they don't get answered.
Firstly: We have not had a drink with you
Secondly: the entire nature, conditions, and application of my own (and not only) strategies, I thoroughly understand, and up to the point of absolute "no way".
Thirdly: if you understood what I am trying to say, you would never have expressed yourself so pejoratively - we (you) develop MTS (mech trading system), then we should clearly understand how and with what results it works at different stages........ - at least to reject casual results :)....... manual trading - it is a completely different issue.
Fourthly: we can be on a first-name basis :)
Firstly: We have not had a drink with you
Secondly: the entire nature, conditions, and application of my own (and not only) strategies I thoroughly understand, and up to the point of absolute "no way".
Thirdly: if you understood what I am trying to say, you would never have expressed yourself so pejoratively - in this forum and on this subject many "kopecks" have been broken - if we (you) develop MTS (mech trading system), then we should clearly understand how and with what results it works at various stages........ well, at least to reject casual results :)....... manual trading - it is a completely different matter
Fourthly: we can be on a first-name basis :)
I've let myself get agitated, I'm sorry... You're just criticising and I'm responding in the same way. I don't know about Pardo, but in my opinion, each timeframe should have different oscillator periods and different characteristics of strategies, the coincidence on different timeframes is a logical confirmation of the reliability of the system, because in principle, if there are more reliable factors, the strategy will show better results on different periods with higher probability. All factors of the strategy have the correct interpretation. So please note that we are all talking about the same thing, despite the different formulation of the question. And I also know that the minute strategy tools clearly shouldn't work on a 4 hour clock. And I'm used to calling timeframes not "stages". Here we discuss PRS strategies in any of their variants, and manual trading can be made automatic. So be clearer about what you mean.
Added arrows.
Are the strategies separated by timeframe in this repository? Or is there no separation in the repository, i.e. the same 20 strategies are given for all currency instruments and different timeframes?
There is no division.
Is it correct to run two PRSs in parallel from the same computer?
It makes no difference if it's from one or a hundred computers. The PHP repository script doesn't log sessions by IP address, and cookies are not involved at all - there's no need for that.
But it does save time. That is, you can install to one PC multiple terminals and multiple SSBs for each of them. For example, each separate terminal will be for a separate instrument. Start one SSB for one tool, wait until optimization is completed, start the next one, etc. We get real paralleling of processes.
I sketched out a rough algorithm for working with the PRS, I would like to hear your comments:
1. I start SBS and choose the smallest timeframe (1min), which provides the largest number of trades and, consequently, the greatest statistical significance of the tests.
2. Select a currency instrument with the highest volatility (e.g. EURJPY, GPBJPY), which will probably give the highest profit and run the PRS.
3. I do my work, and then I save the first 5 strategies, which are already consistent with the principles of testing in different currency pairs and timeframes (because that is how the SBS algorithm works).
Did I understand correctly, the repository indicator values do not change when testing with different symbols and timeframes?
4 I test strategies on all historical data (1999-2009) and the latest historical data (2009.01.01-today). I select strategies with the highest profitability (total profit/total loss), expected payoff, number of trades, the lowest drawdown in dollars, the smoothest balance curve, etc. The selection can also be done for different currency pairs (or timeframes as well?), and for which of them the best values are shown.
The result: one of the five strategies, the currency instrument and the time frame that achieved the best values.
5. With those selected in step 4, I conduct a phased forward test using R. Pardo's method. I optimize parameters with values +10, -10 from the values of parameters given by the strategy.
Result: a) Check for strategy consistency. If it is not successful, I start from point 1 in a week or two, waiting for more stable strategies. And also, work with the repository, using other currency instruments and timeframes, thereby increasing the stability of its strategies. (I mean, the more people use the repository, the more reliable its strategies become?
b) on the last forward test, I select those parameters that have shown the best results according to step 4.
Maybe we should simplify the forward test: optimize 2008.01.01-2009.01.01 + forward test 2009.01.01-today, and then select the parameters according to p.4? Is there any sense in carrying out such a complex forward analysis, if the strategy is already on the 1-5th place in the repository and its soundness has been practically proved?
6. If the strategy is successful, check the strategy and its parameters on the demo account.
7. If the demo is okay, add stop loss, trailing stop, money management to the strategy. Optimizing the values of these variables individually, without changing the indicator parameters. Selecting values of stop loss, trailing stop, money management according to the indicators in point 4.
8. Real trading, if demo account test is passed.
9. monitor real trade results checking profit and loss indices against the test trade from the last forward test. If there are discrepancies, we start from step 1, as the repository may already have more stable strategies.