Sunday, 10 February 2019

ELgos: Trading by the Numbers

We have just finished the first few days of beta testing of ELgorithmics and the new website. Thanks to the great feedback we have received from our beta testers, we identified some issues with the website, how we deliver the ELgos to users, and how users install the ELgos and start trading. 

As we are getting closer to going live, let’s talk about the probabilities behind the ELgos. 

Our goal when we create an ELgo is to have the highest probability of making a profitable trade on the next trading session.

When we data mine as part of the creation of our ELgos, our mining criterion includes a number of important outcomes. The most important of these are:
Win rate is over 50%
Average win is greater than the average loss

Generally, the majority of profits come from a minority of trades. When the market makes a large vertical move, something that happens less often, we can get large profits per contract. This is why we look for average win being greater than average loss.

However, the issue with algorithmic trading is, and always has been, the ability of the trader to sit through draw downs. The question has always been: "Is this a draw down or is the algo broken". Algos always eventually degrade. What we have done with ELgorithmics is take away that disadvantage of degradation by producing a new algo every trading day and relying on the win rate and win to loss ratio for profitability.

When we produce an ELgo, we also produce a new historic profile that each ELgo is based on. On the website, we will be posting the historical profile and the related data, such as win rate and average win.
Let’s look at an example historical profile for Gold (GC). This is similar to what you will see on the website for each ELgo.

Looking at the above historical profile we can see that GC had a 64.18% win rate and a win/ loss ratio of 1.789. In simpler terms, the ELgo has over a 64% chance of being a winning trade and if it is a winning trade, it is likely the number would be nearly twice as large as a win compared to if it was a loss ($420 v -$240)

The math behind the ELgos lead us to expect that the statistics from the historical profile should apply for the next trading day and if that is the case, we have an edge for the next trading day. 

If our ELgorithmic process continues to deliver historical profiled ELgos of the same quality then we believe that our methodology over time should be profitable although we are mindful of the so called 80-20 rule: 80% of profits come from 20% of the trades. 

Please read all the disclaimers including past profitability is no guarantee of future results.

No comments:

Post a Comment