Determine the future operability of the vehicle. - page 6

 
What does the spread have to do with it? The spread is not even a hindrance to the trader.
 
olltrad писал (а) >>
swaps change constantly? twice a year at least? and what kind of system makes profit based on swaps? or is it so "profitable" that it can't cover them?

<too few trades to analyse, the system must be profitable over the whole history


We are talking about the performance of the TS in the future. When optimizing the Expert Advisor on the semi-annual or annual history, it will last a week at most.

To identify the patterns, you need a longer history. If you test any of your EA on a longer history separately for long and short positions, you will see the difference.

When I look at the spreads, i am not going to waste my time, although there are some methods of fighting it.

 

For example, I would like to show the reports of a TC that is optimised for only 2 months, but has been running steadily for more than half a year. Although there is some belief that the TS should run 20% of the optimisation period. Why is this the case? I can't figure it out. Even the stats on OOS+real are better than on the optimization period. How do I know this in advance from reports?

All tests are performed with 0.1 lot.

Optimization period

Symbol EURUSD (Euro vs US Dollar)
Period 1 Hour (H1) 2007.11.01 00:00 - 2007.12.28 22:59 (2007.11.01 - 2007.12.31)
Model By open prices (only for Expert Advisors with explicit bar opening control)
Parameters
Bars in history 1958 Modelled ticks 2914 Simulation quality n/a
Chart mismatch errors 0
Initial deposit 10000.00
Net profit 971.67 Total profit 1796.22 Total loss -824.55
Profitability 2.18 Expected payoff 19.43
Absolute drawdown 43.00 Maximum drawdown 456.82 (4.22%) Relative drawdown 4.22% (456.82)
Total trades 50 Short positions (% win) 25 (40.00%) Long positions (% win) 25 (48.00%)
Profitable trades (% of all) 22 (44.00%) Loss trades (% of all) 28 (56.00%)
Largest profitable trade 309.13 losing transaction -98.00
Average profitable deal 81.65 Deal loss -29.45
Maximum number continuous wins (profit) 5 (375.18) Continuous losses (loss) 10 (-206.83)
Maximum Continuous Profit (number of wins) 464.59 (4) Continuous loss (number of losses) -253.52 (4)
Average continuous winnings 2 continuous loss 3

OOS+real

Symbol

EURUSD (Euro vs US Dollar)

Period 1 Hour (H1) 2008.01.02 10:00 - 2008.07.24 23:59 (2008.01.01 - 2008.07.25)
Model By open prices (only for Expert Advisors with explicit bar opening control)
Parameters
Bars in history 4489 Modelled ticks 7975 Simulation quality n/a
Chart mismatch errors 0
Initial deposit 10000.00
Net profit 3397.46 Total profit 6699.17 Total loss -3301.71
Profitability 2.03 Expected payoff 21.92
Absolute drawdown 154.52 Maximum drawdown 529.09 (4.07%) Relative drawdown 4.07% (529.09)
Total trades 155 Short positions (% win) 78 (60.26%) Long positions (% win) 77 (62.34%)
Profitable trades (% of all) 95 (61.29%) Loss trades (% of all) 60 (38.71%)
Largest profitable trade 341.70 losing deal -231.60
Average profitable deal 70.52 losing deal -55.03
Maximum number continuous wins (profit) 6 (354.60) Continuous losses (loss) 5 (-117.87)
Maximum Continuous Profit (number of wins) 499.52 (5) Continuous loss (number of losses) -375.88 (4)
Average continuous winnings 2 Continuous loss 2

And the most important question - well, how long will this TS last? )))

 

Leonid, the trouble is that the question you asked has no answer!

More precisely, the answer can only be given by the one who "manages" the flow of quotes :). Look: for example, you have created a TS, which until now has been more or less stable and brought you a profit, so we can assume that it uses some of the regularities that are present in the market at the moment. So, to be sure the system will "work" (bring profit) in the future, you can only know for sure, that these same patterns will be maintained in the future. But, unfortunately, virtually no one can give such guarantees. Well, to say that the system will last "so long", looking at the results of the optimization/training period and OoS is as "promising" as trying to predict Close!


So perhaps the question should be phrased a little differently: "What should be done to increase the probability of the TS working profitably in the future for as long as possible? ".

Or like this: "How to determine with the least risk for the deposit the moment of the end of the profitable operation of the TS? (the moment of overtraining/change in strategy)".

For example, if you have trained the network, got good results in the training/optimization section and immediately put it to the real account, then the probability of its successful operation, and even more so for a long time, is unlikely to be high.

Checking the network for OoS (with a positive result) increases this probability(but not to 1 at all!).


P.S.1 All IMHO.

P.S.2 I am looking for answers to these questions myself, but so far with mixed success.

 
AlGor писал (а) >>

For example, you have created a trading system that has more or less consistently made you a profit so far, so we can assume that it uses some of the patterns currently present in the market. So, to be sure the system will "work" (bring profit) in the future, you can only know for sure, that these same patterns will be maintained in the future. But, unfortunately, almost no one can give such guarantees.

Therefore, perhaps the question should be phrased a little differently: "What must be done to increase the likelihood of profitable operation of the TS in the future for as long as possible? ".

The answer to your question "what should I do..." you yourself gave in your post just above: "know for sure that these patterns will remain in the future". Well, to do that, following your own lead, you need to become the one who "steers" the flow of quotes. A very rational thought.

And everything else, judging by the fact that the answer is "only the one", is hardly helpful.

:-)))

 

'Trading sessions or how important time is'

There is a tester report at the end of the page

take a look at it and tell us your opinion

I "managed" to write a system without indicators in summer of that year

it still "works" today without readjustment of parameters

but I've already taken it off the micro-real

 
Korey писал (а) >>
If you take 100% as a tester, the demo will be 90%, realisation down to 50%

>> Why would you do that?

 
Sometimes it is the other way round
 
LeoV писал (а) >>

For example, I would like to show the reports of a TS which is optimised for only 2 months, but has been running steadily for more than half a year. Although there is some belief that the TS should run 20% of the optimisation period. Why is this the case? I can't figure it out. Even the stats on OOS+real are better than on the optimization period. How do I know this in advance from reports?

All tests by lot 0.1

...

And the main question - well, how long this TS will work further? )))

Why don't you want to run it on 1999-2007 history (OOS = entire period before optimization)?

Because the market has changed since then?

So relative to this TS it means 9 years of changing market (albeit taken from the past and not from the future, but it hasn't seen that market anyway).

I suggest that TS with opt=8 month may be more stable, as a confirmation - the larger profit for 9 years.

TS with opt=2m may turn out to be more profitable in the short term (only in the near future). But it will need to be re-optimized more often (for 9 years it is likely to be minus).

 

Yurixx, that's not what I meant...

The point of my post was this: it is impossible to say for sure (100%) that the system will or will not work in the future, much less specify the exact duration of its success. Yes, to know that the patterns found will remain 100% accurate in the future, to know Close at least one bar forward or to "manage" quotes - is great, but we, mere mortals, do not get it.

But it is possible and necessary to try to increase the probability of profitable operation of a TS or to try to timely determine the moment when we should overtrain/overoptimize the TS.