Alpha Dive · Short Engine

Selection alpha, measured against one named benchmark

The realised months the Short engine actually ran — taken from the performance workbook, not simulated, with the months the source could not produce left as gaps rather than filled in.

BenchmarkS&P 500 Equal Weight / SPW Index

01 · Headline figures

Compounded, trailing, and the record behind them

Selection alpha is measured as the benchmark’s return less the shorted basket’s return, so a favourable month reads positive — the opposite of the source workbook’s own sign. Every figure below uses that convention; the raw workbook delta is never shown.

BenchmarkS&P 500 Equal Weight / SPW Index

Compounded selection alpha

+4.80%

Since Aug 2023 · 36 observed months

Positive months

58.3%

21 of 36 observed months favourable

Average monthly alpha

+0.15%

Arithmetic mean — reported alongside the compounded figure, never instead of it

02 · Cumulative selection alpha

The compounded path, month by month

Geometric compounding of every observed month’s selection alpha — never a sum, and never the two absent months treated as flat. The line breaks where the source has nothing to report rather than running through it.

BenchmarkS&P 500 Equal Weight / SPW Index

03 · Monthly selection alpha

Every observed month, on its own

The same 36 readings the cumulative line compounds, shown individually and diverging around zero — favourable months above the line, unfavourable below it, coloured for sign and labelled with it too.

04 · Short basket vs. benchmark

What each leg actually did

Selection alpha is a difference; this is the two things being differenced. The short basket’s own monthly return against the benchmark’s, on the same axis and the same 36-month grid.

BenchmarkS&P 500 Equal Weight / SPW Index

05 · Distribution of monthly alpha

How the 36 months are shaped around zero

21 of the 36 observed months were favourable — the shape below is what that 58.3% looks like spread across the full range actually realised, from the worst month to the best.

06 · Security-level dispersion

What the basket-level number is averaging over

One dot per shorted name, per month: the same benchmark-minus-position convention applied security by security, so the single monthly figure above can be seen for what it is — a summary of names that did not all move together.

Identified by security, not by ticker — 1075 of the 1804 source rows carry a blank ticker, while security is populated and unique per date throughout.

07 · Illustrative overlay

A simulation built on top of the alpha above, not another view of it

A 10% net-long simulation, shown separately and styled apart from every chart above it — so it cannot be mistaken for a second reading of the same headline figure.

Illustrative, not alpha

Illustrative 100/90 Overlay

100% long the benchmark and 90% short the basket — a 10% net-long exposure simulation. It is not pure alpha and must not be read as one: it blends the benchmark’s own market return into the line, which the selection-alpha figures above this panel deliberately net out.

Overlay compounded return

+8.16%

Overlay equity line, latest

108.16base 100

100% long SPW minus 90% short basket; 10% net long exposure Shown for completeness only — the KPIs and charts above this panel are the engine’s measured alpha; this overlay is a simulation built on top of them.

08 · Methodology & data quality

What every number above does, and does not, claim

The engine handoff for this page came with specific, verified caveats. They are stated here in full rather than left for a reader to discover by cross-checking the source.

Sign convention: The source workbook stores short basket − benchmark, where a negative number is the favourable outcome. Every figure on this page uses the inverted convention instead — benchmark − short basket, positive is favourable — and the workbook’s own value is never rendered.
Calendar coverage: Every month from Aug 2023 to Jul 2026 is present. An earlier delivery of this source was missing two months; this page checks the gap fields on every load rather than assuming the source stays clean, so this note updates itself if a future delivery reopens one.
Security file coverage: Every month in the monthly summary file is also present in the security-level file — no gap unique to the security data, checked by comparing the two files directly on every load rather than assumed.
Blank tickers in the security file: 1075 of 1804 security-month rows carry a blank ticker. The dispersion chart identifies names by security instead, which is populated and unique per (date, security) throughout the file.
Overlay weighting: The source workbook’s own chart title stated a different short weighting than its formulas actually compute. The formulas — and this page — use the 90% figure as the one source of truth; the chart title’s figure is not used anywhere here.
Weekly performance fields excluded: The security file also carries first-week and last-week performance columns. They are not consistently populated across the full history and are excluded from this page for that reason — every figure shown here is built from the one-month performance and benchmark return columns only.

The Short and Long engines are separate models with separate benchmarks and separate methodologies. Nothing on this page is summed, blended, or compared against the Long engine — there is no combined figure, and none is shown.