Libraries: BestInterval - page 6

[Mikola_2]

fxsaber:

Cosmetics only - more info in the log.

A small bug:

[fxsaber]

Mikola_2:

A small joint:

This is not a joint, but the correct result. This is how much (albeit a negative value) was added to the profit, compared to the previous variant.

These percentages are how much profit was gained by throwing out one more interval.

[Mikola_2]

fxsaber:

Applying the best interval will always improve the result and show a positive profit.

But it should be realised that if the TS is not profitable, BestInterval will only give a good fit.

Even for profitable TS the rule works: the more intervals are thrown out, the more likely the fit is.

That is why I usually throw out no more than two intervals. And still, the library was created for Optimisation.

The description gives a graph with OOS for a reason. Don't be fooled with the fitting.

The library doesn't create a miracle, though it is a must-have for me personally - I use it 100% of the time.

I solve two tasks through it:

1. Not to accidentally throw out a potentially profitable TS.
2. To set up a single result for combat application, exposing the guts of intervals.

I lack a weekly (not daily) analogue...

ZY If the EA is not MT4-style, the library will use 90% of its capabilities. And intervals will be obtained. However, you won't be able to see their application in the Tester at once. The programmer will have to invent something big for this purpose. That's why only MT4-style can reveal 100% of biblical possibilities.

Absolutely agree! A profitable TS is primary. I realise this and have no illusions. :)

Such a question: if I calculated the intervals and then changed the testing range in the Tester (to make sure that I found a pattern and not a fitting), will the intervals be correctly applied to the new testing range?

[Mikola_2]

fxsaber:

This is not a joint, but the correct result. Exactly that much (albeit a negative value)

That's what I'm talking about minus.

[fxsaber]

Mikola_2:

if I calculated intervals and then changed the test range in Tester (to make sure a pattern was found and not a fit), will the intervals apply correctly on the new test range?

Yes.

2018.10.12 23:59:59   BestInterval Action(true - single pass & MT4-style is required) = true
2018.10.12 23:59:59   Calculation time activated intervals is 2018.10.18 10:12:46 - TesterEA (common folder) 00:01:02 ago.
2018.10.12 23:59:59   
2018.10.12 23:59:59   Amount of Delete Intervals = 2
2018.10.12 23:59:59   00:00:00 - 02:15:22 : Profit = 1281.01 (26.00%), Total = 127 (83.46%), PF = 1.70, Mean = 10.09, DD = 600.66, RF = 2.13
2018.10.12 23:59:59   02:32:07 - 05:42:32 : Profit = 1896.41 (38.49%), Total = 262 (81.68%), PF = 1.44, Mean = 7.24, DD = 993.82, RF = 1.91
2018.10.12 23:59:59   21:53:23 - 23:59:59 : Profit = 1750.17 (35.52%), Total = 101 (79.21%), PF = 2.18, Mean = 17.33, DD = 463.34, RF = 3.78
2018.10.12 23:59:59   SUMMARY: 00:00:00 - 23:59:59 : Profit = 4927.59 (100.00%), Total = 490 (81.63%), PF = 1.65, Mean = 10.06
2018.10.12 23:59:59   
2018.10.12 23:59:59   final balance - InitBalance (100000.00) + Profit (4927.59) with BestInterval.
2018.10.12 23:59:59   OnTester - Virtual InitBalance (100000.00) + Profit (-8563.00) without BestInterval. Profit is calculated with TickValue=1 and w/o Commission+Swap.
final balance 104927.59 USD
OnTester result 91437

The highlighted line shows when the applied interval was calculated. The time intervals shown below will be applied. Their other data (profit, PF, etc.) are always shown only for the initial Action = false. That is, when you change the testing interval Action =true, only the time intervals should be viewed.

If you have increased the testing interval, the difference between the red values will immediately show how much the profit has increased/decreased. If there is no adjustment, then, of course, the value should increase.

[fxsaber]

Mikola_2:

That's what I'm saying about the minus.

Minus is correct, mathematically speaking.

[Mikola_2]

fxsaber:

Yes.

The highlighted line shows when the applied interval was calculated. The time intervals shown below will be applied. The rest of their data (profit, PF, etc.) are always shown only for the original Action = false. That is, when changing the testing interval Action =true, only the time intervals should be viewed.

If you have increased the testing interval, the difference between the red values will immediately show how much the profit has increased/decreased. If there is no adjustment, then, of course, the value should increase.

Very good. Thanks for the clarification.

[Mikola_2]

fxsaber:

Minus is correct, mathematically speaking.

Well, that's fine... Excessive perfectionism is evil! (I'm talking about me).

Everything is rubbish except bees... and bees are rubbish too... :))))

In any case - thank you very much for, not even a fishing rod, but for a gorgeous fishing kit!

And for the additional incentive to master MT5.

[Aleksey Panfilov]

fxsaber:

I'm solving two problems through it:

1. Not to accidentally throw away a potentially profitable TS.
2. To set up a single result for combat application, exposing the guts of intervals.

The weekly (not daily) analogue is missing...

And how does a week, fundamentally, differ from a day, if we introduce (designate) the first hour of the week.

Of course, if we are talking about algorithms for detecting intervals within a day and within a week.

         if(
            ..... 
           && ((New_Time[0]-3600*    Start_Hour)        % 86400    < 3600*Period_Hour)         // обусловлено интервалом внутри суток
           && ((New_Time[0]-3600*(96+ Start_Weekly_Hour))%(86400*7) < 3600*Period_Weekly_Hour)  // обусловлено интервалом внутри недели (неделя 168 часов)

           )

[...]

An example of an EA that could not be optimised, perhaps all the results are dropped because in martin-like systems the drawdown usually exceeds the profit?