EN DE
← FVC Academy Our tests

Our tests: backtest, live test and what we threw away

How we test our rules: with companies that later failed, against a fair market, after costs, and with a portfolio that runs in public.

Video coming soonThe video for this lesson is being produced. You can already read it: the text below is the complete script.Read the lesson ↓
Text lesson · Free, no sign-up

You will learn

  • Why a test without the failed companies tells you nothing
  • What the backtest since 2001 shows, and how much depends on the start month
  • Why the strategy does not avoid declines
  • Why we also show you the ideas that failed
Try it yourself nowOpens the exact spot in the tool this lesson is about.
Open →

Transcript

2,361 companies in our test later went bust or vanished from the exchange. We left them in on purpose. A test that only knows the survivors always looks good and tells you nothing. In this lesson I show you how we test: the backtest, the live test, my hand test from 2005, and the many ideas that failed.

A strategy only deserves trust if there is a reason why it works and that reason shows up in the data. You know the reason from the earlier chapters: a low valuation, quality, and the price returning to value. This lesson is about the data.

We ran the rules from 1992 to 2026, with companies from the US and Europe, including the ones that later failed: 5,617 stocks and 1,200 portfolios. The market is the MSCI ACWI Equal Weighted: every stock with equal weight, with dividends, in US dollars, minus 0.25 percent in ETF costs per year. The strategy pays 0.1 percent per buy and per sale. The comparison starts in 2001, because before that most failed US companies are still missing from the data.

From 2001 the strategy made 11.7 percent per year, the market 8.6 percent, so 3.1 percentage points more per year. We start the portfolio in each of the twelve possible review months and report the median; depending on the start month the strategy made 10.4 to 12.3 percent per year. That range is the most honest number in the whole test: it shows how much of the result depends on the luck of the start.

The strategy does not avoid declines. Its largest drop was minus 57.1 percent, the market's minus 57.4 percent, and it swings a little more. And a backtest is a simulation on old data; the rules were chosen on the same history. Taxes are not included, and past results do not guarantee future ones.

That is why a live test has been running publicly on the site since 1 July 2026, as a portfolio since 10 September 2026: every position, every buy and every sale, with the rule that triggered it. It is young, and returns need time. If the rules change, it continues in the same table.

Long before the automated backtest, I tested the method by hand. In 2005 I wrote down 32 German and US stocks that looked undervalued and followed them until 2025. 26 of the 32 reached their fair value, on average after about five and a half years. On average they returned plus 846 percent against plus 162 percent for the market, back then the average of the DAX and the Dow. But the results are very uneven: 9 of the 32 lost money, and the typical stock returned plus 180 percent. That is exactly why diversification matters. It was a small selection of my own; the backtest is the far bigger test.

In 2026 alone we tested well over two hundred variants: rules that raise or dampen the fair value, extra signals, other buy and sell rules. Almost none passed. We tell you this because a test that only shows the winners shows luck, not the method. And when one measurement once looked too good, a cross-check found the error; we rebuilt it the same day.

Every number in this lesson is in the evidence window, with the same rules and the same date. In short: tested with failed companies, against a fair market, after costs, with honest limits, and a portfolio that keeps running in public. Not investment advice.