total return in non-live sims is a bit off

The total return in non-live simulations is a little bit lower than it should be. If your total market value on $100,000 is $12,192.42, the total return should be 12.19%. Instead it shows up as 10.65%. This in turn affects the annual return. See https://www.portfolio123.com/port_summary.jsp?portid=1482285. This is exactly the same as a live port I have, which doesn’t have this problem: https://www.portfolio123.com/port_summary.jsp?portid=1468948. I’ve compared the transactions and they’re exactly the same, as is the total market value. I’ve run a lot of simulations today and it’s a problem for all of them.

Agree,
Last value of Total Equity in simulation performance listing does not match what is shown as Total Market Value (inc. Cash) in General Info.

I found a few quirks with how the first and last date are handled. Until we can resolve it all and make it consistent, you should consider the ‘incorrect’ data as representative of the portfolio one trading day before the end date of the simulation. One more thing to note: the performance chart is ignoring the first date of the simulation, so the return shown there currently excludes one day of performance. Thanks for bringing it to our attention.

Disregarding the chart issue which will remain until a new web server build is deployed, does it seem to be correct now?

The total return is fine but the annualized return is still slightly different from the annualized return for the same sim in my live port. Compare https://www.portfolio123.com/port_summary.jsp?portid=1468948 with https://www.portfolio123.com/port_summary.jsp?portid=1482285. They should be identical.

What number are you expecting? We have two different code bases which produce these numbers (one for backtests which is done once and one for portfolios which is done daily), so I need to know which one should be changed to match, if either is correct.

So the difference is due to either a rounding error of some sort or a slight difference in the annualization methodology? In other words, for the backtest you annualize the total return and for a live port you take the product of the daily returns and annualize that? Without a deposit or withdrawal of cash, they should be the same. Does anyone else want to weigh in on this? When I calculate my own return (with lots of withdrawals and deposits), I use the time-weighted method rather than the IRR, so in practice I use the product of my returns. But when withdrawals and deposits are not permitted, perhaps a simple annualization of the total return is better?

Does anyone have anything to add here? Are the stats still problematic?

Yes. The total return matches the live and simulated portfolios, but the annualized return is different. Once again, compare https://www.portfolio123.com/port_summary.jsp?portid=1482285 to https://www.portfolio123.com/port_summary.jsp?portid=1468948. The sims are exactly the same with the same total return–the only difference is the annualized return. (Also I inadvertently used two different benchmarks, but that shouldn’t affect anything annualized return.)

  • Yuval

If that’s the case, what number are you expecting?

15.85%, which matches the simulated portfolio. The live portfolio gives a higher annualized return which doesn’t seem justified and which I can’t figure out how it was arrived at.

I’ve found that the difference is caused by differing behavior between simulations and portfolios with respect to ‘End Date’. If you look at the ‘Period’ on the summary page, you can see that the simulation includes today, whereas the portfolio ends on the most recent price date. Using 166 instead of 169 for the total number of days yields the higher value.

Ah, that explains it. Thanks.