This article is about a unique and powerful foreign Exchange trading signal delivery service using reinforcement learning AI that was launched in August 2019!
This is MAiMATE’s starting guid.
- Artificial Intelligence Continues to Learn from Failures and Successes
- AI in pursuit of optimal position management
- Rational and unique AI
- Continuous learning for a changing market
- Educate AI and reflect your thoughts.
- How to get started with MAiMATE
- 5. Building an AI Agent_3 – Selecting News
- 6. Creating an AI Agent_4 – Choosing an Education Policy
- 7. Creating an AI Agent_5 – Naming
- 8. wait for the birthday email!
- 9. get your buy and sell signals!
- Grow your own AI and aim for a higher rank by the time it becomes a real deal!
- 10. let’s see the score!
- 11. let’s educate !
- 12. Let’s see the rankings!
- how to recreate an agent.
- Fall 2020, actual trading!
- in the end
Artificial Intelligence Continues to Learn from Failures and Successes
MAiMATE allows users to create their own AI agent (artificial intelligence) by simply answering a few questions on their own.
The AI repeats a number of pseudo trades of the last three years of market prices over and over again according to the conditions you answered.
The AI, which has just started learning, does not even understand the concept of trading.
Therefore, at first, the AI randomly tries “Sell”, “Buy”, “Continue to Hold”, “Close”, etc.
And AI will learn to trade through repeated failures and successes.。
AI’s performance in the first 11 months after the start of the market, which also survives the turmoil
There were approximately 6,000 MAiMATE AI agents in operation as of June 30, 2020. The results report publishes monthly overall trade gains and losses since the AI was created, and for pseudo-trades from July 24, 2019, when it was released, the total gains and losses as of June 30, 2020 were positive for all currency pairs.
The “Foreign Exchange Composite Index” is a further average of the average total gains and losses for each currency pair.
A good result for a popular currency makes the overall result look better, so the index evaluates the overall performance of AI agents by eliminating the difference in the number of AI agents for each currency pair.
*There is no guarantee of future profits.
AI in pursuit of optimal position management
Not long ago, Google’s AlphaGo, an artificial intelligence that beat the top human players, made the news.
In the beginning, AlphaGo doesn’t even know the rules of Go, but it starts learning from a situation where it doesn’t even know the rules of Go, and by repeating thousands or tens of thousands of virtual games at super high speed, it learns tactics that can outperform humans.
In fact, MAiMATE’s artificial intelligence is also based on the reinforcement learning type algorithms that are famous for Alpha Go and other games.
And one of the key features of MAiMATE’s AI agents is that they learn “optimal position management”.
Very often, the so-called trading AI in the world is challenged with “accurate price prediction after a certain period of time”.
Instead of predicting future prices, which is difficult to find an answer to, we focus on what position we should have now.
Rational and unique AI
MAiMATE is an AI agent in the same setting, but with a different personality and different trading. If you have 10,000 AIs, they will all trade differently and uniquely. Just as twin brothers and sisters are similar but have different personalities. MAiMATE values the variety of personalities that each AI possesses in order to confront the ever-changing monster that is the market, where there is no single answer.
As evidence of the importance of individuality, some AI’s have worked hard to learn to trade, but have failed to find a satisfactory trading method.
No matter what you do, it doesn’t work, so those become “withdrawn” and extremely afraid to trade.
Some “combinations of information” can make it difficult to find good trading techniques, and a “reclusive type” can occur.
This is a testament to the cleverness of AI. It is a wise decision to do nothing rather than lose your master by trading without confidence.
Even these reclusive AIs have a quiet willingness to learn, so when they find a trading technique they can be confident in through continuous learning since birth, they are brave enough to venture out into the world of trading.
Continuous learning for a changing market
And an important feature of MAiMATE’s AI is its continuous learning capability.
It is extremely difficult for an algorithm to keep up with ever-changing market conditions. That’s why once a week after MAiMATE’s AI is created, all of them re-learn by incorporating their own trading results and the latest information.
In other words, the AI is designed to look back at your trades after it has been created and improve them as it incorporates new information. And in this day and age when unexpected things tend to happen, we believe that the “continuous learning” feature is essential for automated trading services.
Educate AI and reflect your thoughts.
The AI creator can externally influence the continuous learning process by “praising” or “scolding” the AI for its trades.
This is another great feature of MAiMATE, and it gives the AI creator’s thoughts on trading as a learning tool.
For example, let’s say you keep praising the settlement of a large profit.
The AI learns about this trade by looking for reasons to praise you again next time, as it receives a bonus reward from you in addition to the profit it made.
On the other hand, if you scold, this bonus reward is negative.
No matter how well the AI performs, if you continue to scold it, it will think that no matter what you do, it won’t be appreciated.
