Methodology

The four terms

These four come back everywhere on the site. This is what they mean; the rest of the page is the precise version.

Open-market buy

A director or supervisory board member bought shares in their own company with their own money, on the exchange. That is something other than receiving shares as compensation or exercising an option — those are here too, but kept apart.

Percentile

A rank from 0 to 100. If it says 90, this measurement came out above 90% of the other measurements in the same group. It says nothing about the share price and it is not a forecast.

Coverage

How much of a measurement actually succeeded, from 0 to 1. The lower the coverage, the closer the published percentile moves to the middle (50) — that way a half-measured company does not pass for a fully measured one.

Cluster

Three or more independent buyers at the same company within 90 days. A count of what happened, not a recommendation.

How we classify filings

A filing does not say by itself whether something was bought, granted or transferred. A fixed set of rules classifies every row; borderline cases go into a queue for human review and do not appear on the site as fact. The categories:

• Open-market buy — A purchase on the exchange with the insider's own money — the signal this site is built around.
• Open-market sell — A sale on the exchange. It can mean all sorts of things (tax, a house, diversification) and therefore says far less.
• Option exercise — Exercising options or rights. Usually compensation, not conviction.
• Award — Shares or options granted as compensation. Nothing is bought, so there is no buy signal.
• Administrative — An administrative change, for example a transfer within the insider's own holding company.
• Other (no signal) — Other filings that say nothing about buying or selling.
• Pledge — Shares given as collateral. Ownership does not change.
• Gift — SEC code G: shares given away or received as a gift. It can go either way, and the form says which.
• Conversion — SEC code C: a convertible instrument turned into shares. A change of form, not a purchase.
• Returned to the issuer — SEC code D: shares handed back to the company itself, for example when unvested shares are forfeited.
• Tender or change of control — SEC code U: shares tendered in a takeover or a change of control.
• Other (disclosed in a footnote) — SEC code J: something else, with a mandatory footnote explaining what. Every such line goes to review, so none of them appear here as fact.

An amount of 0 is not an error. On an award or an administrative change the register reports a price of 0: nothing was paid. The Amount column then says “nothing paid” and the Quantity column shows how many shares were involved. We never top such an amount up with an estimated market value.

The four insider registers

The register on Insider Trades holds four sources side by side, and you can filter on any of them. The Dutch one is the AFM register under MAR art. 19: directors, supervisory board members and people closely associated with them. The American one is Form 4 at the SEC: directors, officers and holders of more than ten percent of a listed US company. The German one is the BaFin database of managers’ transactions, also under MAR art. 19, and the Belgian one is the FSMA’s transaction search under the same article. All four are filings by insiders about their own company’s shares, and all four are public at the source.

We do not merge them. The same company can appear in more than one without us claiming it is the same issuer: Form 4 carries no LEI and no ISIN, and the German and Belgian filings carry an ISIN but no LEI from the source, so there is no key that would let us join them honestly. Nothing is added up across a border either.

The date. On all four, the date we count from is the day the filing reached the regulator, never the day of the trade. At the SEC, at BaFin and at the FSMA that is a hard fact from the source: all three publish the filing date themselves. In the Dutch register it is partly derived — the AFM does not publish a filing date, so for older filings we use the last day the law allowed, which is the conservative choice. That difference is one of the reasons the scores stay on one register.

The amount. The AFM and the SEC do not publish a total per line: both give a price and a quantity, and the Amount column is those two multiplied. BaFin does the opposite — it publishes the aggregated amount itself and no share count at all, which is why German rows show an amount but leave Quantity empty. The FSMA publishes all three, so a Belgian row shows a quantity, a price and a filed total — and there the amount is not our multiplication: when an order was filled in parts the filed price is a rounded or representative one, so price times quantity need not match the filed total. We show what was filed. We never convert between currencies: that would need an exchange rate and a rate date, and no filing carries one. And no register is single-currency — the Dutch one alone reports in euros, dollars, pounds, rand, kroner and złoty. A minimum-amount filter therefore compares the reported figure as it stands, in whatever currency the filing states.

The German register only goes back twelve months. BaFin publishes a rolling twelve-month window: what is older than that disappears from their database, and it cannot be requested back. Our own archive of it starts with the export of 26 July 2026 and therefore reaches back to 28 July 2025. Anything BaFin published before that date is not in our copy and is no longer at the source either. Neither the AFM nor the SEC applies such a window: the AFM keeps its register online and SEC filings stay on EDGAR indefinitely. So an empty German year on the site means “before our archive began”, not “nothing was filed”.

