An index fund is just a rule— a written recipe for which stocks are in, which are out, and how much of each to hold. Anyone with a brokerage account can follow a rule. The hard part is writing one you’ll still trust in a drawdown. Here’s the full playbook, same workflow an institutional quant team would use, compressed into eleven decisions.
What a custom index actually is
Before the mechanics: what are we building? An index is three things on one page:
- A universe — the pool of candidate stocks.
- A selection rule — the filter that picks the final list from the universe.
- A weighting scheme — how much of the basket each survivor holds.
Plus a rebalance cadence (how often you re-run steps 2 and 3) and optional constraints(no single name above 8%, no single sector above 30%, etc.). That’s it. Every index you’ve ever heard of — S&P 500, MSCI World, Russell 2000 Value — is a specific combination of those parts.
Step 1 — Define the thesis in one sentence
If you can’t explain the exposure in one sentence, the index will fight you when it drifts. Good examples:
- “Profitable US software businesses at reasonable valuations.”
- “European defence primes and their top-tier suppliers.”
- “Dividend aristocrats with debt-to-equity under 1.”
Bad example: “Good companies with good management and growth potential.” That’s not an index, that’s a wish.
Step 2 — Pick the universe
Every subsequent rule operates on this pool. Pick it wide enough that survivorship bias doesn’t hollow out the result, narrow enough that you can actually research the companies.
| Universe | Typical size | Use when |
|---|---|---|
| Russell 3000 | ~3,000 stocks | broad US exposure |
| S&P 500 | 500 stocks | US large-cap only |
| FTSE All-Share | ~600 stocks | broad UK exposure |
| MSCI World | ~1,500 stocks | developed-market global |
| Sector / thematic universe | 20–200 stocks | narrow thesis (AI, defence, clean energy) |
Step 3 — Filter the universe to a finalist list
The selection rule is where opinions enter. Common filters:
- Fundamental: market cap, revenue growth, ROE, gross margin, debt-to-equity, free-cash-flow yield.
- Factor-based: value (low P/B or P/E), quality (high ROE, low leverage), momentum (12-month price return), low-volatility.
- Qualitative / thematic: business description, revenue exposure, supply-chain position, NAICS code.
- Exclusions: ESG screens, sanctioned jurisdictions, blacklisted tickers.
Step 4 — Choose a weighting method
| Method | Formula | When to use |
|---|---|---|
| Equal-weight | 1 / N for every holding | you trust the selection but not the ranking |
| Market-cap | weight ∝ market cap | you want to mirror aggregate investor conviction |
| Factor-weighted | weight ∝ score on chosen metric | you want the biggest positions to be the strongest on your thesis |
| Inverse-volatility | weight ∝ 1 / σ | risk parity — each name contributes equal risk |
| Custom | weight = f(your rule) | you have a specific view the standard schemes can't express |
Equal-weight is a great default. It beats market-cap weighting on long horizons in most backtests — at the cost of higher turnover.
Step 5 — Set a per-holding cap
Any weighting method can blow up if you don’t cap individual positions. An 8% cap is a common institutional default. On a 20-name index that means the top holdings can be at most ~2.5x the smallest — aggressive enough to concentrate, safe enough to survive a single-name blow-up.
Sector caps (e.g. “no sector >30%”) are worth adding if your universe is thematic — otherwise you’ll wake up to find one sector is 60% of the book.
Step 6 — Choose a rebalance cadence
| Cadence | Turnover | Tax cost | Best for |
|---|---|---|---|
| Monthly | high (~40–80%/yr) | high | momentum / trend strategies |
| Quarterly | medium (~20–40%/yr) | medium | most factor strategies |
| Semi-annual | low (~10–20%/yr) | low | fundamental / quality |
| Annual | very low (~5–15%/yr) | very low | buy-and-hold tilts |
The trade-off: more frequent rebalancing keeps you close to your target exposure but incurs trading costs and realises gains. Tax-sheltered accounts (ISA, 401k, SIPP) can ignore the tax side; taxable accounts should lean towards semi-annual or annual unless the strategy genuinely needs speed.
Step 7 — Backtest honestly
A backtest that uses this year’s index membership to pick last year’s winners will always look spectacular. Real backtests have to handle:
- Survivorship bias. Delisted companies must be in the historical universe at the time they existed.
- Look-ahead bias. Only use data that was actually available on the rebalance date — financials lag by 45–90 days.
- Corporate actions. Splits, spin-offs, dividends all need to be reflected.
- Transaction costs.At institutional scale it’s 10–20 bps a turn; at retail size with a liquid universe it’s closer to zero, but don’t assume zero.
Step 8 — Stress-test the rules
Before you trust the backtest, try to break it:
- Run it across regime shifts — 2008, 2020, 2022. Does the drawdown stay inside your stomach?
- Vary the parameters. If halving the universe size collapses returns, the result was over-fitted.
- Shuffle the rebalance date. A strategy that only works rebalancing on the 15th of the month is not a strategy.
- Compare to a benchmark you’d hold otherwise. If you can’t beat SPY on a risk-adjusted basis, just buy SPY.
Step 9 — Execute at the broker
With fractional shares, a 50-name basket takes a single notional allocation and the broker slices it up. On Arithmos Pro this is one click — “execute $25,000 across the index” — with a dry-run preview before anything is submitted.
Without the button, the workflow is:
- Export the holdings to CSV with target weights.
- Multiply each weight by your total allocation to get a dollar figure per stock.
- Upload the list as a basket order at a broker that supports them (IBKR, Fidelity, Schwab, Alpaca).
- Review the pre-trade summary and submit as a single batch.
Step 10 — Document the rules
A one-page document you can show yourself when markets crash is the single most underrated risk-management tool in passive investing. Include:
- The thesis (one sentence).
- The universe and the selection filter.
- The weighting method and caps.
- The rebalance cadence.
- The conditions under which you would change the rules. (Hint: “big drawdown” is not one.)
Step 11 — Review on schedule, not on vibes
Put a calendar reminder for one year from launch. On that day, read the document, check the live performance against the backtest, and decide if anything about the thesis has genuinely changed. Do not review because the market is down. Do not review because the market is up.
Shortcut: have the AI do it
The whole eleven-step process is what Arithmos does in thirty seconds from a plain-English prompt. The value is less in the speed and more in the discipline — every output carries the selection rule, the weighting scheme, the backtest, and the stress tests in a format you can actually review and share.
Open the builder and paste any of the example prompts. Read the generated rationale line by line before you hit execute — that’s the whole point.
FAQ
Do I need a special account to build a custom index?
No. Any brokerage that supports fractional shares can hold a custom basket. The only difference between “buying an index fund” and “running your own index” is that you own the underlying stocks directly, which is what unlocks custom exclusions and tax-loss harvesting.
How many stocks do I need?
Academic work from the 1970s onward suggests diminishing diversification benefits past ~30 holdings. 20–50 is the practical sweet spot. Below 20, name-specific risk dominates; above 100, you’re basically re-buying the benchmark at higher cost.
What’s the minimum capital?
With fractional shares and zero commissions, you can run a 50-name basket on £1,000. Tax-loss harvesting and custom exclusions only start paying for themselves in the low tens of thousands, but the mechanics work at any size.
How do I know my rules are good?
You don’t. Nobody does. The test isn’t “will this beat the market next year” — it’s “can I live with this through a 30% drawdown without changing it”. If the answer is yes, you have an investment policy. If not, the rules are wrong for you.