Now Showing: Historical Performance
“Backtesting”: both an exciting and a dreadful word. Exciting, because when we design an algorithmic investing strategy, we’re eager to show how it should perform. When we released the first iteration of the LendingRobot algorithm in 2014, we did not have any trading history. But what we had was a performance estimate, based on all the data provided by Lending Club. So we started showing those numbers, along with careful explanations about how they were calculated, and how it was only hypothetical performance. In theory, there is no difference between theory and practice! Although we’re using the exact same algorithms both for back-testing and real-use, sometimes the two can differ: either a loan is no longer available, or the exact amount we invest is different from the theoretical one. In addition, sometimes loans make ‘weird’ or off-schedule payments.
Today we have accumulated enough data to show our REAL historical performance.
You can see the interactive chart on https://www.lendingrobot.com/#/resources/performance/
This chart shows the evolution in the value of the notes we invested in, with the exact amount we invested for our clients. It is a ‘composite’ performance, in the sense that it combines all the investments we made when our scoring models were used (fully automated or advanced rules, conservative or aggressive). As we kept growing, more and more money came into LendingRobot. We discarded notes that were invested without the use of our models because if someone uses LendingRobot to invest in Lending Club loans with his own set of criteria only, such as only sub-grade, home ownership or debt-to-income ratio, we shouldn’t take credit for the performance of his investments. Real cash drag is also taken into account, as we measured, month after month, how much idle cash was in our clients’ accounts.
- Alex Neporozhniy
- 2 Comment