Strategy: Kurosawa

Type: Trend / Momentum / Breakout

Based on the analysis of the price and volatility of the asset, the strategy creates a dynamic trading corridor. It consists of a main channel and high volatility bubbles. As soon as the price goes out of the main channel and is inside the high volatility bubble, the strategy sends a basic signal to enter (going up outside the channel - long, going down - short). Next, the base signal is filtered - it is checked for its compliance with the short-term trend and the presence of sufficient momentum.

If the filtering is passed, the strategy starts to cautiously gain a position as long as the market keeps the momentum, as well as places dynamic TP and SL orders related to volatility. When entering a position, the strategy assumes that any momentum degrades rather quickly, so TP orders that initially try to trailing the movement will eventually switch to a quick exit mode.

Strategy: ARMMA

Type: Range / Reversal

Based on the analysis of the price, volume and volatility of the asset, the strategy creates a dynamic trading corridor. As soon as the price with a sharp impulse leaves the main corridor and then returns to it, the strategy receives a basic entry signal (upward exit outside the channel - short, downward exit - long). At the same time, the strategy checks the absence of a trend in the direction of the impulse to confirm the reversal signal.

If the entry criteria are met, the strategy opens the position and places a dynamic SL order. If the outcome is favourable, the trade is closed upon reaching the inverse side of the corridor, but the SL order constantly follows the price and provides a quick exit from trades in the absence of a strong reversal movement.

Synergy of strategies

As you can see in the image below, our algorithms support each other by covering different market phases. The Kurosawa strategy is adapted for strong impulsive movements and works on short-term trends, while the ARMMA strategy, on the contrary, tries to avoid trends, working in a calm market with an established trading range.

Optimisation system

Both strategies are connected to the optimisation system, which allows to automatically prepare strategies for launch on new coins, as well as quickly re-optimise them in case of performance degradation on already running assets. Currently most strategy parameters are optimised using a combination of annealing and multi-gradient descent.

Risk management

We proceed from the concept that any predictive model for a particular asset can suddenly degrade. The solution to this problem is broad diversification - If we use not one but 30 models capable of adapting to different market conditions, the probability of successful operation of the system as a whole is significantly increased. Degradation of one model ceases to affect the productivity of the system as a whole, and quick optimisation mechanisms allow easy replacement of “corrupted” elements.

Both strategies operate in parallel on each selected asset. Their basic logic does not allow positions to be opened in the same direction, as a result more than 30% of total trading time is spent on an equal ratio of short and long positions across all assets.