German IFO Days

I have been analysing historical data (from 2010 to 2020) on IFO days.
Below I mention the best results, only based on the direction of the 10:00-10:30 bar and additional the IB direction.
First, I looked at setups with entry time 10:30 (30 minutes after IFO announcement):


The green marked setup seems to be interesting.

Next, I looked at setups with entry times 10:30, 11:00, 11:30 and 12:00:


Some have a nice win rate, but in general the risk-reward doesn’t look good.
To decrease risk/exposure it might be worth to choose for instance a lower SL or a shorter trading timeframe with a slightly lower result.

When adding other filters (like with the daily trades), it will give different outcomes. For instance, looking at the length of the 10:00-10:30 bar will have a positive effect on the results:


There are several ways to drill down on the data/results, but here I just wanted to show you the outcome of this analyses on IFO days

Hello Deetee,

if you add here and there filters and filter out the very thin data basis even more, you are making cherry picking. This will not result in a stable system. You should split your data basis in a training- and test set and compare the results with fixed rules. You should also measure the size distribution of the trade results. If you make any good looking "profit per trade" numbers with only very few trades, these could be outlier which skews your results. But these are the basics of backtesting which were already well described here at futuresio.

Regards,
Koepisch

Deetee 

German IFO Days I have been analysing historical data (from 2010 to 2020) on IFO days. Below I mention the best results, only based on the direction of the 10:00-10:30 bar and additional the IB direction. First, I looked at setups with entry time 10:30 (30 minutes after IFO announcement): The green marked setup seems to be interesting. Next, I looked at setups with entry times 10:30, 11:00, 11:30 and 12:00: Some have a nice win rate, but in general the risk-reward doesn’t look good. To decrease risk/exposure it might be worth to choose for instance a lower SL or a shorter trading timeframe with a slightly lower result. When adding other filters (like with the daily trades), it will give different outcomes. For instance, looking at the length of the 10:00-10:30 bar will have a positive effect on the results: There are several ways to drill down on the data/results, but here I just wanted to show you the outcome of this analyses on IFO days

Koepisch 

Hello Deetee, if you add here and there filters and filter out the very thin data basis even more, you are making cherry picking. This will not result in a stable system. You should split your data basis in a training- and test set and compare the results with fixed rules. You should also measure the size distribution of the trade results. If you make any good looking "profit per trade" numbers with only very few trades, these could be outlier which skews your results. But these are the basics of backtesting which were already well described here at futuresio. Regards, Koepisch

Thank you. I am looking help, so your feedback is appreciated.

I am aware of the thin data basis, and that with filtering down the selection becomes cherry picking. I wasn’t sure about my approach. About when the selection becomes cherry picking, about the population/sample size, and how to approach this data in finding setups. I have been searching on how to handle statistics (without the answers I was looking for), but not on back testing; will do. Is there perhaps a good thread you can recommend?

For my understanding, for example, taking the IFO case above, the bottom setup filters the IFO days, 10:00-10:30 bar green within a (length) range, and triggers a long trade from 11:30-14:30 with a SL of 20pts. Can’t this be seen as fixed rules? I guess not, but is that because only 19 trades remained?

Thanks again,
Deetee

Tuesday

In the Tuesday files, when I add filters on IBS, ‘Up gap & gap closed’ , IB 9:30 position ‘Inside’ the results look good but conflicting.

For a short trade the best result is 445 pts (12 win/9 loss), but this includes a large win of 353 pts.
For a long trade the best result is 466 pts (12 win/9 loss), this include large winners of 222 + 146 pts.

So, no trade today




edit 24-2-21 7:50: I just now notice that the short trade was for 9:30-11:00 (SL50) and the long trade for 11:00-15:30 (SL40). The short would have given 142 pts profit and the long 40 pts loss

Wednesday

This morning I noticed that yesterday the trading times of the short and the long trades were not conflicting (as the trading times were not overlapping) and would have given a nice profit (102 pts). I was a bit in a rush and didn’t do the preparation work properly. Lesson learned.