The Belgian register starts on 21 May 2026, and grows backwards. The FSMA does not publish an export or a bulk file; its register is a search form, so we read it one publication day at a time, at the speed its own robots.txt asks of us. Every publication day from 21 May 2026 onwards has been read, without a gap; everything before that is still being worked through, and it is a lot — we are reading backwards towards 2020 and the source answers slowly. An empty Belgian month before that date therefore means “not yet read back”, not “nothing was filed”. This is the youngest of the four registers by a wide margin, and the smallest.

About three in five Belgian filings are held back for review. That is by far the highest share of the four registers (the Dutch is about one in eight, the German one in three, the American one in twenty-five), and it is not a flaw in the filings. We checked: not one Belgian filing is missing an ISIN, a date, a quantity, a price or an amount. What makes the difference is that the FSMA asks its filers to label a transaction more precisely than the other three regulators do — and the labels it gets back often describe something that is not a plain purchase or sale of shares. A share award bundled into a remuneration package, an option or warrant rather than a share, a subscription to a capital increase, a trade agreed off the exchange. Those go to a human queue instead of onto the site as an ordinary buy or sell. A more precise source produces more review here, not less.

Two Belgian figures that are not comparable over time. First, the reporting threshold changed: until 3 December 2024 a filer had to report once their transactions passed €5,000 in a calendar year, and from 4 December 2024 that became €20,000. A Belgian series that spans that date is not comparable with itself. Second, the FSMA’s search shows at most fifty results per query and never says how many there were in total, so completeness is our own bookkeeping rather than something the source confirms. We read one day at a time to stay well under that limit, and we count every day that comes close: across the 471 days read so far, not one has reached it.

The scores stay Dutch. The insider score, the quant score, the timing score, the percentiles, the clusters and the signal log are calculated on the AFM register alone. The American, German and Belgian rows are facts: dates, roles, quantities or amounts, prices, and the source’s own transaction wording. They carry no score and take part in no ranking. They also have no company page (no LEI to key it on) and no filer page (that index is built on the Dutch register).

Sources. Dutch filings: AFM register of notifications under MAR art. 19. US filings: SEC EDGAR, Form 4. German filings: © Bundesanstalt für Finanzdienstleistungsaufsicht / www.bafin.de. Belgian filings: Bron: FSMA (www.fsma.be), CC BY 4.0 — the same filings are available free of charge from the FSMA’s own transaction search.

The insider score

The score (0–100) is a fixed calculation: the same input always gives exactly the same result — no network, no randomness, no AI. It weighs five factual components:

• What kind of transaction, and by whom — a purchase on the exchange weighs more than a sale; a CEO or CFO more than a supervisory board member or a holding company.
• Size — how large the transaction is relative to the position, as far as the source allows. The AFM detail page gives no total position, so this component is usually reweighted away.
• Cluster — several independent buyers within a rolling 90-day window, after counting the same person and their own holding companies as one buyer.
• How recent — a fresh filing weighs more than an old one.
• What came before it — for example a purchase shortly after a sharp price drop.

Every stored score carries the version of the model it was calculated with (currently 0.6.0). Old scores are never recalculated: that would change the track record after the fact.

What the score is not. Our own backtest shows the score is usable as an ordering within a period, and not as an absolute number or a threshold (“above 70 is worth buying”). We therefore publish it as context, never as advice. The real test runs forward and lives in the track record.

The twelve-month rank

The score level drifts over the years. That is why we also express a score as a rank: where does this buy sit relative to all scored open-market buys of the preceding twelve months? A rank of 90 means the buy came out above 90% of them. Orderings and any later thresholds are defined on that number, never on the score itself again. If there are too few observations in the window, we show no rank. The underlying score remains the source — the rank is derived from it, not a recalculation.

The quant score

Alongside the insider score we calculate a company score (0–100): how the business is doing by its own numbers, entirely apart from who is buying or selling. This too is a fixed calculation with its own version number. The three scores (insider, quant, timing) stay side by side: we deliberately do not add them into one total, because then nobody could see where the figure came from.

