How to use Evolize

Here’s a full step by step description of Evolize graph and what it consist of.

Agents

single-agent

Keywords like Artificial Intelligence, deep learning and predictions might help you guess, that we are using computer algorithms to predict future foreign exchange market movements. But if you look at the Evolize graph closely you can find yourself wondering where the data is coming from. Imagine a person, lets name him John, sitting at his desk, trying to predict tomorrow’s EURUSD highest point. After doing extensive 4 hour analysis John comes down with a prediction ‘EURUSD highest point will be 1.024 tomorrow!’. Let’s say John the analyst works at your company InsertYourName Ltd. You have hired him around 4 years ago and spent half a year testing if his predictions are correct and it occurred that John is 80% accurate. Would you trust his predictions while trading?

two-predictions

Now let’s say you also have another analyst named Peter. Peter’s only job at your company is to predict tomorrow’s lowest points. Like John, Peter is a good analyst (you wouldn’t hire him if he’s not good, right?) and his predictions are also 80% accurate.

If you trust both John and Peter, you already have Highest and Lowest tomorrow’s EURUSD price, with 80% accuracy for both. Is it enough for you to perform profitable trades?

Multiple agent system

At Evolize, we use multiple agents, except they are not humans, they are algorithms that designed to learn from the information they receive and predict possible future market movements. Our algorithmic agents are studying huge amount of information each night and after scientific research, which was published on 7th International Conference “Business and Management” in 2010 (you can read it here) it was found, that single agent have 65% accuracy while predicting 3 days to the future. Sixty five percent is not bad, but also, not that good, so scientists didn’t gave up and decided to try out “Ensemble method” where instead of using 1 agent for High and 1 agent for Low predictions, they use hundreds of them, while each of them had slightly different parameters (publication here).

To be precise, at Evolize we currently use 176 agents for high and the same amount for low predictions.

multiple-agents-high

multiple-agents-low

How to deal with so many predictions?

We will definitely need some math here. Let’s say we have 10 predictions, 5 for highest point and 5 for lowest.
High predictions:
1.024, 1.02, 1.027, 1.018, 1.02

Low predictions:
1.015, 1.012, 1.014, 1.018, 1.010
histogram-agents
Let’s group our predictions on the scale from 1.030 to 1.010 (Illustration above). Can you make an informative decision based on such a graph? Well you could but it would waste you a lot of time, i guess.

At Evolize we use Freedman-Diaconis rule to place all predictions into Bins of Histogram so that the predictions would be more readable. Since we don’t know which of the predictions is the most accurate, we treat them as they have equal probability each.

histo-agents-actual-transition

This way the more predictions Bin on the histogram has – the bigger the probability of reaching certain price. As I’ve mentioned before, data analysis is very difficult and uses a lot of time and resources. To be precise, it takes 6 hours to count 4 currencies on 32 processors of 1.8GHz. This way today’s prediction at each time zone (currently London UTC and New York UTC-5) we get at 6am. Most of the markets are still closed at his time so we still can place our trades with such predictions.

More details into graph

histo-candles-mix

Evolize graph consists of not only prediction data, but current market prices also, so we could see which of the points are already reached today. In our example, the highest point today was lower than 1.025, but bigger than 1.020. Lowest point in this graph have reached 1.015, but never went any further. Looking at this graph, now it makes a bit more sense, right?

What still missing are probabilities! We can easily calculate it. Here’s an example of how it’s done:

hist-100-percent

Now we need to separate all of the reached predictions from others.

hist-percent-all

But what does the percent mean? You can see it as probability to reach certain point.

hist-percent-final

And we also can have row by row predictions for each “Bin”:

hist-probabilities-by-line

Hope, that now you are ready to get some profits while using Evolize! If you still have some questions take a look at our FAQ section, there are plenty of valuable tips there. And don’t forget to hedge. Consider using one of our Forex trading strategies, or come up with one yourself!

And here is how our actual graph looks like:

2016-11-29

At 7AM London time (UTC)
gbpusd-11-29-7-00

Graph at the same day 10AM London time (UTC)

gbpusd-11-29-10-00

The same graph at the end of the day

gbpusd-11-29-24-00