What “backtesting” means here
This tool replays dividend-adjusted (total-return) monthly price data for whatever tickers you choose, month by month, starting from any historical date the data supports. It's not a curve-fit strategy optimized in hindsight — it's the same real allocation, tested against real history.
That distinction matters: a strategy built by searching for whatever weights happened to perform best over a specific past window will almost always look great in that window and worse going forward. Testing a fixed, chosen allocation against every available historical starting point — rather than searching for the best-fitting one — is what keeps a backtest honest.
Why allocation weights matter more than any single pick
Diversification changes how a portfolio behaves even when the individual holdings look similar in isolation. Combining assets that don't move in lockstep — U.S. stocks, international stocks, bonds — changes the overall path a portfolio takes, which is why it's worth testing an allocation as a whole rather than judging one ticker on its own.
The built-in presets — Bogleheads 3-Fund, S&P 500 & Chill, 60/40 Classic, Tech Growth, and Dividend Income — are useful starting points for exactly this reason: each represents a genuinely different allocation philosophy, so backtesting a few of them side by side shows how much the weighting itself, not just the specific tickers, drives the outcome.
What counts as a meaningful backtest
A backtest is only as informative as the number of independent periods it actually tests. A 5-year window run once tells you about one path; testing every rolling 5-year window the data supports (which this tool does automatically) might mean dozens of overlapping starting months — more evidence, though still not fully independent of each other since adjacent starting months share most of the same underlying history.
Longer test windows (10 years, full history) naturally have fewer independent starting points to test, since a 20-year window can't start from as many different months within a fixed dataset as a 5-year one can. That's worth keeping in mind when comparing a “100% survival rate” across a 5-year test versus a 20-year one — they're not drawing on the same amount of independent evidence.
How rebalancing is handled
Every simulation on this site assumes the portfolio is rebalanced back to its target weights every month. That assumption matters: without rebalancing, a portfolio identical at year one can drift substantially by year ten as winners grow into a larger share of the total — this tool holds the target weights steady throughout so allocation comparisons stay apples-to-apples.
As a hypothetical illustration: a 60/40 stock/bond split that isn't rebalanced during a strong multi-year stock rally could drift toward 75/25 or higher purely from stocks outgrowing bonds, which changes the portfolio's actual risk level without anyone deciding to take on more risk. Monthly rebalancing, as modeled here, keeps that drift from happening silently.
Comparing two portfolios
Because everything updates live, there's no need to export or save anything to compare allocations: run one allocation, note the median, best, and worst ending balance in the historical performance summary, then change the allocation and rerun. The numbers update in place for a direct comparison.
Reading the results honestly
The best and worst ending balances represent genuinely different real historical entry points, not statistical noise — they're worth examining individually, not just the median. A high median doesn't mean a guaranteed outcome; checking the worst-case starting month specifically gives a fuller picture of how a portfolio actually behaved across history.