InsidersTradesSigma

Understanding the Top 50 Insiders by 3-Year Batting Average, Sigma Journal

Evaluating the Top 50 Insiders by 3-Year Batting Average

How reliable are insider buy rankings based on a 3-year batting average? This article examines the methodology and challenges behind such leaderboards.

By Sigma Research·May 17, 2026·12 min · 2,740 words

A good insider can look like a genius right up to the moment you count properly.

Top 50 insiders by 3-year batting average

The seduction of a leaderboard is obvious. Put names in rank order, attach a percentage, and the market starts to look like a batting cage. Some insiders seem unable to miss. Others, despite enviable titles and expensive tailoring, manage to buy stock at exactly the wrong moments.

The trouble is that insider trading data, the legal kind disclosed to regulators, is not a clean sports dataset. It is sparse, lumpy, full of role changes, repeated buys in the same drawdown, and contaminated by mechanics such as option exercises, automatic plans, tax sales, and family holdings. A 3-year batting average, defined here as the share of disclosed open-market buys that were followed by a positive forward return over a fixed horizon, can be informative. It can also flatter noise.

This article sets out how to think about a "Top 50 insiders by 3-year batting average" leaderboard when the objective is profile writing rather than numerology. The angle is intentionally narrow: hit rate, minimum five buys, names plus companies, with caveats carried in plain sight rather than hidden in footnotes like contraband.

Before we get to names, one blunt point. No article-specific query output was supplied for this piece. That means there is no defensible way to publish an actual ranked top 50 with percentages, company names, and ordering without inventing data. We do not do creative writing with leaderboards. So this piece does the next best thing: it gives the exact editorial and methodological framework for such a ranking, explains where it goes wrong, and sets out the fields that should appear once the live query is run.

Why "batting average" works, and why it cheats

A hit rate is attractive because it is intuitive. You do not need to explain annualisation, factor neutrality, or the emotional support required to read a t-statistic. If an insider bought ten times and seven of those buys were followed by a positive forward return, the reader understands the claim immediately.

That simplicity is also the first trap.

Hit rate is robust to outliers, but blind to magnitude

A batting average ignores whether the wins were tiny and the losses catastrophic. An insider who buys before six modest rebounds and one collapse can still post a handsome hit rate. In market terms, that can mean being directionally right more often than not while still destroying capital.

For profiling, this is acceptable if stated plainly. The article is about consistency of timing, not payoff asymmetry. But the distinction matters. A leaderboard by hit rate should usually sit beside, or at least acknowledge, complementary metrics such as median forward return, average excess return versus sector, and drawdown after purchase.

Three years is recent enough to be relevant, short enough to be noisy

A 3-year window has editorial appeal. It captures current management teams, current business models, and current market regimes better than a decade-long sample. It also tends to include at least one stress episode, which is where insider buying becomes more informative.

The cost is sample size. Even active insiders may only have five to ten qualifying buys in that period. That is enough for a screen, not enough for sainthood.

Minimum five buys is necessary, not sufficient

The threshold of five buys removes the most embarrassing cases, namely the executive who bought once near a local bottom and now sits atop the rankings like a statistical prank. But five is still a very small denominator. A 5-for-5 record is interesting. It is not proof of repeatable edge.

The ranking only works if the plumbing is boring

The best methodology is usually the one that makes readers yawn. That is a compliment. A leaderboard should be built from rules so dull and explicit that nobody suspects the editor of falling in love with a narrative.

Define the event properly

At minimum, each row in the ranking should be tied to an event definition with four elements:

Who: the insider identity, normalised across spelling variants and role changes.
What: an open-market purchase, with transaction codes harmonised across jurisdictions.
When: the filing date and, where available, the trade date.
Outcome: the forward return horizon and benchmark.

A common choice is to score a hit if the stock's total return is positive 90 calendar days after the trade date. A stricter variant asks whether the stock outperformed its local market index or sector over the same period. The latter is usually better if the aim is to identify insider timing rather than simple beta exposure.

De-duplicate clustered buys

Insiders often buy repeatedly over a short period, particularly during sharp drawdowns. Counting each purchase separately can overstate conviction and inflate hit rates if the stock then rebounds. There are two sane approaches.

First, issuer-window collapsing: treat all buys by the same insider in the same issuer within, say, 30 calendar days as one signal, using the volume-weighted average price. Second, event spacing: only count the first qualifying buy until the forward window expires.

Neither is perfect. Both are superior to pretending that four buys in eight trading days are four independent acts of prescience.

Weighting matters more than readers think

Should a €5,000 purchase count the same as a €500,000 purchase? For a pure batting average, equal weighting by event is defensible. For a conviction-aware profile, value-weighted supporting statistics are useful. An insider who buys frequently in token amounts is not the same as one who writes large personal cheques into weakness.

A clean article can keep the ranking equal-weighted and then add two side columns: total capital committed and median purchase size. That lets readers distinguish the enthusiastic nibbler from the serious accumulator.

Jurisdictional rules can quietly distort the table

Insider disclosure is regulated, not standardised. A cross-market leaderboard that ignores filing rules is like comparing sprinters while some are running uphill.

Europe is broadly harmonised, but not identical in practice

Under the EU Market Abuse Regulation, persons discharging managerial responsibilities, and closely associated persons, must notify transactions once a threshold is crossed, and issuers must make the information public promptly. In practice, data quality still depends on national implementation, issuer disclosure habits, and how cleanly vendors map transaction types.

France is a useful example because AMF disclosures are accessible and the market has enough mid-cap activity to produce meaningful insider-buy clusters. But even there, transaction coding and related-party nuance require care.

Market	Regulator	Rule	Deadline	Notes
FR	AMF	MAR Art 19	T+3	Managerial transactions disclosed under EU MAR framework; data quality depends on issuer notices and coding.
US	SEC	Section 16 / Form 4	T+2	Rich transaction coding, but Rule 10b5-1 plans and derivative transactions require filtering.
EU ex-FR	National competent authority	MAR Art 19	T+3	Broadly similar legal basis, uneven issuer formatting and dissemination.

Disclosure rules shape the observable dataset, especially for timing and transaction taxonomy.

What the Top 50 should look like on the page

A profiles article needs names. It also needs restraint. Once the live query is run, the ranking should present the top 50 insiders by 3-year buy hit rate, subject to the minimum-five-buys filter, with company names and enough context to stop the list becoming a hall of fame for statistical accidents.

Recommended columns

For each insider, include:

Rank
Insider name
Primary company
Role at time of most recent qualifying buy
Number of qualifying buys in the 3-year window
Hit rate
Median 90-day return after buys
Total disclosed value of qualifying buys
Most recent buy date
Sector
Country or primary listing market

If an insider traded across multiple issuers in the period, the article should either assign a primary company based on the most recent or most frequent issuer, or split the person into separate insider-issuer rows. The latter is usually cleaner.

Names plus companies, but not without context

A leaderboard row that reads "Jane Doe, ABC SA, 80%, 5 buys" is not enough. The profile note should tell the reader what kind of insider this is. Founder-chair? CFO? Independent director? Turnaround CEO? Family holding vehicle? These distinctions matter because information sets and incentives differ.

A founder with concentrated wealth and operational control is not the same signal as a non-executive director making periodic purchases. Both can be informative. They should not be treated as interchangeable.

A model row, because blank spaces are honest

Since the live query was not provided, the only responsible way to show the table structure is as a template. That is less glamorous than a fabricated ranking and more useful than pretending otherwise.

Rank	Insider	Company	Role	Qualifying buys, 3Y	Hit rate	Median 90D return	Total buy value	Last buy date	Market
1	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a
2	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a
3	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a

It is not pretty. It is accurate.

The caveats that separate a leaderboard from a parlour trick

Every ranking in this category should come with caveats prominent enough to annoy the marketing department.

Small samples create fake legends

Suppose an insider bought five times, all during a sector panic, and all five trades were followed by a relief rally. That person will rank highly. The result may reflect genuine timing skill. It may also reflect the fact that many cyclicals rebound after panic selling. Without a benchmark and some sector control, the leaderboard will reward courage, beta, and luck in unknown proportions.

Role changes and issuer events contaminate histories

An executive can move from one company to another during the 3-year window. A company can be acquired. A spin-off can change the return path. A rights issue can alter the economics of subsequent purchases. If these edge cases are not handled consistently, your table becomes a scrapbook of exceptions.

Buys are more informative than sells, but not always for the same reason

Academic literature has long found that insider purchases tend to be more informative than insider sales. Sales are often driven by diversification, tax, or liquidity needs. Buys involve putting more personal capital into the same risk already tied to the insider's career. That asymmetry is real. It does not mean every buy is a signal. Some are symbolic, some are board optics, some are confidence theatre at exactly the moment confidence is required.

Price impact and liquidity can flatter outcomes

In smaller names, insider buying itself can attract attention and tighten the register. If the stock rallies after disclosure, part of the measured "success" may be the market responding to the signal rather than the insider foreseeing fundamentals. From a trading perspective, that distinction may not matter. From an attribution perspective, it does.

How to read the eventual Top 50 without embarrassing yourself

A leaderboard is a starting point, not a portfolio.

Look for repeat behaviour, not just rank

The most interesting insiders are not necessarily those at the top of the table. They are those whose purchase patterns are coherent across time and circumstances. Do they buy after earnings disappointments but before guidance stabilises? Do they add in tranches during industry-wide stress? Do they buy only when valuation is compressed relative to history? Pattern beats percentile.

Separate conviction from ceremony

Some boards expect directors to buy stock after joining or during periods of scrutiny. Those purchases can be sincere and still uninformative. The tell is often size relative to compensation and existing holdings. A purchase that changes economic exposure is different from one that decorates a governance slide.

Compare within sectors and market caps

A 70% hit rate in volatile small-cap biotech is not the same animal as a 70% hit rate in a regulated utility. Sector structure, liquidity, and event cadence all affect the opportunity set. A good article can include sector tags and perhaps mini-leaderboards by industry. A better one will resist claiming that all hits are created equal.

Use the names as a watchlist for future filings

Profiles become useful when they point forward. Once you know which insiders have a credible record, future buys by those individuals deserve immediate attention. That is the practical value of the ranking. Not "copy every trade", but "know whose filings earn a second look".

Editorial standard for this specific article

Because the requested angle explicitly asks for "Top 50 insiders by 3-year batting average" with "names + companies", it is worth stating the publication standard plainly.

What we can say now

We can define the metric, state the filter of minimum five buys, explain the likely distortions, and specify the output fields required for publication. We can also note that the source corpus contains 162,000 filings, which is ample for a serious leaderboard once the article-specific extraction is run.

What we cannot say now

We cannot publish the actual top 50 names, companies, hit rates, or rankings because no live query output was supplied. Any such list would be invented. Invented rankings are very popular on the internet. They are less popular in journals that expect to be read by people who can count.

The exact query choices that should be locked before publication

To turn this into a finished profiles piece, the data team should confirm:

Jurisdiction scope, for example global, Europe only, or France only.
Eligible transaction codes, especially treatment of derivatives and plan trades.
Event de-duplication rule.
Forward-return horizon, for example 30D, 90D, or 180D.
Benchmark rule, absolute positive return or excess return versus market or sector.
Insider identity resolution across issuers and spelling variants.
Currency normalisation for disclosed transaction values.
Handling of delistings, mergers, and suspended securities.

Once those are fixed, the article can be populated with actual names and company profiles in a way that survives contact with an auditor, an issuer, and a sceptical reader. A rare trifecta.

What a finished version should add once the data arrive

The final published piece should not stop at the ranking table. It should include short profile capsules for perhaps the top 10 to 15 insiders, each around 100 to 150 words, covering role, company context, buying pattern, and one caveat. For example:

whether the buys clustered around a capital raise,
whether the insider is a founder or hired executive,
whether the company sits in a sector with high rebound propensity,
whether the purchase sizes increased over time,
whether the person has a longer history outside the 3-year window.

If the data support it, a small time-series block showing the count of qualifying buys by year would also help readers understand whether the leaderboard is built from a broad opportunity set or from one crisis-heavy interval.

The article should also consider a companion chart, not necessarily in the same piece, showing how hit rate changes as the minimum-buy threshold rises from 1 to 3 to 5 to 10. If the names at the top remain stable as the threshold tightens, confidence improves. If the leaderboard reshuffles violently, the "top 50" is probably a top 50 of who happened to be active during a convenient patch of volatility.

That is the real editorial question. Not who ranks first, but whether first place means anything durable.

The next concrete step is simple: run the article-specific query with the minimum-five-buys filter, lock the event definition and benchmark, then publish the top 50 with sample sizes and company context attached. The open question is the one that matters more than the rank itself, namely whether the same names still dominate once clustered buys are collapsed and returns are measured against sector benchmarks rather than a forgiving rising market.