A lot of trading signals capture specific market participant behavior and predict future market price moves. A simple example would be trading signals that try to detect order flow coming from high-frequency trading participants and use that to get a sense of what portion of available liquidity is from very fast participants with the ability to add and remove liquidity at prices very fast, sometimes faster than other participants can react and trade against.

Another example would be trading signals trying to capture participant behavior in related markets, such as cash markets or options markets, to gain an advantage in other related markets, such as futures markets, for similar trading instruments. Sometimes, if a large amount of market participants that these trading signals capture and leverage exit the market, become more informed, or are able to disguise their intentions better, then these trading signals that depend on these participants no longer retain their predictive abilities and profitability. Since market participants and market conditions change all the time, signal decay due to absence of market participants is a very real and very common occurrence and something that all profitable market participants have to account for and deal with.

This involves having teams of quantitative researchers always searching for new predictive trading signals that are different from existing trading signals to counteract the possibility of currently profitable trading signal decay. The signal-parameter optimization aspects we covered in the previous section also help to alleviate this problem by using existing signals but with different parameters to get information from new participants, as information gleaned from existing participants decays over time.