Eclipse Technologies

The objective of any trading strategy is to differentiate itself from the market in terms of volatility, return, or diversity. Today, I will focus on strategies that aim to outperform the market’s return while reducing volatility.

The conventional approach to strategy development involves analyzing price movements using data and selecting statistical or fundamental factors that, on average, yield positive expected value in trades. Ideally, a perfect strategy would enable shorting at the market’s peak and going long at the bottom of every trough within a specific timeframe, such as 5 minutes. However, achieving such perfection is impossible due to the requirement of flawless information encompassing all market influences at any given moment. While delving into the reasons behind the impossibility of perfect information is beyond the scope of this discussion, it’s important to note that even with perfect information, predicting price movements accurately is challenging since some information is subjective and doesn’t solely lean towards bullish or bearish sentiments.

Keeping this in mind, we can pursue one of two approaches. The first approach involves seeking insights from external information sources to predict the bias of buyers and sellers before it is reflected in market prices. The second approach is to utilize real-time buying/selling information (order flow) to make decisions based on the current market consensus. Since the market is always “correct,” having information that doesn’t align with the market’s behavior (inefficient information) isn’t necessarily valuable. Following this line of thought, two ideas emerge for the second approach. Firstly, we can identify the factors that the market “cares” about and solely utilize them to predict price movements. However, this approach has drawbacks as it may omit certain information, and the market’s concerns can vary depending on various factors. For instance, interest rates may hold more significance in one week compared to a week where the risk of war is high, and interest rate concerns take a backseat. Implementing this approach is challenging due to the interconnectedness and overlapping of data, which would lead to over-sensitivity in the model, rendering it inaccurate.

The alternative approach is to disregard any external data and assume that all factors are already embedded in price data and order flow. This relies on the assumption of market efficiency, although we know that the market is not always efficient. Nonetheless, this assumption presents a more manageable problem compared to disentangling the overlapping data affecting our predictive factors.

With this approach, our task would be to predict when the market is behaving inefficiently, indicating a discrepancy between the market and the present information available. However, in this context, there is no single “correct” price as the current price is considered “right.” Our focus would shift to reacting to order flow as it unfolds, only entering trades during periods of significant bias in one direction. This strategy offers a high expected value per trade, especially over very short time horizon. Additionally, we can employ statistical frameworks such as the Copernican lifetime equation to measure the expected duration of price impulses (impulses are not trends; trends appear on higher timeframes), both long and short, adjusting trade size and duration accordingly.