When we evaluate past performance, we’re doing more than flipping through history. We’re trying to answer: how likely is it that this return pattern reflects genuine skill—rather than luck or beta exposure?

This is where statistics becomes our ally. The Central Limit Theorem—that dusty old concept from stats class—actually holds the key. It tells us that as we collect more independent observations, our sample averages start to approximate true expectations. That’s how we gain confidence in what we’re seeing.

One key formula is:

Standard Error = σ / √n
(where σ is the standard deviation and n is the number of independent observations)

Here’s what it means in plain English: If you want to halve your estimation error, you need four times as many independent data points. Confidence grows slowly—and only if the observations are truly independent.

This is why daily NAV data seems attractive: 250+ data points a year! But there’s a catch. Returns often move in clusters. Yesterday’s return affects today’s, especially in strategies with low turnover or tightly held positions. That correlation quietly chips away at the number of independent data points you really have.

And here’s the kicker: If you care about more than average return—say, volatility, skewness, or tail risk—you need exponentially more data.

  • Standard deviation error ≈ σ / √(2n)
  • Skewness error ≈ √(6 / n)
  • Kurtosis error ≈ √(24 / n)

 

Bottom line: The deeper you want to go into the risk profile, the more data you need. A one-year track record might help with return estimates—but it tells you almost nothing about the strategy’s worst-case behavior.

 

You Have to Clean the Signal

Now let’s add another layer. Even if a fund’s returns look impressive—are they really the result of the manager’s decisions? Or just the market doing the heavy lifting?

This is where return-based attribution earns its keep. Strip out the things the manager doesn’t control: equity beta, factor exposures, macro tilts.

A simplified regression looks like this:

Fund Return = Alpha + (Beta 1 × Market Return) + (Beta 2 × Value) + (Beta 3 × Momentum) + … + Residual

The residual is what you care about. That’s where the manager’s unique value lives. If you’re going to analyze a track record statistically, it better be this cleaned-up version—not the raw returns inflated by bull markets or style drift.

 

It's Not Just About Time. It's About Bets.

Here’s where things really start to get practical.

We often think of a “three-year track record” as some kind of magical threshold. But what matters more is: How many independent investment decisions were made during that time?

Think about a high-frequency futures trader—someone making dozens of independent trades a week. You can learn a lot about them in six months.

Now compare that to a concentrated value investor who holds 10 names for years at a time. Even if they’ve been running for five years, they may have made fewer truly independent decisions than that futures trader did in a single quarter.

A better frame is:

Effective Sample Size = Time Period × Decision Frequency × Portfolio Breadth

In other words: don’t just count days or months. Count bets. Count convictions. That’s the true currency of manager evaluation.

 

The Ergodicity Trap: When the Past Stops Being Useful

Now for the uncomfortable truth. Even if a track record is long—and full of decisions—it still might not tell you what you need to know about today.

That’s because strategies evolve. People leave. Teams grow. Markets change.

In statistics, we call this non-ergodicity—when time averages don’t match ensemble averages. Or put more bluntly: what happened back then doesn’t necessarily reflect what will happen next.

Maybe the manager got better. Maybe the strategy got bloated. Maybe the opportunity set disappeared entirely.

A ten-year return history sounds comforting. But if the process or personnel has changed materially, you’re staring at a museum exhibit, not a roadmap.

Heraclitus said it best: You never step into the same river twice. The same goes for investment strategies.

 

What If the Numbers Don’t Help? (They Often Don’t.)

So what do you do when the track record is too short… or too stale… or too hard to interpret?

You go back to fundamentals: Who is this manager? What is their process? What’s under the hood?

Let’s say you’re looking at a new credit fund. It’s 18 months old. That’s not enough for statistical confidence. But:

  • The team previously ran credit portfolios at respected shops.
  • The risk process is structured, documented, repeatable.
  • The operational setup is tight.

Even without a five-year track record, you can form a view. How? By comparing it to funds with similar people, process, and setup. This is interpolation—not extrapolating from a thin sample, but benchmarking against known entities.

In this kind of qualitative analysis, you’re expanding your dataset sideways—not just backward in time.

 

A Practical Framework for Evaluating Track Records

Here’s a simple checklist you can apply:

1. Start with the Strategy

  • How frequently does it trade?
  • How many independent positions does it hold?
  • What kind of tail risks does it carry?

 

2. Apply the Stats (But Don’t Blindly Trust Them)

  • Use standard error formulas to assess what you can meaningfully estimate.
  • Double-check whether your data points are really independent.

 

3. Strip Out the Noise

  • Run return attribution models to isolate true alpha.
  • Focus your analysis on the residuals—not the raw return.

 

4. Estimate Effective Sample Size

  • Think in terms of decisions, not months.
  • Adjust for serial correlation and portfolio overlap.

 

5. Check for Process Drift

  • Is the current strategy team and framework still comparable to what it was years ago?
  • If not, discount or discard the older track record.

 

6. Lean on Qualitative Analysis

  • Evaluate the manager’s process, pedigree, and organizational setup.
  • Compare against similar funds to interpolate expected

 

Rule of thumb: The less statistical certainty you have, the more qualitative clarity you need.

 

Final Thoughts: Track Records Are a Mirror, Not a Map

Track records reflect decisions. But they don’t explain them. They don’t account for luck, shifting regimes, or changes in process.

So instead of asking "How long is the track record?" ask:

  • How many independent decisions were made?
  • How relevant is the past to the present?
  • How much of this performance came from skill—not beta?
  • And what does this track record say about what comes next?

 

There’s no magic number. But with a sharper lens—and the right mix of quantitative tools and qualitative judgment—you can see a lot more clearly.

Subscribe to our Newsletter