I trade Nasdaq futures, and for a long time I worked from a feeling rather than a number. CPI was “big”. JOLTS was “probably nothing”. I sized positions off that feeling, and when a release did something I had not expected I told myself it was unusual, without ever checking whether it was.
Every economic calendar I could find tells you what a report printed and when the next one is due. None of them tell you what the market actually did afterwards. That is the only part I care about as a trader, so I built it.
The first thing it did was correct me. I had payrolls filed as the biggest mover on the calendar. It is tied with CPI on the median and its worst days are smaller. I had retail sales filed as background noise, and its biggest reaction is more than four times its median, which makes it the kind of release that is nothing until suddenly it is not. Neither of those was a hunch I would have given up without seeing the count.
Release dates and times come from the publishing agencies themselves - the Bureau of Labor Statistics, the Census Bureau, the Bureau of Economic Analysis and the Federal Reserve. Using their own schedules rather than a calendar rule matters: payroll reports have landed on dates no rule would predict, and a date that is wrong by a day measures the wrong minute of the wrong session.
Prices are one-minute bars on Nasdaq 100 and S&P 500 index CFDs. They trade around the clock, which is why they capture the 08:30 ET releases that land an hour before the equity market opens. They track the futures closely but are not the CME contracts, and every figure here is a percentage of the index level rather than a contract point.
The figures themselves are read from two independent sources, and where both have a number for the same release the two are compared at every rebuild. A disagreement is flagged and the agency's figure is the one shown. That check exists because I do not trust my own parser, and it has already earned its place. See below, and the methodology page for every definition in one place.
The headline figure on each page is the swing: the highest price minus the lowest price in the thirty minutes after a release. It is deliberately not the net change. A report can leave the index exactly where it started after moving half a percent in both directions, and the net number would call that a non-event while anyone holding a position would disagree.
Thirty minutes is a choice, not a law, and it is worth knowing what it hides. Roughly half of a CPI reaction is already over sixty seconds after the print. FOMC is the opposite: the median move at two hours is more than double the move at thirty minutes, so for that one release the half hour mark is closer to the middle of the story than the end of it.
I am listing these because a site full of numbers should tell you where its numbers have been wrong, and because both mistakes are the kind that stay invisible until something forces them out.
The word “unchanged”. The parser reading the BLS releases treated the phrase “changed little” as a reading of zero. At the BLS that phrase does not mean flat, it means the change was not statistically significant. Three payroll reports that actually printed +73,000, +22,000 and +64,000 were recorded here as zero before the cross-check against the second source caught the disagreement. Only a literal “unchanged” is a real zero, and that is now the only thing the parser accepts as one.
Releases that share a minute. Some reports print at exactly the same time as another. GDP and Core PCE land together; retail sales sometimes lands with PPI. On those days the move belongs to both releases and to neither, and counting it twice flatters the quieter one. GDP looked about 1.5 times larger on its shared days than on the days it printed alone, and none of that was GDP. Every figure on this site now excludes shared-minute releases, which moved GDP from a respectable mover to one of the two quietest things on the calendar.
No red and green for direction. A hot CPI and a hot payrolls print land on the same side of the arrow and mean opposite things to the market. Colouring one of them “good” would be editorialising, so direction is shown in plain ink with the word next to it.
No “beat” or “miss”. Those words carry a judgement about whether a number was welcome. The site shows the figure, the forecast where one exists, and the distance between them, and lets you decide what that means.
Counts, never forecasts. Nothing here predicts anything. Every page shows how many observations it is built on so you can decide for yourself whether a pattern across twenty three releases is worth anything. Some of them are not, and the pages say so.
The whole site is rebuilt once a week, on Sunday, from the agencies’ published schedules and a fresh pull of price history. