How to Read the Track Record of an Investment Method (and Why We Don't Publish One)

Screenshots of spectacular gains cost nothing to produce. You only ever see the position that worked, over the period that suits. They prove nothing — and yet they convert, which explains their abundance.

The only serious proof is a dated, continuous and complete record, losses included.

Three biases that inflate published performance

Survivorship bias

Measuring a strategy on the companies that still exist today mechanically erases bankruptcies and delistings. The result is flattering and false. An honest record includes the companies that disappeared along the way.

Look-ahead bias

Using data published after the fact to justify a past purchase is the most common amateur backtesting error. Annual accounts are not known on December 31: they appear weeks later. A serious backtest is point-in-time — it uses only the information genuinely available at each decision date.

The chosen period

Every strategy has its favorable window. A record starting precisely at the March 2009 low tells you more about the start date than about the method. Insist on complete, consecutive calendar years.

Questions to ask before paying

• Does the period cover at least one full bear market?
• Are transaction costs and slippage deducted?
• Were the entry and exit rules published before the measured period?
• How often is the selection recomputed, and what exactly makes a name leave it?
• Are the simulated portion and the genuinely tracked portion distinguished without ambiguity?

That last question is decisive. A backtest remains a simulation, however rigorous. Presenting it as a live record is, at best, negligence.

The four numbers that actually matter

A track record never reduces to a single figure — and the one pushed hardest is precisely the least informative.

Cumulative return

This is the advertising number: "+185%". It says nothing until you know over how long. +185% in three years and +185% in fifteen describe neither the same strategy nor the same risk.

Annualized return

The average compounded return per year. It is the only figure that lets you compare two methods, because it neutralizes duration. When a pitch shows a spectacular cumulative figure but omits this one, the omission is rarely accidental.

Maximum drawdown

The gap between the highest point reached and the low that followed. This is the number that decides whether you will hold on. A strategy with an excellent annualized return but a 60% maximum drawdown will be abandoned by almost everyone who follows it — and abandoned at the worst possible moment, namely the bottom.

Recovery time

How many months to regain the previous peak. Two strategies with identical maximum drawdowns are not comparable if one returns to its high in eight months and the other in six years.

The ultimate test: reproducibility

A method whose rules are published can be recalculated by a third party. It is the only check that does not depend on the seller's good faith. If nobody, even given time and effort, can redo your calculations, your track record is not evidence: it is an assertion.

Ask the question plainly: given your rules and public data, would an attentive reader arrive at the same ranking you did? The answer sorts methods from sales arguments remarkably fast.

This is also why any element of human selection, however small, changes the nature of what is being promised. The moment a manager "adjusts" a holding, the record stops being that of a method and becomes that of a person — which you can neither verify nor reproduce, and which walks out of the door the day they do.

There is a second, quieter benefit. Rules written in advance protect the person applying them. In a bad stretch the temptation to intervene is overwhelming, and it is precisely then that discretionary decisions do the most damage. A published rule removes the question.

What we do not publish, and why

Everything above applies to us, and it is the reason you will find no performance record on this site.

We could produce one. A mechanical method, backtested point-in-time, yields a curve like anyone else's. But hold that curve against the standard set out above and the objection is immediate: you would have no way to rebuild it. You do not have our universe as it stood on each past date, our fundamentals with their real publication lags, or the companies that were delisted along the way. You would have a number, our word for it, and no means of checking either. A figure a reader cannot reconstruct is not evidence. It is an assertion with a decimal point.

There is a second reason, less flattering to the industry. A published record is read as a promise however carefully it is captioned, and it quietly sets an expectation about what the next twelve months owe you. No method owes you twelve months. Anyone who has read the section on drawdowns above already knows why that expectation is the thing most likely to make a reader abandon a sound approach at the worst moment.

What we publish instead: the rule

What can be checked is the rule, so that is what we publish. Which readings a company gets, what puts it on a list, what takes it off, and on what data. Health, Upside and Odds are set out on the method page; each list states the question it answers; and the lists are rebuilt every trading day whether the day's result flatters us or not.

That is a deliberately narrower claim than a performance figure, and it has the one property a performance figure lacks: you can test it yourself. Read the rule, look at today's ranking, and judge whether the first produces the second. Nothing in that sequence requires you to trust us — which is the only position from which the four preceding sections mean anything at all.

The trade is plain. A record you could not verify would sell better. A rule you can read is the only thing worth handing you, because it is the only one that survives your checking it.

This article is information and education. It is not personalized investment advice. Past performance is not a guide to future performance. Investing carries a risk of capital loss.