Change from fixed VaR to variable VaR in the DARWIN assets
On January 11th 2020 we will be making changes in the way in which the Darwinex risk manager works.
Here we will explain:
- The reason for the change
- The operation of the new manager
- The effects achieved
- The implementation of the change
- Appendix: Examples of popular DARWINs
1. What the reason for the change is
The risk manager is in charge of ensuring that all DARWIN assets trade with a known risk, regardless of the risk taken on by the underlying strategy on which the logic for market entry and exit decisions is based.
This means that the DARWIN replicates the asset and the timing of the trader’s operation, but investor volume is defined by Darwinex through the risk management logic. The manager’s involvement thus makes the respective returns of the underlying strategy and of the DARWIN different.
As of today, all DARWINs have the same target risk – 10% monthly VaR. Setting it at a given value and not allowing it to be variable has a series of benefits.
- It makes it possible to compare DARWIN returns in a very intuitive way.
- The investor knows the risk of any DARWIN portfolio composition at any given time.
- It counters the bias of many traders who, for emotional reasons, are unable to take on significant investment amounts and, to protect themselves, lower their strategy risk, even though there is no evidence that their predictive capacity has worsened.
- It removes the statistical illusion of trading strategies whose performance is mainly based on the gradual use of leverage, such as more aggressive martingales, which are so widespread in copytrading.
- It makes it easier to estimate the maximum capacity that a DARWIN can absorb, that is, the maximum investment that it can manage without impairing its investors’ returns too much.
- It makes it possible to apply management fees adjusted to the potential returns of the DARWIN in question.
However, in addition to these features, the risk manager should be as least “intrusive” in a trader’s strategy as possible, so that the trader does not feel that the risk manager is harming their investors. Otherwise, the best managers will be unwilling to provide their systems to our market.
In the past, we received many complaints along these lines. Many of them were baseless, or only came when the trader was harmed, but we were not thanked when it was the other way round (that’s the human condition).
Obviously, we must continue to improve the risk manager to make it less “intrusive”, but it’s hard to establish what that means.
After compiling feedback, we have reached the conclusion that many traders do not always trade with the same risk. There may be long periods (even 1 year long, but usually 1-6 months) in which the trader, because they believe that their system will not perform well due to market conditions, because they are making adjustments, or for any other reason, lowers their risk.
If these periods in which traders lower their risk last more than 45 days (the benchmark period used to calculate the underlying strategy VaR), at the end they significantly lower the benchmark VaR calculated by the platform. Then the DARWIN multiplies the trader’s leverage by a higher factor than usual, and consequently there is a sudden loss of proportionality between the underlying strategy and the DARWIN.
If the trader was right and their strategy does not perform well during that period, the DARWIN will lose significantly more than the underlying strategy with respect to its usual performance (if the trader was not right, the opposite would be the case, obviously).
Most traders are UNABLE to predict the market. Consequently, lowering their risk usually generates a random performance from the DARWIN. The problem is that it is precisely the best traders who ARE ABLE to adjust their risk to market conditions.
We have reached the conclusion that the risk manager should cease to have a fixed VaR to adjust better to the best traders when, for whatever reason, they believe that it is better to lower their strategy risk for reasonably short periods (less than 6 months).
Having said this, risk should fall within a range – otherwise, we would cease to comply with the rest of reasons why we preferred to make it fixed.
2. The operation of the new manager
With this change, DARWIN risk becomes variable, between 5% and 10%.
To determine the target VaR, historical VaR data is taken into account, starting with the most recent data with a look-back window of maximum 6 months, until the ratio between maximum and minimum VaR is 2:1.
The current VaR then gets divided by the maximum VaR calculated before. Last, this ratio gets multiplied by 10%. The resulting VaR will move between 5% and 10%.
Should there be no 2:1 ratio in the last 6 months, the maximum in the last 6 months is taken into account.
Current VaR: 8%
Maximum VaR: 12% one month ago
Minimum VaR: 6% five months ago
Target VaR: (8%/12%)*10%=6.67%
Current VaR: 9%
Maximum VaR: 14% 2 months ago
Minimum VaR: 8% in the last 6 months
Target VaR: (9%/14%)*10%=6.43%
In summary, the Risk Manager tolerates changes in Var up to 2 times (rises or falls) with the aim to better adapt to the trader’s risk management.
2.2 Additional change
To protect investors from abnormally high leverage, there is an additional control measure that does not allow investors’ D-Leverage to exceed in any case the 20-15-15 threshold for 15 min-30 min-1 hour positions.
This restriction mainly affects short-term DARWINs, or DARWINs that use short-term leverage peaks.
This change also makes that the DARWIN and the underlying strategy look more “similar” in shape.
3. The effects achieved
DARWIN performance at times of drawdown is significantly improved, without penalising too much total returns when the DARWIN is recovered. That is, an improvement in the return/drawdown ratio is achieved, and above all, we achieve greater similarity in DARWIN performance with respect to the medium-term risk decisions of the underlying strategies.
To better understand the impact of the change in the risk manager on DARWINs, we will give a very illustrative example. This is LZB, backtested using the current and the new manager. In both cases, a backtest is used because we have no real data using the new manager, and so the comparison is better using the same trading data (backtest).
The example shows that the current risk manager is responsible for the strategy drawdown being hugely amplified in the DARWIN, and later the strategy recovery is not apparent to the same extent in the DARWIN. This is an obvious case in which the current Darwinex manager damaged the strategy.
By contrast, the new manager, by offering a lower risk (between 5% and 10%) is able to perfectly replicate the proportionality with the underlying strategy.
The change will be automatically implemented. Neither traders nor investors need do anything.
Investible attributes will continue to function, at least for the time being, as though the VaR were still fixed and 10%. This has implications for certain notes, but they may take a long time to change, and we have opted to make the risk manager our priority.
Past graphs and charts won’t be recalculated.
5. Appendix: More examples of popular DARWINs
For discussion and feedback on this change, we invite you to visit this community post.