• Welcome to the Darwinex Forums, these forums are member-run and managed by CavaliereVerde. Member-run forum rules may differ from the site guidelines.

VaR and Risk Manager

I can see this problem very well.
LSC and NFO are exactly the same strategy portfolio, same tradesize and same equity.
LSC has a darwin VaR of 4.64%
NFO has a darwin VaR of 5.83%
That is private info but everyone can see DLeverage/position : 1.20 vs 1.65 .
The Rs algo would be much better to trigger a target VaR than the primitive logic used now.

I also don't understand why NFO does not have a target VaR of 6.5 as the max. and min. values of the VaR never had a worse ralation than 2:1 in the last 6 months as shown under the Rs attribute.
Do you know it?
 
I also don't understand why NFO does not have a target VaR of 6.5 as the max. and min. values of the VaR never had a worse ralation than 2:1 in the last 6 months as shown under the Rs attribute.
Do you know it?
Look to the Rs chart, max VaR on 6M is 5.32 , min has been 3.56 actual is 4.07
So 4.87 for the darwin makes perfect sense.
To have 6.5 you need to be at ATH on 6M.
You need min = max that is impossible to have in real world.
There are 6 darwins with Rs > 9.6 , all with supershort trackrecord.

Risk stability works perfectly, it is the best algo IMO , it is variable var the bullshit.
 
Last edited:
They also did not update all pages of their site, so I got the wrong examples. The correct ones are here:
https://help.darwinex.com/risk-manager
1621713501835.png

With theses examples I cannot understand what they could mean with the tolerance in their last sentence:
"In summary, the Risk Engine tolerates changes in VaR of up to 2 times (up or down) with the aim to better adapt to how the trader manages its risk. "

Long story short:
A good risk manager cuts losses more than profits, but this one does too often significantly the opposite because of that dubious logic of the target VaR.
 
Long story short:
A good risk manager cuts losses more than profits, but this one does too often significantly the opposite because of that dubious logic of the target VaR.
From what I remember, the reason for the introduction of the variable VaR was as a result of traders' trying to manage risk by reducing exposure in difficult markets. By doing this the VaR of the trading account reduced sharply causing the multiplication factor between the strategy and the associated Darwin to increase significantly, thus turning the Darwin into a wild uncontrollable thing!
Whether the solution has achieved the desired result is a matter for debate.
 
From what I remember, the reason for the introduction of the variable VaR was as a result of traders' trying to manage risk by reducing exposure in difficult markets. By doing this the VaR of the trading account reduced sharply causing the multiplication factor between the strategy and the associated Darwin to increase significantly, thus turning the Darwin into a wild uncontrollable thing!
Whether the solution has achieved the desired result is a matter for debate.
Originally they used 21 days for calculating the VaR, with the risk manager 2.0 they changed it to 45 days.
Then they added the target VaR looking into another 6 months in the past. As I'm still in discussion with Darwinex about it, they wrote on Friday "we have reached the conclusion that usually 1-6 months traders make adjustments or lowers their risk."
I assume they took the maximum period of this analysis to get all instead of looking for an optimized point and compare the results with different periods.
 
Flexibility was just an excuse to reduce risk.
Variable VaR + Pivot = deadly combo.
Now we have not enough volatility for retali investors and not enough alpha for institutionals.
The asset class is running at 4.8% VaR, nice spending review for Darwinia.
 
Flexibility was just an excuse to reduce risk.
Variable VaR + Pivot = deadly combo.
Now we have not enough volatility for retali investors and not enough alpha for institutionals.
The asset class is running at 4.8% VaR, nice spending review for Darwinia.
I agree with your points. The final nail in the coffin for me was the addition of the management fees. It makes it almost impossible to consider it as a longer term investment with all the fees associated with any return.
The swap cost is still high too. Charging swap for holding EURUSD short, whilst other brokers will pay you to hold the same position.
 
From what I remember, the reason for the introduction of the variable VaR was as a result of traders' trying to manage risk by reducing exposure in difficult markets. By doing this the VaR of the trading account reduced sharply causing the multiplication factor between the strategy and the associated Darwin to increase significantly, thus turning the Darwin into a wild uncontrollable thing!
Whether the solution has achieved the desired result is a matter for debate.
I think it is definitely been an improvement
 
I think it is definitely been an improvement
The idea about a variable VaR itself can make sense, but the parameters should be changed.
This was my last idea in the discussion with Darwinex - they will look into it:
"IMO the current width of the VaR channel shown under the Rs attribute with its logarithmic scale is much more significant to measure the stability of risk and VaR than the absolute values of the VaR of the last 6 months."
 
Exact , it is the perfect example of the recent simplicistic changes.
While risk manager is pretty sophisticated this additional risk reducer is naive, just like DScore depending only on past return and DD.
We were used to another quality standard.
A bit of flexibility could help but it should be logarithmic and bidirectional, like Rs and the underlying normalizer.
 
Friends,
I don't quite understand the point of showing the "return/risk" ratio SINCE INCEPTION, instead of PER UNIT OF TIME.
This way an older DARWIN always shows an advantage over a younger one, making a comparison impossible.
Your thoughts?
 
Hey AYC!
I disagree with you. Originally the Risk Return Ratio formula calculates Return divided by the Risk. It's fair to compare strategies to each other by their whole period of live. That is, taking their whole history. The indicator will be much trustworthy as long as the calculation period is longer.

Otherwise you could get biased conclusions. For instance: You can have an excelent Darwin with a long history that last year deteriorated for whatever market reason and a terrible young Darwin that started a year ago with a lucky strike. If you compare both by 'unit of time 1 year' you could get the conclusion that the bad young Darwin is much better than the excellent veteran Darwin, which would be a terrible mistake.

We could argue, to avoid that bias, that we could take the average of several annual periods (or unit times). Well, in that case we would be changing one indicator for another, not sure which one. Could be Sharpe ratio or Calmar ratio? And the battle is served. Which indicator is better?

Between two excellent Darwins. The older will have better indicators on the Darwinex platform: Better return/risk, better D-Score,.. That's perfectly fine. The veteran Darwin has proved to overcome more market regimes than a young one. I think time and number of trades are critical to evaluate a Darwin, among many other things. Of course.

Same as you, I have a young Darwin. But the only way to prove ourselves is by standing firms. That is resilience.

Personally, I like indicators that take whole history. Except for those who take absolute values as Return, or % Return. Well, in those cases a mediocre strategy can have a 50% return. That says nothing if it has taken 10 years to get it.
 
Good work of the risk manager can be seen in September 2022 on Darwin BQH.

While the trading account shows significant losses in September 2022, ....
1664728080930.png



the Darwin shows a profit of 7.55% in this month ...
1664728042778.png



and raised a DarwinIA allocation of more than 45k.
1664728007569.png


Congratulations to the trader - oder better to the risk manager? ;)
 
Top