In the Wednesday files, when I add filters on IBL, ‘No gap’, IB 9:30 position ‘Inside’ and Prior Day Hi/Lo on extreme, the results are not the best but the win rates are OK.

For a short trade the best result is only 136 pts (9 win/3 loss), from 11:30 – 12:00 SL20.
For a long trade the best result is 186 pts (9 win/3 loss), from 12:00 – 14:00 SL45.

I will take both trades, only for the long trade using a 20 pts stop (178 pts, 9 win/3 loss) to have a better risk/reward.

DAX Short trade
Entry 11:30 @ 13971
Exit 12:00 @ 13991 (SL, was exactly the high of this 30min bar. I traded the CFD IBDE30 and escaped from the SL and ended with 1 pts loss)
SL 20pts
Result - 20 pts

Historical results with this setup:
136 pts (9 win/3 loss)

= = = =

DAX Long trade
Entry 12:00 @ 13974 (not the best place to go long)
Exit 14:00 @
SL 20pts
Result - 20 pts

Historical results with this setup:
178 pts (9 win/3 loss)

Let's see



edit 17:48 Well, at least I'm creating some statistics. Hopefully better tomorrow

Thursday

In the Thursday files, when I add filters on IBS, ‘Up gap & closed’, IB 9:30 position ‘Above’ it turns out that there were only 7 occurrences (out of 269 Thursday IBS days in 2010 to 2020). I’m taking the ‘Above’ as the IB closed near prior day’s high, which then gives 39 occurrences.

For a short trade the best result is 574 pts (18 win/21 loss), from 9:30 – 15:00 SL30.
For a long trade the best result is 461 pts (24 win/15 loss), from 11:00 – 12:30 SL25.

The win rate should at least be 1 so the short trade is off. I will take the long trade, also because when I filter it down with prior day OC and the day before, results still look good.

DAX Long trade
Entry 11:00 @ 13971
Exit 12:30 @ 13946 (SL)
SL 25pts
Result - 25 pts

Historical results with this setup:
461 pts (24 win/15 loss)

Let's see



edit 12:30 So far, not so good. Anyway, I'm learning

Friday (last day of the month)

Good morning, today it is the last trading day of the month, which I analysed separately.
There are 84 occurrences between 2010 and 2020. So, the sample size is small. When adding Friday and IBL to the filters only 7 occurrences remain… Too thin for a trade.

Anyway, just for the record:
For a short trade the best result is 112 pts (4 win/3 loss), from 11:00 – 12:00 SL20.
For a long trade the best result is 153 pts (5 win/2 loss), from 12:00 – 14:00 SL45.

No trade today

Results

The results:


6 trades, 6 losses. This journal might looks like a joke, but it's a real thing
The only winning day (2 trades, 102pts, last Tuesday) I missed by 'mistake'

Basicly I see 10 setups (see table above) and 6 of them lost their first trade. I would like to think 'you win some, you lose some', but 6 losses out of 6 (or 7 losses out of 8 if I include last Tuesday) doesn't sound good.

Please comment on the approach as I can use some help to understand if this approach is a waste of time or not.

Anyway, have a good weekend and Monday another try.
Cheers,
Deetee

Monday

In the Monday files, when I add filters on IBL, ‘Up gap & not closed’ and a Prior Day and the day before open/close ranges, the short trades give the better results.

I picked this setup because the win rate is good, and because on a higher level the results were ok as well:

DAX Short trade
Entry 12:00 @ 13937
Exit 15:00 @ 13914
SL 20 pts
Result 23 pts

Historical results with this setup:
253 pts (9 win/2 loss)

Let's see

Tuesday

In the Tuesday files, when I add filters on IBL, ‘No gap’ and on Prior Day and the day before open/close ranges, the long trades give the better results.

For a long trade the best result is 342 pts (19 win/6 loss), from 10:30 – 15:30 SL35.
For a short trade the best result is 215 pts (4 win/12 loss), from 09:30 – 15:30 SL10.

DAX Long trade
Entry 10:30 @ 13997
Exit 15:30 @ 14091
SL 35 pts
Result 94 pts

Historical results with this setup:
342 pts (19 win/6 loss)

Let's see



edit 15:33: A good winner