This is an attempt to use the earnings date metrics from EventVestor to take advantage of market inefficiencies that could be created by mean reversion algorithms. If a company reports poor earnings and its stock price falls, mean reversion algorithms will buy its stock anyway and hold it for a few days assuming that the market overreacted. My theory is that the market may not have overreacted and the stock could fall further after the mean reversion algorithm exits the position. I tried it with the long side as well but it seemed like alpha was negative.
Hedged this with SPY and used a fixed value to represent the change in market cap, it might work better to use the top 20 stocks and the bottom 20 stocks or something like that.
The algorithm loses money during the financial crisis but generates much more alpha if the backtest starts in 2011, I used free data so don't know what happens over the last two years.
