Remember that a trading signal has input parameters that control its output/behavior. Similarly, prediction models, which are combinations of trading signals, have weights/coefficients/parameters that control how trading signals interact with each other. Finally, trading strategies also have many parameters that control how trading signals, predictive models, and execution models work together to send the order flow to the trading exchange in response to incoming market data, how positions are initiated and managed, and how the actual trading strategies behave. This is the final finished trading strategy that gets backtested and deployed to live trading markets.

Let's discuss this in the context of a trading strategy we're already quite familiar with. For example, in the trading strategies we saw in Chapter 5, Sophisticated Algorithmic Strategies, there were static parameters as well as volatility-adjusted dynamic parameters that controlled thresholds for buy/sell entries, thresholds to control over-trading, thresholds to lock in profits/losses, parameters that controlled position increase/decrease, and parameters/thresholds that controlled the strategy's trading behavior as a whole. As you can imagine, different trading strategy parameter sets can produce vastly different trading results in terms of PnLs and also in terms of risk exposure that the trading strategy is willing to take, even if the trading signals or predictive models themselves do not change.

Another way of thinking about this is that individual trading signals provide opinions about future market price moves, while predictive models combine many different trading signals with different opinions and produce a final opinion about future/expected market price moves. Finally, it is the trading strategy that takes these predictions and converts that into the outgoing order flow to be sent to the exchange to perform trades and manage positions and risk in a way that converts predicted price moves into actual dollars, which is the final objective of all algorithmic/quantitative trading strategies.

Trading strategy parameters are optimized using similar infrastructures, components, and methods to optimize trading signals and predictive models, the only difference is that here the optimization objectives are PnL and risk instead of predictive ability, which is used to evaluate trading signals and predictive models. Continuously evaluating and optimizing trading strategy parameters is another important step in adapting to changing market conditions/participants and staying consistently profitable.