As a result, it will not make any trades.
Maybe it’s a bit like parenting, where it’s important to “praise with consistency” and “not scold too much”.
How to get started with MAiMATE
1. New Registration
All you need is an email address to register with MAiMATE.
2.Create an AI Agent – Login to Invast Securities My Page
When you click the “Create Agent” button, you will be asked to log in to your Invast Securities My Page.
By logging into your My Page, you can create an agent.
If you don’t have an account with Invast Securities, you can receive your My Page ID here.
3. Making an AI Agent_1 – Selecting a currency pair
The first step in the creation screen is to select the currency pair that the AI agent will trade.
Currently, there are three forex dollars, pound sterling and euro dollars.
4. Building an AI Agent_2 – Selecting Technical Indicators
Technical indicators are information to be trained by AI agents.
The AI selects from several parameters and inputs them as training material.
1) Moving average: Averages prices over a certain period of time and is an indicator that mainly measures the direction of prices (trend).
2) “Bollinger Bands”: A band line drawn by using the standard deviation of prices over a certain period of time around a moving average. When the trend and volatility (the magnitude of price fluctuation) are grasped, the indicator can be used to determine whether the market is oversold or overbought.
3) RSI: An indicator that tells us how overheated the market is (overbought or oversold).
5. Building an AI Agent_3 – Selecting News
News, like technical indicators, is information fed to agents.
The information in this section uses proprietary training information from Refinitive (formerly Thomson Reuters) news aggregation information, which we have improved upon.
- Market Conditions: Information related to the current market conditions, such as overbought and oversold conditions. 2.
- Market Analysts’ Market Forecasts: Information on which market analysts’ forecasts are more likely to be bought or sold, etc.
- Emotional Expressions on the Internet: Gathers information on emotional expressions, including screaming and ranting, on the Internet.
4．External Factors, such as Regulations and Other Factors”: This section collects information on external factors that affect foreign exchange, such as interest rates and regulations, although they do not appear in the news frequently.
6. Creating an AI Agent_4 – Choosing an Education Policy
This is not information to be learned, but rather the balance of the reward when the AI makes a profit or loss of the same amount.
The agent continues to learn and evolve in an attempt to maximize this reward.
For example, when a “return-focused” AI makes a profit, it is rewarded more than a “balanced” trade.
On the other hand, trades that result in a loss are rated more lenient and have fewer points to lose compared to “Balance”, so they trade without fear of loss.
Decide to train your preferred AI agent.
It’s the combination that determines the character.
The character is actually determined by the combination of these questions. It’s not random.
7. Creating an AI Agent_5 – Naming
Give the AI agent a name and you’re done creating it.
8. wait for the birthday email!
After you create it, wait until you are done learning.
When you are done, you will receive a birth email to your registered email address.
It will take approximately 7 hours.
9. get your buy and sell signals!
After the birth, you will receive an email report of trading signatures about once a day.
Grow your own AI and aim for a higher rank by the time it becomes a real deal!
Points will be awarded to trainers of top AI’s in the actual trade scheduled for fall 2020. We recommend creating and training AI’s for the top tier while you’re at it.
10. let’s see the score!
You can log in to MAiMATE to check your AI results.
There are five classes, [S] [A] [B] [C] [D].
The higher the total value of each value on the radar chart, the higher the class.
Each value in the radar chart is evaluated relative to the agents in MAiMATE.
The radar chart is not created immediately after the completion of the AI, but after the first signal is sent out. Please wait until then.
11. let’s educate !
Education was a good response to the five days of AI trading last week. Not like this! You can tell them that.
Just select the date and choose “Praise” or “Scold”.
You can change it as many times as you like up to the evaluation confirmation date.
At the evaluation confirmation date, your evaluation will be sent to you and you will be able to train them for the next five days.
MAiMATE does not incorporate “no possession” into the AI’s continuous learning, whether you praise or scold it. (Because it may not do anything.)
12. Let’s see the rankings!
You can also check the results of other AI agents.
They are listed in order of profit and loss for the year.
how to recreate an agent.
You can recreate the AI agent from the Config page.
The previous AI will be deleted.
You can delete your MAiMATE account itself by unsubscribing.
Fall 2020, actual trading!
MAiMATE is a service that allows you to receive trading signals from your AI via email and other means.
Now, by the fall of 2020, you’ll be able to actually trade using AI signals!
MAiMATE is now fully operational as a social Algo trading service.
We will update you within the blog post about the articles that will be added and more.
in the end
In addition to actual trading in the fall of 2020, we plan to introduce a variety of useful features for asset management in the future. Our goal is to create a service where users can learn to invest and achieve their dreams through MAiMATE, a service that brings together the wisdom of the market and allows them to conquer the great enemy that is the market. Thank you for your support.