Finally, we have completed a test version of Orbit v2. We spent several weeks on Orbit v2 testing to study features of it in many aspects such as performance, risk, trading frequency, win ratio, and so on. The test results taught us that Orbit v2 is good to go for the next round. Its performance is above the acceptance level, and MDD is negligible. The MDD convinced us that we could start testing in the real market because we could bear the loss even in worst case.
After a few testings for error checking , we will let the Orbit running on the real market; we will not do anything with Orbit unless there is an error but just monitoring. Orbit and Athena were designed to do its job without human intervention.
Athena
Orbit v2 Trainer
- We implemented a distributed trainer for Orbit v2. Orbit v2’s core logic is coming from machine learning. At a certain point, we need to update Orbit v2’s machine learning model to teach the market trend.
- Due to the trainer, the whole training process could be automated. What we have to do to train ML models is to execute the trainer. Then the trainer will handle everything from beginning to the end. The current version of trainer is not matured yet, but it makes our life much easier.
Bet Sizer
- Bet sizer is another critical component in Athena because it controls the size of the investment; how much money should be invested. If the expected outcome is big enough, then betting big money is acceptable. But risk versus gain is likely to be low, no bet or small bet. Bet sizer’s role is just like that. Without right bet sizing, it is hard to make a profit. Orbit v2 can not predict the future correctly all the time.
- It has an intelligent logic to determine the investment size by checking probability data, consecutive winning ratio, available fund size, etc.
- Bet Sizer and Orbit are the two critical components in Athena.
The below snapshot shows the test result of Orbit v2 conducted on Korea stock market from 2020-01-01 to 2020-10-05
- win_ratio : daily win ratio
- long_count : daily buy count
- short_count : daily sell count
Get early access to the latest updates on the MoA
Get early access to the latest updates on the MoA Investment Solutions
