Forecasting the Lead Week: A Deeper Dive

Because the CSI Ratio is a historical index, basic machine learning algorithms can be applied to study the large scale of its daily data points. The resulting forecasts of the following week's price, volatility, and Sharpe Ratio for the S&P 500 have shown to be statistically significant and materially accurate.

The following gauges represent live 1-week forecasts for price, volatility and Sharp Ratio for the following week. The middle value of each gauge (i.e., needle pointed at 12 o’clock) represents the historical average with anything to the left or right representing relatively higher and lower values, respectively.



It is important to understand the long-term accuracy of the above forecasts to understand their respective viability better. The following table illustrates the correlation score between forecasts and 1-week S&P 500 returns and volatility, over the period from January 1, 2015, to the present date. The results show that:

  1. forecasted PnL and actual PnL exhibit a 0.25 positive correlation
  2. forecasted volatility and actual volatility exhibit a 0.60 positive correlation
Long-Term Accuracy of Forecasts (correlation scores)

Actual PnL Actual Volatility
Forecasted PnL 0.25
Forecasted Volatility 0.60