Questions about Darwinex attributes

[IlIlIlIlI]

To avoid a thread for every single attribute, I try it with this new common thread.

My first question:
Does anybody have an explanation why the 'Monthly divergence & Latency' raised from -0.06 % to -0.08 % in the last two days (Thursday to Friday) while the divergence shown on the investor divergence panel shows green bars?

The long red bar was produced on Wednesday and should be part of the -0.06% calculation on Thursday.

 

[TrainerTXR]

To avoid a thread for every single attribute, I try it with this new common thread.

My first question:
Does anybody have an explanation why the 'Monthly divergence & Latency' raised from 0.06 % to 0.08 % in the last two days (Thursday to Friday) while the divergence shown on the investor divergence panel shows green bars?
View attachment 295490
The long red bar was produced on Wednesday and should be part of the 0.06% calculation on Thursday.

Green Bars are GOOD (=followers got better fills than you)
Increasing positive divergence is GODD as well = makes sence for me.

But when looking at ILR I saw, divergence DECREASE fom -0.06 to -0.08%.
Could be the case, if the SIZE of the following trades are zero (which is likely with your small amount of AuM in combination with high VaR.). If Size of the follower is zero - the 'green bar trades' do not count in the calculation.

 

[IlIlIlIlI]

If size is zero then there is no trade.
But you are right: they calculate divergence on zero size trades. There are a few (about 20-25% of the bars there)
(Edit):
Hmmm .. how can that be done if there was no trade on the investor account?

That is a zero size trade with no data on the left side - no trade - but +0.90 points divergence..

And this is another trade where the positive divergence (+0.8 pts.) is calculated down with the divergence VWApip (-0.8 pts.) which is not shown on the graph.

To say it short: I don't understand what they are calculating there. Can anybody help?

 

[IlIlIlIlI]

When are the attributes calculated?
Usually - I thought always - attributes are calculated after the end of a business day.

But not on Thursday 25th of February (1 business day before end of the month).
On my Darwin ILR there is a DD shown on the Pf attribute while the Darwin closed after an intraday DD with a new all time high.

The assumption is now that they took the data from the late afternoon to calculate Thursday's attributes.
The same is with the VaR calculation, I traded about half the volume in less than half time than average on Friday, but the VaR made a jump up on the Friday value. I assume it is the same reason: not all trades of Thursday are used for the Thursday value, obviously the later trades were taken into account for Friday.

Now I wonder if that was an exception or an error or because of the expected workload at the end of a month. I never saw that before.
It would be bad and unfair for some Traders if it had an impact on DarwinIA eligibility or evaluation.
Intraday calculations shown as daily calculation could also explain the divergence shown in my first question in this thread.

 

[CavaliereVerde]

Since the revision of DScore they recalculate it more often than before, maybe every 4 hours.
I dont' know if they recalculate investable attributes too...

I understand you concern about your darwin but by the way the probability to be allocated with a DScore lower than 40 is close to zero.
It is useless to focus on Darwinia with less than one year of trackrecord.

 

[IlIlIlIlI]

Since the revision of DScore they recalculate it more often than before, maybe every 4 hours.
I dont' know if they recalculate investable attributes too...

I understand you concern about your darwin but by the way the probability to be allocated with a DScore lower than 40 is close to zero.
It is useless to focus on Darwinia with less than one year of trackrecord.

That was not written because of my Darwin, I don't expect DarwinIA rewards before the 3rd quarter because of the D-Score metrics, it was written more for Darwins like FIU made by our friend @Forexintradiadarwin which is excluded with a correlation value of exactly 0.80 .

 

[CavaliereVerde]

Darwins like FIU made by @Forexintradiadarwin which is excluded with a correlation value of exactly 0.80 .

FIU was out even without correlation problems.

 

[IlIlIlIlI]

IMO the correlation calculation is strange and hard to understand. If you can understand it.
I can't see correlation on charts like this (ILR and its closes other Darwin), even if I try and the correlation value shown is 0.61:

 

[CavaliereVerde]

IMO the correlation calculation is strange and hard to understand.

IMO correlation is completely flawed.
Thy prefered to change regularity that was fine even before and they ignored correlation that has aways been BS.

 

[IlIlIlIlI]

I'm curious how divergence of today will be calculated over the weekend (screenshots from ILR):

I've never seen such a bad broker behaviour at Darwinex before (made by their CFD provider) for such a small investment amount.
If you like to guess I'd appreciate your ideas what will be shown on Monday behind "Monthly divergence & Latency".

 

[LongVision]

they ignored correlation that has aways been BS

Correlation between asset class very difficult to master.

Medallion Fund is the Greatest Money-Making Machine of All Time because of correlation mastery between asset class.
If there is any flaw in algo then the result will be disastrous(Remember Darwin HGF).

I agree, It is better to ignore the flawed darwin correlation.

 

[CavaliereVerde]

Ok but here we ware speaking about correlation between darwins.
Nothing will ever be perfect or predictive but instead of calculating on 3 weeks they could calculate it on 3 months.
What is the point of comparing the behaviour of 2 darwins for 3 weeks?
What is the significance of 3 weeks of trackrecord? IMO almost zero.
In 2017 they extended the lookback of var by a factor of 3, from 15 working days to 45 working days, they had to do the same for correlation.
It is not rocket science, for some unknown reason they decided to to let it unchanged.

 

[IlIlIlIlI]

What is the point of comparing the behaviour of 2 darwins for 3 weeks?

If one of the two Darwins does not trade, correlation is unchanged for a longer period, also if the other is trading and producing results in a defined direction.
The current calculation of the correlation is pure nonsense and abused to keep some Darwins randomly away from DarwinIA prizes if the D-Score is lower. The most relevant value of this D-Score is meanwhile the influence eligibility for DarwinIA.
With this DarwinIA stays a lottery to fool traders.

 

[CavaliereVerde]

It is not pure BS but it is 70% of what it should be like much else in Darwinex.
It is an unfinished job.
Darwinia would be a lottery even with perfect correlation and perfect DScore because it is based on the result of one month that is vastly random.
What is not so random is the performance of the following semester.
Also stacked allocation tend to be on good darwins.

 

[IlIlIlIlI]

Darwinia would be a lottery even with perfect correlation and perfect DScore because it is based on the result of one month that is vastly random.

I think that's wanted to give a lot of Darwins the chance to participate in DarwinIA.
At Darwinex there are a lot of blown off accounts which are still traded because of that.

What is not so random is the performance of the following semester.

That saves a lot of money 😄
The premium of the allocation should be paid by investors (fee and commision) for the better Darwins and by the traders themselves (commision) for the others.

 

[CavaliereVerde]

There is the other topic to discuss about Darwinia.
Going back to correlation, it was designad to exclude clones from Darwinia and in 90% of cases it works for that task.
It could work better with a longer lookback.
The real point is that with 3 weeks lookback correlation is completely useless to build uncorrelated portfolios, that should be the main purpose of correlation.

 

[CavaliereVerde]

Correlation algo has gone for a walk...

 

[IlIlIlIlI]

Correlation algo has gone for a walk...
View attachment 309479

They don't show correlation on ILR since months.
I'm tired to ask for repair again and again.

 

[CavaliereVerde]

Just noticed.
If you pick a darwin without investors correlation is not calculated.
for example I have correlation under LSC but not on CVL and NFO .

Probably Darwinex has gone to some kind of "energy saving " mode
"Useless" darwins are ignored in correlation that is a very consuming algo that have to create a matrix with all darwins.
a matrix of 300x300 is more sustainable that one of 3000x3000.