Subscriptions are offered in the EU and EEA only, so you can read everything here but not subscribe. This is a deliberate limit while the project is a pilot, not a comment on where you are.
Institutional screening for private portfolios

Beat the S&P 500 by
owning less of it.

Twice a year all 500 of the largest US companies are ranked on the cash they generate, and the weakest half is dropped. You get the companies that survive, how much of each to hold, and the exact number of shares to buy — enough to build the portfolio outright, or to check against what you already own.

19a month · seven days free
Start 7-day trialSee a portfolio
Free for seven days, then €19 a month. Cancel any time.
Concentrated · 10 years, backtested
21.35% a year
the S&P 500 15.47% over the same period · to Aug 2026
+100%+300%+500%
Jan 2017Aug 2026
Concentratedthe S&P 500

Growth on money invested at the start, dividends included, net of trading costs, against the S&P 500 Total Return. A backtest is not a live account. How this is calculated.

16.76%
a year, backtested since 1998
+7.41%
a year ahead of the S&P 500
83x
invested money over 28.5 years
-36%
from a high

Backtested on 28.5 years of point-in-time filings, dividends included, net of trading costs, against the S&P 500 Total Return. Past performance does not predict future results.

What arrives, and what keeps arriving

A new book each January and July, in three parts: which companies to hold, how much of each, and how many shares that comes to for the amount you have. Below is Balanced as it stands today, with the company names hidden.

Part one
The list

Every company the screen kept this period, with its sector. The 250 it excluded are published in full, names and all.

Technology
Consumer Cyclical
Industrials
Healthcare
Technology
Consumer Defensive
Energy
+ 28 more
Names shown once subscribed
Part two
The weights

How much of the book each position is, worked out from company size. The largest can run to about a fifth — big positions are not trimmed back.

19.4%
12.1%
9.6%
7.8%
6.2%
5.1%
4.4%
+ 28 more
Weight distribution is real
Part three
The order list

Whole shares, worked to the amount you enter. Nothing rounds to a fraction no broker will sell.

Buy14×$2,140
Buy9×$1,210
Buy31×$960
Buy6×$780
Buy22×$620
Buy4×$510
Buy18×$440
+ 28 more
Sample at $25,000
A year of the subscription
January
A new book

Every position re-ranked and re-weighted on the latest filings. About half the names change.

February to June
Changes only

If a holding is bought, delisted or stops reporting, its replacement arrives with the trade to make.

July
A new book

The second reset. Even where the names hold, six months of price moves have pulled the weights off.

August to December
Changes only

The same watch on the holdings, and the book left alone until January.

Start 7-day trialThe whole book, with the names, opens on the Balanced page.

The companies behind the record

The businesses the screen has held across 57 half-yearly periods since 1998. The companies in the book right now are withheld; the ones it has held in the past are not.

Most-held companies
313 held at least once
TickerCompanyPeriodsAvg return
CLColgate Palmolive35+3.40%63%
HDHome Depot27+10.48%81%
MSFTMicrosoft23+8.00%70%
INTUIntuit22+10.76%82%
VRSNVerisign22+7.15%64%
ORCLOracle21+15.57%67%
TGTTarget20+2.67%45%
GWWWW Grainger19+9.38%68%

Periods counts the half-years each company was held, out of 57. Avg return is the mean six-month return over those periods only, not over all 57. Up is how often it rose while held.

The eight most-held of 313 companies the screen has held at some point. Companies that later went bust or were bought out are counted in the record too, so the returns above are not just the survivors.

The last eight half-years, against the index
ConcentratedS&P 500
2022-07
-4.28%
-0.44%
2023-01
+29.71%
+13.52%
2023-07
+4.94%
+6.43%
2024-01
+18.15%
+14.77%
2024-07
+27.13%
+10.12%
2025-01
+32.76%
+5.64%
2025-07
+41.51%
+10.13%
2026-01
-2.95%
+8.56%

Eight of the 57 half-years in the record. Across all 57 the book finished ahead in 31 of them and behind in 26. Winning slightly more often is not where the difference comes from: the wins are bigger. An average half-year ahead was 10.37% clear of the index; an average half-year behind was 3.70% short of it. Compounded over 28.5 years that gap is the whole gap.

The 250 companies we will not own

The weakest companies on cash generation trailed the rest by about four points a year, in the first half of the record and again in the second. The list is published in full, free, with no account. Check it against what you already hold.

See the list

151 trading strategies, tested the same way

Each one was specified on 1998–2014, then measured once on the eleven years set aside before any of it started. Thirteen survived that test. Three of them shape the portfolios below.

Cash, not reported profit

Reported profit depends on judgement calls — when a sale is booked, how quickly an asset is written down. Cash actually collected does not. Of 108 measures tested, cash generation was the strongest.

Take companies out, not pick them

The weakest companies on cash trailed the rest by about four points a year, and by a similar margin in each half of the record. Selecting winners was markedly less reliable, so the method removes 250 companies.

Two resets a year

Rebalancing more often did not improve the result and did cost more. The only thing that moves the book in between is a takeover or a failure.

The failure register
14 named in full on the evidence page

Most of these are other people's published research. A strategy that fails here failed on this universe, over this window, under our reading of its rules — which is a statement about the test, not about the original paper.

StrategyFamilyOutcome here
12-1 month price momentumotherdid not replicate here
The accruals anomalyqualitydid not replicate here
The asset-growth anomalyotherdid not replicate here
The Best Six Months ("Sell in May and go away")seasonaldid not replicate here
Best Six Months with MACD timingtimingtrue, but not worth trading
The book-to-market value premiumvaluenot measurable here
Open the full register →10 did not replicate here · 2 not measurable here · 2 true, but not worth trading

Portfolios and pricing

Over the last ten years the three returned within 1.5 points a year of each other. The deepest fall ranged from -28% to -19%. Over that window the setting being chosen is the ride, not the return.

Concentrated
Higher volatility
19/ month
19.53%
a year, 10 years
+3.85%
vs the S&P 500
22%
-28%

Ahead of the index in 31 of 57 half-years — a little more often than not. The gap comes from the size of those wins: an average of 10.4 points ahead in a good half-year against 3.7 points behind in a bad one.

Start 7-day trial
Works in whole shares from $10,000
~49 orders a year
Most chosen
Balanced
Moderate volatility
19/ month
18.03%
a year, 10 years
+2.35%
vs the S&P 500
20%
-24%

Ahead of the index in 35 of 57 half-years — a little more often than not. The gap comes from the size of those wins: an average of 6.5 points ahead in a good half-year against 4.3 points behind in a bad one.

Start 7-day trial
Works in whole shares from $15,000
~63 orders a year
Broad
Lower volatility
19/ month
18.44%
a year, 10 years
+2.76%
vs the S&P 500
16%
-19%

Ahead of the index in 31 of 57 half-years — a little more often than not. The gap comes from the size of those wins: an average of 5.8 points ahead in a good half-year against 4.5 points behind in a bad one.

Start 7-day trial
Works in whole shares from $15,000
~78 orders a year

Ten years to 2026, net of trading costs, against the S&P 500 Total Return. Over the full 28.5-year record the spread is wider — 16.76%, 13.93% and 11.63% a year — and the ordering matches the volatility ordering. Both full record for each is on its own page. Switching between settings is free and the price does not change.

Size a position before subscribing

Put in what you have to invest and how much of a fall you could sit through. It picks the setting that matches, shows the range of five-year results that setting actually produced, and tells you whether your amount is enough to buy the book in whole shares — which is what decides whether any of this is practical for you.

And in a bad year
Held for
MatchesBalancedModerate volatility
Five-year outcomes in the record
4.03% to 17.00%a year, median 11.00%

Eight in ten of the 48 overlapping five-year stretches fell in that range. The worst ran at -0.15% a year for five years. This is a record, not a forecast, and the next five years may fall outside it.

Covers 97% of the book

25 positions round to a whole share at this amount. Those that do not come to 2.52% of the book between them. The whole book fits from $15,000.

$345,861

after 20 years at the full-record rate, against $161,259 in the index, with the €19 a month taken out of the portfolio annually.

The things you are right to ask

What exactly arrives, and when?

A reset book in January and July: the list, the weight of every position, and an order list in whole shares for a stated amount. In between, a note whenever a holding is taken over, delisted or stops filing, with the replacement and the order to make the swap.

Why keep paying between resets?

Two resets a year is the visible part. The rest is the watch: across the record, holdings were taken over or ceased trading often enough that a book left untouched for a year drifts away from the method it was built on. A lapsed subscription also means the next reset does not arrive, leaving a snapshot of a ranking that has since moved.

Do I need to create an account?

No. Card details go to the payment processor at signup, never to this site, and there is no password to set. A link arrives by email and opens the book; it works on any device and a fresh one can be sent whenever it is needed.

Is this financial advice?

No. Every subscriber receives the identical list, and nothing is assessed against any individual's circumstances, objectives or tax position. It is general investment research published to a fixed method, with the disclosures required under EU market-abuse rules on every page.

Why not simply buy an index tracker?

A tracker costs almost nothing — roughly €7 a year on €10,000 — and in exchange it owns the weakest half of the market alongside the strongest. This ranks the same 500 companies on the cash they generate and leaves the weaker half out, at the cost of two sets of orders a year and a subscription. Whether that trade is worth it is the question, and a tracker is a perfectly sound answer.

How much trading does this involve?

Two sittings a year. There is no daily signal and nothing to monitor. Trading costs are already deducted from every figure published here, at 0.2% of the fraction of the book that changes hands.

What if the amount available is small?

The calculator above answers it for any amount. Below roughly $10,000 the smaller positions stop rounding to a single share, and what remains is a partial version of the book that will not track the published record exactly. The calculator says so when that happens.

These are US shares. What about currency and tax?

The holdings are US-listed and priced in dollars, so a return measured in your own currency also carries the exchange rate movement. US withholding tax on dividends applies and is reduced by treaty for most EU residents once a W-8BEN is on file with the broker. Tax treatment is a matter for an individual's own adviser and nothing here accounts for it.

What happens on cancelling?

Cancelling inside the first seven days charges nothing. Afterwards the subscription runs to the end of the paid month and the book already delivered stays readable. No further resets arrive.

What if the method stops working?

It may. When academics re-test a published market edge on fresh data, about a quarter of the advantage disappears, and more than half of it goes once the finding has circulated widely and others trade on it. That is why eleven years were set aside here and examined once. Where a setting failed on those held-back years it was not published.

Who produces this?

The producer is Purple Turtle, the publishing name for research produced in a personal capacity in Denmark by Christian Mortensen, Pakhusgården 28, 5000 Odense, Denmark. That is stated in full on the terms page and linked from every disclosure block, because EU market-abuse rules require the producer of research to be identified when it is published and not at the point something is sold. Not authorised or supervised by Finanstilsynet: general research published identically to every reader does not require it. The producer holds the same portfolios, and those positions are disclosed with every ranking.

Seven days, then €19 a month.

The list, the weights and the order list, unlocked in place. Cancelled inside seven days, nothing is charged.

Start 7-day trial
Card taken by the payment processor · available in the EU and EEA · not personal advice