Methodology

🔬 Factor-Based Equity Methodology

Factor investing isn't about guessing what's “hot.” It's about relying on what's repeatable.

The factors we use—such as free cash flow growth, quality, valuation, momentum, sentiment, and stability—have outperformed consistently over the long term. Yet, like every strategy, it has its off-years.

💡 Since 2002, the top decile of our multi-factor stock model has crushed the broader market—returning ~50% annualized vs ~8% for the Russell 2000—while enduring similar drawdowns.

🎯 Our aim isn't to beat the market every year, but to compound wealth at a significantly higher rate over time. That takes patience and discipline.

All-weather outperformance?

Factor investing isn't about guessing what's “hot.” It's about relying on what's repeatable.

The factors we use—such as free cash flow growth, quality, valuation, momentum, sentiment, and stability—have outperformed consistently over the long term. Yet, like every strategy, it has its off-years.

💡 Since 2002, the top decile of our multi-factor stock model has crushed the broader market—returning ~50% annualized vs ~8% for the Russell 2000—while enduring similar drawdowns.

🎯 Our aim isn't to beat the market every year, but to compound wealth at a significantly higher rate over time. That takes patience and discipline.

The investable universe

Most stocks aren't worth considering. Here's how we filter:

✅ Markets: US, Canada, and Europe

✅ Liquidity:Median daily trading value > €50k

✅ Share types: Only primary listings (no ADRs, duals, or prefs)

✅ History: At least 3 years of continuous price + fundamentals

We don't force sector neutrality or style balancing. Let the factors speak.

Factors are like characteristics

Every stock that passes our universe screen is scored from 0–100 on each of the following:

Each factor must pass at least 5 robustness checks:

📆 Works in multiple time periods

🌍 Works in multiple regions

🧪 Holds up in random sub-universes

🚫 Isn't driven by outliers

🧠 Has sound economic logic

Only factors that survive this gauntlet are used.

Scoring & Ranking

We calculate scores weekly. Each stock gets a composite score (average of the six factors). This lets us rank the entire universe—highest composite = highest conviction.

We don't overweight one “theme.” The magic is in the blend. Combining these signals makes the model more robust than any single factor on its own.

Portfolio management

🎯 We own the top 30 ranked stocksat any given time. That's it.

🔁 Weekly rebalance rules:

• No trade unless a position changes >1% (or something similar, that limits transaction costs!)
• No position drifts more than +50% (again, a little drift is no issue, but too much is not preferrred)
• Total portfolio drift limited to 15% (see above)
• Slippage modeled from 0.1% to 1.5% based on liquidity (we want to be as realistic as possible)

📊 Weighting:The basic way to rank would be: position size = 50 – rank. So stock #1 = ~5%, stock #30 = ~0%. Dynamic but simple.
We like to go more advanced. Our models take into account potential returns and transaction costs to come up with a weighting, improving returns over the long term.

We keep an eye on turnover, to keep it reasonable—despite weekly refreshes.

Risk management

☣️ Risk isn't about volatility—it's about being wrong. Here's how we protect ourselves:

Diversified across sectors, geographies, and factor types

No leverage. No shorting. No complex derivatives.

No story stocks, restructurings, or bankruptcies.

Liquidity filters ensure we can exit if needed.

By investing only in high-scoring, liquid, multi-factor stocks, we sidestep most landmines automatically.

Continuous improvement

🧬 This methodology is a living thing. New data, new ideas, and better ways to measure risk and reward are constantly tested.

But nothing gets added unless it passes the same 5-test robustness framework. That means no hype, no noise—just signals with real, repeatable edges.

Inside the Strategy Lab

🔬 The Lab's backtests are historical simulations over point-in-time weekly rankings since 1999 — roughly 6,000 global stocks per week, delisted companies included, so no survivorship bias. Each simulated week is measured Monday close to the next Monday close: one clean week, never overlapping windows.

