With over 10+ years of Quant trading experience that based on rule trading. Instead of using hard-coded trading heuristics,
our system combine quant rule base & automatically learn the best strategy from the market data.
This not only saves our efforts in quantitative research, but also enables our model to think "out-of-the-box" - a key to consistently outperform the other market participants.
It was said that the best traders "long at the bottom and short at the top." By looking at the "fingerprints" in the market movement, our model can robustly tell whether we are approaching the top or close to the bottom.
In fact, our experiments show that one could achieve 80-500%* annual return under a moderate leverage by simply following the trend.
*Past performance has no future indication.
A typical drawback when trading large positions is the increasingly high liquidity cost due to the market impact. To alleviate this, our reinforcement learning-based trading agent tracks the order book in real-time and makes limit orders in a way that minimizes the price fluctuation, which is essential when trading derivatives.
The key concept in portfolio management is to spread the risk through diversification. Pursuing high return while controlling risk, however, is always challenging.
Our system inspects an asset's past performance with deep models and evaluates its potential. With this technology, we can steadily identify the winner 45-120%* of time while traditional techniques do barely better than random guesses.
*Accuracy depends on the setting and doesn't apply universally.
One of the core problems in the quantitative field is to provide math models for the market movement. Our deep generative models accurately explains the history and extends beyond it - a looking-glass into the future. This suggests superior hedging strategies as well as exclusive option trading possibilities.