The question is how intelligent is the Artificial Intelligence here. Is it discussed in the whitepaper? For me, it's just another iteration of a Poloniex trading bot.
Hey,
ePesoInitiative!
I understand that concern. We will publish the simulation results a bit later and it will be clear that our system is much more accurate than others.
The main difference of our system is that it is
modular and
scalable.
Crypto trading bots that exist today are: 1) Based on one
single strategy and implemented in a hard and static logic of
if ... then ... 2) Based
only on past prices history
3) Have a
hard-coded logic of money management and position sizing
Mirocana system is: 1) Based on hundreds of strategies (and some strategies has hundreds of sub-strategies) and it uses deep-leaning neural nets to calculate the weights among all strategies in the system to achieve a cumulative effect and create a "super-strategy" that is better than any individual strategy by its own. It allows to create the system is not static and is able to adjust to changing market behavior.
2) Based not only quotes. We have strategies in the system that is based on the trading activity of other traders, based on news (they use sentiment analysis), based on macro data, based on correlations of assets. We will also analyze Bitcoin and Ethereum transactions and will try to create the strategy that will be based on that. In total, we have 40+ sources of financial data.
3) Each simulation has a logic of money management and position sizing that is not static. It takes into account a number of investment opportunities and previous investing performance.
To achieve that we have created: 1) Our own syntax and framework for writing new strategies and backtesting them.
2) A modular system of data collection that allows us to store and collect data from multiple data sources in real-time.
3) Our own framework for simulations that allow us to go through the history and understand how the system would behave. We run multiple experiments of weight distribution and see what is performing better. Then we deploy best performing simulation into production.
To make Mirocana system more accurate we will: 1) Release
Mirocana Alpha platform where any interested algo-trader and quantitative analyst will be able to add his or her own strategy. If this strategy actually finds any patterns in data he or she will be rewarded with our tokens.
2) Release
Mirocana Target platform where any person will be able to leave his or her future price forecasts and if these forecasts are accurate and consistent he or she will be paid with our tokens.
3) Release
Mirocana Sigma platform where any data scientist or mathematician will be able to participate in our predicting tournament and try to create a predictive model of weights distribution among our strategies. If this model is better than ours, he or she will be paid with our tokens.
We successfully manage our own fund using the current version of the system and want to make it accessible to those who will support our research.
Show Posts
Feel free to join and stay tuned, many great infos.