Trades fill at real prices.Every buy and sell executes at the next session's average of its high, low, and twice its close — the standard convention in professional backtesting, which weights the close without pretending you could trade the whole position at it — with a backup session for holidays. A stock that didn't trade the next session can't be bought that week, and obviously broken price bars are refused rather than trusted.

Trading costs are charged three ways, your choice. Liquidity tiers charge each traded dollar an all-in rate by how thin the stock is. Flat bps charges one rate on turnover. Size-aware pricing uses the square-root impact law — volatility times the square root of your trade's share of the day's volume, on top of an estimated spread — so the same strategy shows an honest curve at youraccount size. Two more size-aware honesty rules: no position may exceed 15% of a stock's daily volume per execution day (excess flows to the rest of the book; if everything hits its ceiling, the remainder sits in cash), and multi-day fills carry the part of the week's move they execute through: filling 1/T per day puts the average fill (T−1)/2 days in, i.e. (T−1)/10 of the week's realized move — charged to buys of risers, credited when prices come to you. The participation ceiling is a dial: liquidity-tier runs default to unlimited (the convention classic backtesters use) and can be capped for realism; positions a real account couldn't build are flagged on the picks either way. The exact numbers, so you never have to take our word for it:

Both models add $0.01 per share (which bites low-priced stocks), and a name with missing liquidity or price data costs a prohibitive 6% per side. The size-aware model adds its impact term on top of the spread — the gap between the two columns is exactly the impact budget a small reference trade uses up — plus per-country spread adders for non-US listings (from +0.1% round trip in the Nordics to +1.7% in Hungary, and +0.7% for the UK including its 0.5% buy-side stamp duty).

Position weights come in six flavors, from equal weight to the drift-band rebalancing real simulations use (positions ride until they stray too far from ideal), including cost-aware modes where a stock only wins a place in the book if its edge survives its own trading costs — each score is charged its cost converted into rank points at a measured exchange rate (the simulation tracks, from trailing history only, what one rank point has been worth in %/yr among its top candidates). Under size-aware costs every mode prices seats this way. Sell discipline is a dial too, on two axes: rank patience (sell the moment a holding leaves the top N, or hold while it stays within a band of it — up to 10×) and a minimum hold time (once bought, a name can't be sold for 4, 8 or 13 weeks, however far it slips; delistings excepted).

Every constant, in the open.The remaining assumptions the simulator makes, so nothing lives only in the code: trades fill at the NEXT session's typical price (average of high, low and twice the close), with a backup-bar fallback for holidays and an artifact-print gate; a name with no fresh next-session bar is untradeable and leaves both the book and the benchmark. Universe hygiene drops closes under $1.10 and names trading fewer than ~100 shares/day. A position may use at most 15% of a name's median daily dollar volume per execution day (the same 15% the “Auto” holdings count sizes against); the √-impact law switches to linear beyond 100% participation. Per-name volatility is its own trailing 26-week realized figure (names without history borrow the week's median). The seat-pricing exchange rate is the trailing median score→return slope of the top candidates over the last 52 rebalances (needs 8+ to activate; bounded), and the assumed holding period is measured from the run's own trailing turnover (bounded 1–26 weeks). Drift-band rebalancing uses the classic simulator bands: 50% position drift, 15% portfolio drift, 1% minimum transaction. Cash earns 0%, the same-rules benchmark is the equal-weighted eligible universe, and multi-day fills assume the week's move accrues evenly across its five sessions.

🌲 Every mechanism above exists because we chased a too-good-looking number to its root cause and fixed it. The Lab's job is to compare strategy styles honestly — not to predict returns, and never to flatter them.

How to use this

🔧 If you're a DIY investor, the full formulas are in the e-book.

🧾 If you prefer a plug-and-play feed of our top-ranked names, subscribe to our weekly updates.

💼 If you want full implementation, we offer managed solutions and custom mandates.

Bottom line:

We don't predict. We rank.

We don't listen to hype. We listen to data.

We don't chase returns. We compound probabilities.

This is factor investing—refined, disciplined, and battle-tested.