CAGR

Posted January 26th, 2008 by ndevr

I am fairly new to mechanical investing but not new to investing in general. Its seems that some of the canned screens have fantastic cagr but I am skeptical as to whether or not an investor can realisitically come close to those types of returns. Does anyone have a good guidance on what one can realistically expect versus what the screen performance has theoretically been? Are these results too good to be true? Even if you were to discount the returns by 40% you would still do better than 99% of most investors? Thanks in advance for your inputs.

ndevr

‹ WWL - Is there another way FoxPro .dbf to MySQL data issues for backtesting ›

Yes, returns are typically

On January 28th, 2008 Keelix said:

Yes, returns are typically somewhat overstated because stocks can not be bought and sold at the same time. The backtester - due to data limitations - assumes sales and buys happens on Friday closing prices, despite the fact that the underlying data for rankings does not become available until Sunday. Estimates (based on testing) of the effect of delaying transactions until Monday typically lower the CAGR by 5-10 percentage points. Another source of errors are spreads, specially for low liquidity stocks. You typically loose a percentage point or two to spreads as well, which the backtester does not account for.

Then there are some other uncertain factors to be aware of as well, such as the fact that CAGR (or even the CAGR/GSD ratio or other measures) may or may not be predicative of future performance. People doing MI believe they are to a certain degree, but personally I lean towards it being a method of trying to "shuffle the cards right" rather than "a bulletproof method of getting nice returns".

And finally, it's disipline. You really need to look at/measure the drawdowns as well and try to get a feel for how you will react to it. Keep in mind that the backtester only measures drawdowns based on weekly data. Daily data, or even intra-day data, may lead to even larger drawdowns. When you are 25% in the read, will you still stick with the methology or not? If you bail out, the "backtested bounce back" will definitively not get you back on track.