I noticed P123 was down in the afternoon today (~ 3-4pm ET). Afterwards I re-ran some of my best rolling sims and I noticed I’m getting different results. Most periods in the rolling sim (rolling 1 yr, offset 1mo) look the same but there are a few that are different; enough to change by homebrew metrics and calculations by ~100 bps (median returns) on just about all my sims.
I can got into more depth if needed, but just curious if there was a data/algorithm/UI update today?
Thanks
John
Yes, there was a change released to track cash accruals from dividends and include them in performance in backtests. We planned to announce it next week when we rolled out support for live portfolios. But none of this should affect end-of-period returns on backtests, so I’d be happy to look into whatever difference you’re seeing. It would help if you could point me to a specific rolling test and sample. I’ll try to eliminate any related issues you’re seeing today.
Hi Aaron,
Port https://www.portfolio123.com/port_summary.jsp?portid=1508889 shows a difference.
But I think I know why. I use Weight to rebalance a holding to a target level whenever it becomes disportionately large. If dividends are now part of the return calculation, it makes my transaction history different. In fact, the sim referenced above never buys NVDA where as the live port that it’s based on did. Overall the results are down only marginally.
I’m OK with this change.
Walter
Thanks, Walter.
This difference you saw was likely caused exclusively by the weight rule. The new functionality should only affect the computation of daily performance. This is the case since, although it’s now tracking a cash accruals metric, the money is not included as part of the net liquidation value until the pay date. This approach results in higher backwards compatibility and should prevent the system from deploying dividends before they’re paid.