The score weighs four components, and deliberately uses one metric per component. Two metrics measuring the same thing would reward that effect twice.

• Quality (35%) — gross profit relative to total assets.
• Valuation (30%) — operating profit relative to the value of the whole business.
• Momentum (22%) — the price return over twelve months, excluding the most recent month.
• What goes to shareholders (13%) — dividends and the company’s own share buybacks, relative to market value.

If a component is missing, we spread its weight over the components that are there, and show alongside the stock how much could be measured. We compare in three steps: first within the company’s own sector, and if that group is too small (fewer than 8 stocks) against all measured companies together. Only if that fails too do we fall back on fixed absolute thresholds.

What that means in practice. That second step is not an exception here but the main route: more than half of all sub-scores rest on the comparison with all companies, because only a handful of sectors reach eight comparable stocks. The fixed thresholds — the last fallback — have so far not been used for a single sub-score. A “sector comparison” here is therefore more often a comparison with the whole universe than the name suggests. We say so, because it changes what such a sub-score means.

You can check every step. On the company page and in the quant ranking each component shows the value we measured, the middle value of the group it was ranked against, how many companies that group holds, and — if the component is missing — the reason it is missing. Those figures are stored with the measurement itself, so a page from months ago still explains itself. What you will not find is a letter grade: a percentile presents itself as a measurement, a report card as a verdict, and only the first is what this is.

Honest about the data source. Our main source for annual figures provides no publication date and no figures-as-they-were-known-then. The quant series therefore starts at the first measurement and grows forward; we do not reconstruct company scores for earlier years. Momentum and timing come from the price series and are real history.

One exception, and it is a real improvement. For the companies that also file with the American regulator, we now read the annual figures from those filings directly. Each figure there carries the date it was actually filed, so for those companies the publication date is a fact rather than an estimate. Where both sources have figures for the same company, one source supplies the entire annual report — never a mix, because ratios that divide two figures assume they come from the same balance sheet.

Different currencies. A company can report in dollars while its shares trade in euros. Two of the four scores divide a figure from the annual report by a figure from the market, so those two would otherwise divide different units. We convert using the official European Central Bank reference rate of the measurement date. If we have no rate for that pair on that date, we leave the score out rather than guess — and we check the outcome for plausibility, so a mislabelled amount cannot quietly become a confident wrong number.

Which companies take part, and how often they count

The four components assume a company with an income statement and a balance sheet. For a listed investment fund or a pure investment holding, three of the four do not exist in any meaningful form: the “revenue” is price appreciation and the “assets” are the portfolio itself. Such funds therefore get no quant score. That judgment is recorded per row in our reviewed ticker table — assessed per instrument, not applied automatically from a sector label: a holding company that consolidates operating subsidiaries does report ordinary annual figures and simply takes part. Excluded funds stay in the database and are counted and named on every run.

One company can appear in the registers with two ISINs — after moving to another country, or with an old listing next to a new one. Both keep their own company page with the same score, but in the ranking and the sector comparison such a company counts once. Otherwise one administrative duplicate would shift the position of every other stock.

What happens when we don’t know something

The percentile is a rank transformation: it says how a stock compares with the rest on that date. Until recently it said nothing about how much sat underneath that score — a stock with one measured component got a percentile that looked just as confident as a fully measured stock. Since the latest model version the published percentile moves toward the middle, in proportion to the coverage:

percentile* = 50 + (percentile − 50) × coverage

At full coverage nothing changes by definition; at half coverage half the distance to the middle remains. The underlying score is unchanged — the percentile is derived, not recalculated, exactly as with the insider score. We deliberately do not apply a hard floor: pulling a stock out of the list entirely because one component is missing is a heavier intervention than is warranted, and it would hide almost a third of the list.

Unknown is not zero. If the source provides no line at all about dividends or buybacks, we treat that component as not measured and reweight to the rest — we do not book a zero. If the source explicitly states that nothing went to shareholders, that does count as zero. That distinction is a measurement choice and not a statement about the company, and it visibly lowers the coverage.

Risk labels and timing

Risk labels are flags, not plus points and not penalty points. There are six, and they measure exactly this:

• Market value below our floor — the market value on the measurement date came out under the floor this model version uses.
• Thinly traded — average trading volume came out under the floor this model version uses.
• Stretched balance sheet — net debt relative to operating earnings came out above the level this model version uses.
• Share count grew — over twelve months the number of shares outstanding grew by more than this model version allows for.
• Little could be measured — the coverage sits below the threshold, so the published percentile is pulled toward the middle.
• Annual figures are out of date — on the measurement date the most recent annual report was older than this model version counts as current.

