The problem

The common advice for a retirement saver behind on their goals is “just DCA into an index fund, time in the market beats timing the market.” I wanted to actually test that claim instead of repeating it. Specifically: does scaling your monthly contribution up when the market is down — a common tactic — actually guarantee you come out ahead over a realistic 15- or 20-year investing window, or does it just feel disciplined while quietly depending on when you happened to start?

This isn’t just a personal finance question. Any team that builds financial planning tools, retirement products, or investment guidance — a wealth manager’s client materials, an HR benefits team picking a default 401(k)/pension glidepath, a fintech’s “set and forget” investing feature — is making the same implicit claim to its users. Most of that guidance is backed by a handful of illustrative examples, not by testing the rule against every historical starting point. That’s the gap this project targets.

The approach

I built two Excel simulation engines on daily S&P 500 data (ticker ^GSPC) going back to 1950: one running every 15-year investing window from a 1950 start through a 2010 start (so the last window ends in 2024), and one doing the same for 20-year windows through a 2005 start. Each workbook has a summary sheet plus one sheet per starting year, applying a conditional contribution rule — $200/month normally, $300 if the index is down 15-30% from its all-time high, $400 if down more than 30% — and computing the resulting CAGR for every single starting year, not just a cherry-picked one or two.

That simulation exposed the actual problem: for some historical starting years, scaling contributions by drawdown alone does not guarantee a positive total return over 15 years. So the second half of the work designs and tests a fix — a “glidepath” or tranche strategy, where the investment is split into tranches and the first tranche is sold off before the 15-year window ends if specific recovery conditions are met, rather than holding everything to the last day regardless of what the market is doing. I compared equal-glidepath vs. conditional-glidepath variants and tested different drawdown thresholds to see which held up best across the full historical distribution of starting years.

All of that got written up as a ~15-page academic-style research paper (abstract, sections, conclusions, bibliography) generated and iterated through Claude, working from a Word document of research notes plus the two simulation workbooks as source material, going through a review draft, a v0.5 pass, and a final v1.0. The paper closes with a practical implementation section: an actual ranked list of physical-replication, accumulating S&P 500 ETFs available on US and EU markets, sorted by TER and fund size, so the strategy isn’t just theoretical — someone could open a brokerage account and execute it with named, real funds.

What I learned

The headline finding surprised me: “average in more when it’s down” sounds like a complete strategy, but tested across every historical 15-year starting point, it doesn’t guarantee you avoid a loss — outcome depends materially on where in the cycle you began. The fix wasn’t a smarter contribution formula, it was a structural one: pre-committing to sell part of the position early under defined conditions instead of holding the entire investment to a fixed end date, which meaningfully improved the worst-case outcomes across starting years.

The other lesson was about process, not finance: the paper got noticeably better through a deliberate critique loop — generate a full draft, then spend several rounds questioning and pressure-testing my own conclusions with the model before calling it final — rather than accepting a single generation pass. Treating the first draft as a starting point to interrogate, not a finished answer, is the difference between a paper that sounds authoritative and one that’s actually been stress-tested.

Where this could go

The obvious extensions are more markets and more data: run the same starting-year-by-starting-year simulation against other indices, other countries’ equity markets, or a mixed-asset portfolio, and refresh it with each new year of data instead of freezing the analysis at 2024. It could also become an interactive tool — plug in your own start date and contribution rule and see where you’d land — rather than a static paper.

The bigger pattern generalizes past investing: test any “rule of thumb” a company puts in front of customers or employees — a default retirement glidepath, a savings-rate recommendation, a marketing claim about a strategy’s safety — against the full historical range of starting conditions, not a single illustrative scenario, before publishing it as advice. That’s a cheap, repeatable due-diligence step any wealth-management, HR-benefits, or fintech product team could run before shipping guidance that people will act on with real money.