Another solution to dealing with live trading performance deviating from the expected simulation performance is to have sophisticated analytics on live trading strategies. This is another way of saying that instead of relying completely on backtesting performance and behavior, you can also invest in adding enough intelligence and sophistication directly to live trading strategies to reduce the likelihood of simulation dislocations derailing an algorithmic trading business. This, again, is an imperfect approach to solving the problem, but can be a good alternative to help with limitations and errors in backtesters. The idea is to deploy trading strategies to live markets with very small exposure, collect statistics on each strategy action, and properly instrument and collect statistics on why those decisions were made.

Then we resort to an extensive Post Trade Analytics (PTA) framework to dig through these strategy action records and classify winning and losing positions and statistics on strategy actions that led to these winning and losing positions. Often, performing this kind of PTA on trading performance from live trading can reveal a lot of insight about problems/limitations for that particular trading strategy. These insights can be used to guide the development and improvement of the algorithmic trading strategy and improve profitability over time. In many ways, this boils down to the approach of starting trading strategies at very small risk exposures with intuitive parameters and using feedback from live trading to improve the strategy's performance.

This is not a perfect approach, since it requires the trading strategies to be simple enough where they can be run under live trading conditions with easily understood parameters, in addition to the fact that we might have to run a trading strategy that is not profitable in live market conditions for a short time, while accruing losses that we don't want.