A label never subtracts anything from the score and never takes a company out of the ranking — it stands next to the figure. A label we cannot evaluate because an ingredient is missing is not set: missing data should not be punished twice. They appear on the company page and, since they are notes about the measurement rather than a verdict, in the same words and without a warning color in the ranking.

The timing score (0–100) is a deliberately simple, technical measure of the price series, and it consists of four checks that count equally: where the price sits relative to its 200-day average, where it sits in the past year’s range, what it did over the last twenty trading days, and whether trading volume picked up alongside. The company page shows all four with the figure each one produced. It says something about the entry moment only, and must never push a fundamentally strong stock aside.

A short price history leaves checks out. Each check needs a minimum number of trading days — the 52-week range needs a full year, the 200-day average needs 200 days — so for a recent listing some of them cannot be calculated. The score is then the average of the ones that could be, and the company page says so and shows which. If none of the four can be calculated there is no timing score at all, rather than a zero.

The track record

The track record logs two things, and only for filings the pipeline sees arrive live: (1) a purchase on the exchange of at least €10,000, and (2) a new cluster of three or more independent buyers within 90 days. The score and the rank are shown as context; they are not a threshold. Every logged signal gets a reference price and follow-up measurements after 20, 60 and 120 trading days — the negative ones just as much as the positive.

Buying along with every filing

On some filer, fund and member-of-Congress pages there is a block that answers one question: what would the share price have done if you had bought after every purchase that filer reported? It is arithmetic on register facts, in the same family as the track record — not a portfolio, not a strategy and not a suggestion to copy anyone. The definition is one definition: the same entry rule, the same windows, the same weighting and the same rounding for all four registers. What differs is only which lines in a register count as a purchase, and which date the register makes public.

The definition, in full:

Which filings count

For a person: purchases on the exchange only, classified and outside the review queue, with a filing date. Awards, option exercises and sales are not in it. For a fund: the five holdings that grew the most in a quarter, measured against its own previous filing, by the reported value of the increase. Holdings that merely became worth more without the share count rising are excluded, and so are put and call lines.

Why a fund’s growth is not always buying

A Form 13F reports what was held on the last day of a quarter, never what was traded. A share split multiplies a holding without a single share being bought, and the form does not say which of the two happened. Where a quarter-on-quarter comparison cannot tell them apart, our own figures mark it — and the holding is still counted among the five. On the five years of filings we hold that is 78 of 2,041 selected holdings, and 38 of those are the largest of their quarter. We would rather name that than quietly present a split as a purchase.

Which date counts for a fund

The day the filings behind that quarter reached the SEC — the later one where a quarter was amended, because until every filing is public nobody could work out which five positions grew the most. A Form 13F may be filed up to 45 days after the quarter it describes, so the positions are already at least that old on the day a reader could first see them. That is a property of the form, not a shortcoming of the measurement.

What we could not price

A fund reports a position by its CUSIP; a price series is found by ticker, and no public source connects the two. We keep that bridge as a reviewed file with evidence per line. A position whose CUSIP has no confirmed ticker yet produces no measurement and is counted, with its own reason, in the coverage list on the page — it is a queue on our side, not a gap in the register.

Which purchases count for a member of Congress

Every line a member reported as a purchase in a Periodic Transaction Report, provided the ticker on that line has been matched to a listed symbol. Sales and exchanges are not in it, and neither is a line that a later amendment withdrew — the filer says themselves that it does not belong there. Lines on a spouse’s, a joint or a dependent child’s account count too. Nearly half of all reported lines are of that kind; the filing is the member’s and the law makes them responsible for reporting it, so leaving them out would drop half the register on a distinction the form itself does not make for reporting. The owner code stands on every line on the member’s page, so anyone can see whose account a purchase was on, and the figure never says the member traded. A ticker still waiting for a judgement produces no measurement and is counted with its own reason in the coverage list — a queue on our side, not a gap in the register.

Which date counts for a member of Congress

The day the register of that chamber published the filing — the Clerk of the House, or the Senate Office of Public Records — never the day of the trade. A member must report a trade within 30 days of learning about it and no later than 45 days after it happened; in the filings we hold the median gap between trade and publication is four weeks, with a tail of years. The day of publication is the first day anyone outside that household could have acted, so measuring from the trade date would count a period nobody could have traded in. That is the same point-in-time rule the rest of this site follows, and it is the single biggest difference between these figures and the ones published elsewhere.

Entry

The closing price of the first trading day strictly after the filing date — the same entry rule as our backtest and our track record. If there is no trading day within five calendar days, there is no measurement.

Exit

The closing price on the measurement date, which is the most recent trading day in our price cache. You buy and you keep holding; there is no selling rule, because a filer’s sale is a separate decision we do not follow here.

The figure

Exit divided by entry, minus one. Gross, in the currency the share trades in, before costs, taxes and dividends — the same convention as the track record. Averaged unweighted across filings: one filing counts once, whatever it was worth.

The windows

Six months, one, two, three and five years, counted back from the measurement date on the entry day. They are nested, so the five-year figure contains the three-year one. Identical filings on the same day for the same company count once. Where a window reaches further back than the register we hold for that filer, the page says so next to the figure instead of letting the label stand for a depth we do not have.

The reference next to it

Every figure carries a second one beside it: what an index did over exactly the same days. Not the index over the same calendar period — for every filing we take the index level on that filing’s own entry day and on the measurement date, and then average those changes exactly as above: unweighted, one filing counted once, the same window, the same rounding. The reference is the AEX for people who report in the Dutch register, and the S&P 500 for the funds and for members of Congress, because both of those registers are American by definition — one reference series per index, not two. It has its own count, because a filing we measured is not always one the reference could follow — a reference series that starts later than the entry day, or was not trading around it, produces no figure and is listed with its reason in the same way. The two numbers stand next to each other. We do not subtract them, we do not rank on the difference, and neither number is colored.

Why the reference is not measured on identical terms

Both references are price indices: the dividends their constituents paid are not in them. The follow-along figures use closing prices adjusted for corporate actions, which reinvest dividends as far as the price series carries them. Over long windows that difference is real and it runs one way — measured on the S&P 500 series we hold, from January 2015 to August 2026 the price index rose 276.1% while the same series including reinvested dividends rose 355.4%. We chose the index over a tradable fund on purpose: a purchasable product placed next to a filing reads as an alternative on offer, and this block offers nothing. So the reference is the more conservative of the two figures, and it says so here rather than in a footnote.

Where the figures are ordered, and the one rule that governs it

On the three index pages — the fund filings, the insiders we measure and the members of Congress — the highest measured figures are shown at the top, with the full sortable list underneath. That is a deliberate change of 18 August 2026 to our own earlier presentation rule, and it is limited to the ordering: the count, the window and the reference stand in the highlighted row itself, there is no judgement wording, no good/bad coloring and no difference between the two numbers anywhere. The soft background behind a percentage is the same whether that figure is positive or negative: it frames the number, it does not grade it.

One threshold decides what may be highlighted — at least 3 measured observations in the window on display — and it is written on the page. The number here is the same constant those pages read, so the two cannot drift apart. Whoever falls below it is not hidden: they are in the list underneath with their count next to them, marked as below the threshold. The threshold exists because at one or two observations an average is one or two numbers, and a figure that cannot rise above its own noise does not belong at the top of a page.

Where that list carries a search box, it narrows only that list — the figures at the top stay the highest of all of them, because “the highest measured” is a statement about everyone we measure and not about whoever you searched for. While a search is active the page says so, and says that the counts around it still cover everyone.

The estimated weighted variant

Next to the unweighted list on the members-of-Congress page stands a second block that weighs the same measured purchases by an estimated point value taken from the amount band on each line: the midpoint of the band, so $15,001 - $50,000 counts as est. $32,500; a line that carries an exact amount counts as that amount; a band without an upper bound carries no estimate and is counted out with its reason. The figure is the sum of estimate × price change divided by the sum of estimates — a percentage, so a member with many small filings does not weigh more than one with a single large one. Two identical lines on the same day each carry their own weight: two reported amounts are two amounts, which is why this block counts lines where the unweighted one counts deduplicated filings.

The House form carries no amount, only a band, and the lowest band is a factor of fifteen wide. Every figure in that block is therefore an estimate, is marked est. wherever it appears, and stands next to the same calculation with the lower bound of every band as the weight, so a reader can see how much the choice of point value moves it. Purchases only: sales are reported and shown, but they are not in the figure and no label suggests they are. Everywhere else on this site a band remains exactly the text the form carries. This variant has its own version counter, separate from the unweighted definition, and its changes are recorded in the changelog like every other one.

Why the count next to it is the important number. Half of all filers in the register report two or fewer purchases in five years. An average over two figures is a fact about two figures, and nothing more. That is why the count stands next to every percentage, why every filing appears in its own row underneath, why nothing is highlighted below 3 observations, and why nowhere on this site does a figure like this carry a label, a color or a word that says whether it is good.

The filing date is an assumption for older filings. The AFM does not publish when a filing became public. For filings from before July 2026 we use the transaction date plus three business days: the last moment the law allows. That is the conservative end — the real publication can only have been earlier — but it is an assumption, and it is stated on the page itself. For funds the date is the day the 13F reached the SEC, and that form describes holdings that are already at least 45 days old.

The block is shown on the pages of chief executives and chairs of the management board, on the pages of filers at the fifteen largest listed companies we cover, and on the pages of the funds whose 13F filings we read. It is calculated for everyone; which pages show it is a reviewed list, not a property of the figure. A fund page shows one block per registered filer identity: two identities of the same house are never added together, because the filings do not say which shares were counted twice.

What the share price did since a filing

Every reported line from a congressional filing carries two figures where we can work them out: what the share price did since the filing became public, and what the S&P 500 price index did over exactly the same days. Two facts, side by side. We do not subtract one from the other and we do not call one of them better; the tint behind a figure is the same whether it is positive or negative.

The definition, in full:

From which day

From the day the register of that chamber published the filing — never from the day of the transaction. A member may file up to 45 days after trading, and the median gap in our data is 28 days; on the day of the transaction nobody outside could have known. Measuring from then would credit a figure to information that was not public. The entry price is the closing price of the first trading day strictly after the filing date, and if there is no trading day within five calendar days there is no figure — the same entry rule as our backtest, our track record and the block above.

Until which day

The most recent close we hold for that share. If a series has stopped — a company taken private, a symbol renamed — the line shows no figure rather than a price from months ago presented as today’s.

Which share

The form carries a ticker as the filer typed it, and that is not an identification: it is not validated, not dated and not linked to any register. We match it to the symbol the SEC itself publishes, and only where that match is unambiguous — one form, exactly one symbol. Every other case goes into a reviewed file with the evidence on the line and waits for a human judgement. A line whose ticker is not confirmed says so and carries no figure. Nothing is guessed, and an empty result is not proof that the security does not exist: the SEC list is not exhaustive.

The reference

The S&P 500 price index, aligned on the same two days (the last index close on or before each of them, at most five calendar days back). Because it is a price index, dividends paid by its constituents are not in it. The share figure uses the adjusted close, so splits and distributions are in that one. The two are therefore not measured on identical terms, and that is a limitation of the reference rather than a correction we apply.

The figure

The most recent close divided by the entry close, minus one. Gross, in dollars, before costs and taxes. It is not a return you could have had: you cannot buy at a closing price, and the line may be a sale, a spouse’s account or a partial disposal. It says what the share did, and nothing about what anyone earned.

How a strategy is calculated

Two kinds of page use the same engine and the same assumptions, and they ask different questions. A member page asks what a portfolio would have done that bought whatever that member reported buying and sold whatever they reported selling. A strategy page asks the same about a rule over the whole layer — the ten companies most reported, or the ten whose insider purchases score highest — rather than about one filer. Both are back-tests on register facts. Neither is a recommendation, a prediction, or something anyone actually traded.

Three assumptions decide every figure on that page, so they belong here in plain language rather than in a footnote. Change any one of them and every number moves.

1. What it starts with

USD 100,000 in cash, the same for every strategy, on the trading day before the first purchase could have been made. That date is shown on the page as the back-test start date. The amount is purely a scale: double it and every dollar figure doubles while not a single percentage changes.

2. How much goes into each reported purchase

Every reported purchase gets an equal share of the portfolio. One reported purchase counts once, whatever the amount band says — the form gives a range, not a number, and turning that range into a weight would be a guess doing the work of a fact. On any day the holdings change, the portfolio is divided equally over whatever is open then; in between nothing is traded and the weights drift with the prices. That redividing is our assumption about the follower, not something the member reported: to buy something new with a fixed pot you have to sell a slice of what you already own.

3. What happens on a reported sale

The whole position in that company is closed, on the closing price of the first trading day after the filing was published. The form does not say how many shares were sold — again a band, not a number — so a partial sale is read as a full one. That is the cautious side: the portfolio never keeps holding something the register says was disposed of. A sale of something this portfolio never held (bought before our data begins) is counted and reported, not silently ignored.

The rest of the definition:

When it buys

On the closing price of the first trading day strictly after the day the filing was published — never the day of the trade itself. A member may report up to 45 days late, so the trade is typically weeks old by the time anyone could read about it. Whatever happened in between is not in this curve, as gain or as loss.

Costs

0.1% of every dollar traded, each way. Opening a position and later closing it therefore costs 0.2% in total. That is lower than the 0.4% round trip used by the event studies elsewhere on this site, and the difference is deliberate: those measure a single entry and a single exit months apart, while a strategy here can rebalance every week. At that frequency the cost assumption stops being a rounding detail and becomes the largest single drag on the result, so it is set to what trading a liquid U.S. listing actually costs rather than to a conservative catch-all. No tax and no interest on cash.

Currency

Every curve is measured in USD, so that two of them can be read side by side. A price quoted in another currency is converted at the European Central Bank reference rate of that day — the last published rate on or before it, never a later one — and so is the index it is compared with. That matters for what the numbers mean: for a strategy that holds European listings, part of the return is the exchange rate rather than the company. A day with no usable rate is dropped from the series and counted, never filled in at one-to-one.

What it is compared with

An index, drawn as a second line from the same starting amount on the same day: the S&P 500 for the strategies and members that trade U.S. listings, and the AEX for the one that holds European ones. Which index a page uses is shown on that page. Matching the index to where the portfolio actually trades is the point: a European portfolio measured against a U.S. index would be measuring two things at once. It is placed beside the strategy and never subtracted from it. Beta, alpha, information ratio and Treynor are a different matter: those are named measures with their own definitions that use the benchmark as an input. Alpha is not “return minus benchmark”.

What the key figures mean

CAGR is the annual rate that would turn the starting amount into the ending amount; it is left blank under a year of history, because annualizing four months is an extrapolation, not a measurement. Max drawdown is the deepest fall from any earlier peak. Volatility is the spread of the daily returns, scaled to a year. Sharpe and Treynor divide the return by that spread and by beta respectively; both assume a risk-free rate of zero, because we hold no interest-rate series — and that choice makes both look better than a real money-market rate would. Win rate is the share of all positions with a positive return after costs, open ones valued at the last day of the curve; a position that returns exactly zero counts in the total but not as a win.

When a figure is blank

It is blank because it could not be calculated, never because it is zero. Below twenty days of history there is no volatility, beta, Sharpe, information ratio or Treynor; without a benchmark line there is no beta and no alpha. An empty box says “not measured”.

Which members get a strategy

Those with at least 5 reported purchases from the past 3 years — counted from the date of the pipeline run — that we could match to a listed company and give an entry price. Below that a curve is in practice the share price of one or two companies with a politician’s name above it, and without the window a member who stopped trading years ago would still rank on old entries. Every member is calculated either way, over their full history; the threshold and the window only decide what is shown.

What is missing from the portfolio

Anything the member owned before our records start, anything reported in a form we could not match to a ticker, and everything that is not a listed share — bonds, funds, options. The curve is what the reported purchases did, not what the member owns.

What this data cannot do

• Not everything can be matched to a price. Missing ISINs, bonds and unlisted entities fall outside the price series. Those cases are counted and logged, never quietly dropped.
• Vanished companies stay in. Companies taken off the exchange remain in the database; they do not disappear from the history, because then the past would look rosier than it was.
• Old filings lack a filing date. For those cases we conservatively assume the latest permitted moment: the transaction date plus three business days. For filings that arrive live we measure the real delay, and so test that assumption.
• Everyone sees the same state at the same moment. The site shows the latest state each pipeline run has written — there is no publication delay, and no path by which an alert or newsletter can run ahead of the site.

Questions about the exact weights, thresholds or version history? They are in the project documentation. Feel free to get in touch.