How to Evaluate an Investment Strategy Before You Trust It With Money
You cannot verify a forecast. You can verify a process.
That is the whole point of evaluating a strategy before you fund it. Anyone can describe returns they intend to produce. Far fewer can show you the ruled columns behind them: what was decided, when, at what price, and what happened when the decision went wrong. The work below is a sequence of checks, not a personality test. Each step points at an artifact you can ask for and read.
Read the sequence as of September 19, 2026. The method does not expire. The examples do.
Start with the artifact, not the argument
Before you evaluate a systematic investing process, decide what evidence you would accept. Not a narrative. Not a chart with the axes cropped. An artifact.
Three artifacts carry most of the weight:
- A decision log with dated entries
- A published record that includes losses, on the same schedule as wins
- A statement of method precise enough that a stranger could reproduce the rules
If none of those exists, you are not evaluating a strategy. You are evaluating a pitch. Stop there and save yourself the month.
Ask for the decision log
A decision log – sometimes called a trading decision log – is the dated record of what the process decided and why. It is the difference between process versus stock tips in physical form.
Ask for these four things in every entry:
- The date of the decision and the date it was acted on, kept separate
- The instrument, the direction, and the size
- The stated reason, in the words used at the time
- The fill: the price actually paid, and on which account
A real log is boring. It has entries you would not have chosen to publish. It records the decision that was later reversed and says so on the same page.
A retrospective summary – the kind assembled after the fact into a tidy table – is not a log. A log is written forward and dated forward. You can usually tell the difference within ten entries, because forward-written logs contain the small embarrassment of ordinary decision-making.
Read the record on the same schedule as the wins
This is the fastest filter available. Ask one question: when do you publish losses?
If losses appear late, quietly, or inside a footnote, the rest of the record is decoration. Publishing investment losses on the same schedule as wins is not modesty. It is the mechanism that makes track record transparency possible. A record you can only see the good half of is not a record.
Check three details while you are there:
- As-of dates. Every figure should carry the date it was measured against – “as of September 19, 2026,” not “recently” or “since inception.”
- Return basis. Net or gross, and after which costs. A record that never states its basis is not comparable to anything, including itself.
- The losing column. Find the worst stretch and read what the process said while it was happening.
If the losses are cancelled from the window entirely, treat the remainder of the numbers as unverified.
Check whether dissent survives the meeting
A single decision-maker with no published disagreement is not a committee. It is one opinion with better formatting.
Look for dissent in the record. If a process is run by more than one person, the log should show disagreements, the arguments made, and how the vote resolved. Dissent that appears in the log is a signal that the record reflects what happened rather than what the writer would prefer you to believe.
If every decision was unanimous, either the group is small and aligned to the point of redundancy, or the disagreement is being removed before publication. Both are worth knowing before you commit money.
Test the numbers that carry information
Most published statistics about a strategy are decoration. A small number carry information. Sort them accordingly.
Carries information:
- Drawdown disclosure. The depth and duration of the worst peak-to-trough decline, with the dates. An investing strategy drawdown disclosure tells you what the ride actually felt like.
- Number of decisions, not just outcomes. Twenty decisions with three winners and seventeen losers is a different animal from three decisions with three winners, even if the final return matches.
- Fills versus signals. The gap between the price the process identified and the price it actually transacted at. Wide gaps mean the backtest was better than the portfolio.
- Account-level detail. If the same process runs across several accounts, the differences between them are information about execution, not noise.
Decoration:
- Total return quoted without a benchmark or a basis
- Win rate with no average size of win and loss
- Anything measured from a date chosen because it flatters the curve
A quantitative investing methodology that cannot survive this sort is still a method. It is just not one you can check.
Soundings before you commit
Run this list against any strategy you are seriously considering. Take a sounding at each line; write the answer down.
- What is the exact rule you would follow, stated so a stranger could run it?
- How many decisions has the process actually made, and over what window?
- What is the worst drawdown, with dates?
- What does the process publish, and on what schedule?
- Where are the losses in that publication?
- Who disagreed, and how was it recorded?
- What does the fee come out of, and what is the return basis?
- What would have to be true for you to conclude the process is broken?
The last question matters most. A process with no stated failure condition cannot be falsified, and a strategy that cannot be falsified cannot be evaluated – only believed.
When the process fails your audit
Failing the audit is common, and it is not always fatal. Distinguish between three outcomes.
- Missing artifacts. The process may be real but undocumented. Ask again in writing; if the answer stays vague, the answer is no.
- Present artifacts, weak disclosure. The log exists but the losses are thin. You can still proceed, at a smaller size, while you watch whether the losing column fills in.
- No artifacts, strong claims. Walk away. No amount of confidence substitutes for a record.
The cost of an audit is a few hours. The cost of skipping one is measured in years.
The point of the exercise
You are not trying to predict what a strategy will return. You are deciding whether the process is the kind of thing that can be checked at all. If it can, you will be able to tell – later, when it matters – whether it is working or broken.
That is a modest thing to ask for. It is also the only thing you can verify before the money moves.
Start with the log. Read the losing column first. Everything else is commentary.

Leave a Reply