Fine-tuning our algorithms

As you may have noticed already, we are not resting this summer but are working hard to polish our algorithms. Below you can find the main improvements to our algos introduced this week.

1. Minor bugs (Levels & Performance)

First of all, we have removed minor bugs that were reported by some of our traders.

As you may recall, there are different paths to level up in Darwinex (click here for further info on the different paths). In certain cases, the algorithms were not picking the “right” path to level up. We have found the root of this issue and have fixed it so it won’t happen in the future.

Further, we were reported that the Performance attribute included a minor gap when certain strategies reached 6 D-Periods experience. This has now been fixed, too.

2. VaR Simulations

We are working hard to introduce a major improvement in the way our risk manager replicates trading strategies on behalf of investors. The first step towards that goal has been introduced this week: VaR simulations have been improved significantly. The way we were simulating VaR in the past used to have an effect on the D-Periods’ calculation. This used to cause a delay in the accounts’ analysis that has now been fixed. VaR simulations are now independent from D-Periods, which makes it easier (and faster!) for our algorithms to analyse trading strategies.

This is only the first step in a series of substantial improvements to our risk manager that are underway, we’ll keep you duly posted!

3. Consistency Attribute & swing trading

The main improvement introduced this week relates to the Consistency Attribute.

We have received several complaints from swing traders who felt penalised by the way the Consistency Attribute addressed their trades. In a way, achieving good scores in the Consistency Attribute used to be harder for traders who trade in longer timeframes (e.g. underlying strategies of DARWINS TSB or PGH). With this in mind, we have introduced a major change in the way weight is assigned to each trade in the Return Consistency & Duration Consistency charts.

Snip20160817_11

In the past, Weight used to be calculated as follows:  Weight = Trade’s Leverage* VaR(20%) / VaR of the strategy at trade’s close.

Going fwd, the weight assigned to each trade will contain a correction that takes into account the trade’s duration. The new weight is now calculated as follows: Weight = Leverage * (Square root of trade’s duration) * VaR(20%) / VaR of the strategy at trade’s close.

This improvement has given rise to changes in our traders’ Consistency Attribute score (in general, our community’s score in this Attribute has improved with the amendment introduced).

As a rule of thumb, traders who trade shorter timeframes and keep a specific trade open for longer time than usual will be more penalised on their Consistency Attribute than in the past. On the other hand, traders who trade longer timeframes and close a trade earlier than they usually do won’t be as penalised as much as they used to.

Further improvements to the Consistency Attribute are in the pipeline. In the coming months, we’ll introduce a big improvement in the way “hedging” is assessed by the Consistency Attribute.

4. Upcoming Improvements

Further improvements to our algorithms are underway. In addition to the changes mentioned above,  the main improvement we are working on will affect the risk manager: the way it works today, our risk manager is being too conservative when replicating strategies that trade very frequently and keep several trades open all the time (e.g. QNR). We’ll keep working hard to have this fixed ASAP, we’ll keep you duly posted.

As always, please do not hesitate to contact us at info@darwinex.com should you need assistance. Please accept our apologies for the effects this week’s changes may have caused in DarwinIA’s standings, we promise we do our best to avoid changes halfway of the month!

 

Loading Facebook Comments ...
2 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *