How the 13F consistency score works

Every manager Vestiio ranks carries a consistency score between 0 and 100, computed from public SEC Form 13F filings. This note sets out how, with the weights and thresholds the ranking uses and a worked example. A shorter version of the same explanation sits on every public manager page.

One observation per filing

A 13F lists the US equity positions a manager held at the end of a quarter, and it is filed up to 45 days later. For each filing the score takes the positions it reports, weighted by their share of the reported book, and follows them for 90 days from the day the filing became public. The result is one observation: what the disclosed book returned for someone who could first read it.

• Only positions that map to a listed security with daily prices count. Options, exchange-traded funds and other funds stay out.
• A holding whose price moved below minus 95% or above plus 300% over the window is dropped as a probable data error, and its weight counts as unpriced.
• An observation counts once the full 90 days have passed and at least 70% of the book was priced, and when its return lies between minus 95% and plus 200%. A filing made last month has no observation yet.
• One filing counts per quarter. A restated filing replaces the original, and an amendment that only adds holdings never counts on its own.

Compared with every other manager in the same quarter

A quarter’s return says little on its own, because a rising market lifts every book. So each observation is compared with those of every other eligible manager for the same quarter, and a quarter is used only when at least 20 managers have an observation for it. Each observation gets a peer percentile, from 0 for the lowest return that quarter to 100 for the highest, and an excess return, its distance from that quarter’s median.

Seven inputs

The score reads a manager’s 20 most recent observations and needs at least twelve. From them it takes:

• Peer percentile, averaged with more weight on recent quarters: each quarter counts 0.93 times as much as the one after it. Worth 45% of the score.
• Hit rate, the share of quarters above the median. Worth 25%.
• Top-quartile rate, the share of quarters in the best quarter of managers. Worth 15%.
• Coverage, scaled from 0 at an average of 70% of the book priced to 100 at 95% or more. Worth 10%.
• Evidence, the number of observations out of 20. Worth 5%.
• Bottom-quartile rate, the share of quarters in the worst quarter of managers, subtracted at 15%.
• Volatility of the excess return: each point of standard deviation above 20 subtracts a quarter of a point.

The sum is held between 0 and 100 and then pulled towards 50, by less as the evidence grows. Its distance from 50 is multiplied by the number of observations divided by that number plus eight, so twenty observations keep 71% of the distance and twelve keep 60%.

A worked example

Take a manager with 20 observations and these made-up inputs: a recency-weighted peer percentile of 62, above the median in 60% of quarters, in the top quartile in 30% and the bottom quartile in 15%, an average of 92% of the book priced, and an excess return with a standard deviation of 14 points.

Coverage scores 88 and evidence 100, and the volatility is under 20, so it subtracts nothing. The raw score is 62 × 0.45 plus 60 × 0.25 plus 30 × 0.15 plus 88 × 0.10 plus 100 × 0.05, minus 15 × 0.15, which comes to 58.95. Twenty observations keep 20/28 of the distance from 50, so the score is 56.4.

The same record over only twelve observations, the fewest a ranked manager has, scores 54.2. Evidence falls to 60, the raw score to 56.95, and twelve observations keep 60% of the distance. A short record has to be stronger to rank as high.

Which managers get a score

• Filers classified as active managers. Index managers, operating companies filing for strategic stakes, sovereign wealth funds, pensions, venture capital firms, private equity sponsors, and banks with their broker-dealers are left out.
• Managers with at least a quarter of the book in the ten largest positions, measured on their median filing. The score weights positions by value, and a book spread across thousands of names describes the market more than the manager.
• Managers with at least twelve observations that pass the checks above, and an average of at least 80% of the book priced across them. A shorter or thinner record keeps its holdings page and waits for more quarters.

A ranked manager’s confidence is high with 20 observations and an average of at least 90% of the book priced, and medium otherwise.

The one-year figure

Beside the score, each manager shows a one-year figure: the latest four completed 90-day windows, compounded. It accepts windows with at least 40% of the book priced, so it can exist where the score does not. With fewer than four windows it compounds the ones there are, and its confidence is low.

What a 13F leaves out

Cash, short positions, private assets, many derivatives, fees, leverage and every trade made between two filings. The score describes the disclosed equity book, followed from the day it became public. A fund’s actual returns can differ widely from it, and a low score is no verdict on the manager.

The score answers one narrow question: how the positions a manager disclosed went on to do for someone who read the filing the day it became public.