At the foot of the Performance overview sits a calendar period returns table with six tiles. On the capture I am working from they read MTD 0.00%, QTD -16.13%, YTD -38.59%, 1 year -38.59%, 3 year -38.59% and since inception -38.59%. Four of the six are the same number.
Directly beneath them is the line that explains it. Observations 86, trades per year estimated at 327, days span 96, risk-free 4.5%. The account has 96 days of history. There is no three-year return to compute, so the tile shows the longest span available, and the label on it is aspirational.
Six tiles, one number, ninety-six days
Any calendar table has three options when the requested window exceeds the available history. Show a blank, show an error, or show what exists. Blanks read as broken to most users and errors read as broken to all of them, so showing the available history is the sensible default and every reporting stack I have worked with does some version of it.
It is also a reporting hazard, and the hazard is structural rather than a criticism of the choice. The label and the content of the tile no longer agree, and the thing that reconciles them lives in a footer in smaller type. Screenshots crop footers. Slides drop them. A figure lifted from a tile labelled 3 year travels perfectly well into a document where nobody can see that it was 96 days.
The tell is right there in the table, which is that four adjacent tiles are identical. Identical period returns across different labels is the signature of a padded table, and once you have seen it once you will spot it across every reporting surface you use.
The annualisation next to it is the same hazard wearing a suit
The footer figure for trades per year, 327, is an extrapolation. Eighty-seven trades over 96 days scaled to a calendar year gives you a rate of roughly that magnitude. As a capacity input it is genuinely useful, because if you are modelling commission drag or exchange rate limits you need a rate rather than a count.
As a headline it is the same error as the three-year tile, and it is more dangerous because it does not look padded. Nothing in the number 327 announces that it was built from 87 observations. A reader assumes a year of activity because the unit says year.
The general rule follows directly. Any figure whose unit implies a period longer than your observation span is an extrapolation, and extrapolations get a different verbal treatment from measurements. Say the rate implied by the observed period is roughly 330 trades a year, on 87 trades across 96 days. That sentence is bulletproof. The bare number is not.
The disclosure rule for a short track record
Six rules cover almost every case, and they are worth adopting as house policy rather than deciding case by case under time pressure.
- Never quote a period return whose label exceeds the observation span. When somebody asks for the three-year figure, the answer is that the account has 96 days of observations, followed by the since-inception figure and the dates. That answer is shorter than the alternative and it ends the topic.
- Publish the span and the observation count adjacent to the figure, in the same object. Not in a footnote, not in an appendix. This page prints observations 86 and days span 96 under the table, which is the right instinct and the wrong typography for anything that gets screenshotted.
- Do not annualise a sub-year return for external consumption at all. Internally, label it as implied.
- Prefer monthly granularity below twelve months of history. The monthly table on the Distributions tab reads April -0.4%, May -20.5%, June -8.1% and July -16.8%, and that sequence carries more information than a single -38.59%. It also exposes that April is nearly flat, which is what a partial first month looks like, and that August is blank.
- Distinguish a zero from an absence. The MTD tile reads 0.00% while the monthly table shows August empty. A zero is a claim that the period happened and produced nothing. An absence is a claim that there is no data. They are different statements and only one of them can be true at a time, so establish which before the figure goes into a report.
- State the composite definition and the flow treatment alongside any since-inception figure. Over 96 days a handful of days dominate the result, and if the underlying series is a raw balance carrying deposits, the number is measuring something other than the strategy.
Two aggregations of the same span will not agree, and that is normal
There is a second figure on this page worth handling carefully. The calendar table reads -38.59% for the full span while the monthly table on the Distributions tab totals -45.8%, and the asset class table on the overview shows the crypto row with a total P&L of -45.82%.
I am not going to assert which of those is the right description of the period or how they reconcile, because I cannot verify it from the outside. What I can say is that two aggregations of one span differ for ordinary reasons all the time. A chained sequence of period returns and a sum of per-trade P&L percentages are different arithmetic on different populations. A return computed on an equity series includes days with no trades and days with open marks, and a trade sum does not. Windows and scopes can differ between panels even when both are visible on one screen. And a per-trade percentage and a period return are denominated against different bases.
The practical instruction is not to pick the one you prefer. It is to be able to say which construction your reported figure uses before anyone asks, and to have the export that demonstrates it. The reviewer's question is never why do these differ. It is which one are you quoting and why, and that question has a good answer available.
What a reviewer will ask, and what you should already have on the shelf
Four artefacts, prepared once, cover the entire conversation about a short record.
A dated inception memo stating the first day of live capital, the composite definition, and what was included or excluded and why. A monthly return series with the flow treatment stated. A count of independent observations alongside the span, because 86 daily observations across four months is not the same evidence as 86 monthly observations. And a written statement of what the track record is not yet long enough to support, which is the artefact most people skip and the one that buys the most credibility in the room.
That last one deserves the emphasis. A manager who volunteers that 96 days cannot distinguish skill from a single favourable or unfavourable regime is a manager whose other claims get taken at face value. A manager who quotes a three-year tile built on a quarter of data has spent that credit before the meeting started, and the tile that let them do it was not hiding anything. The footer said 96 days the entire time.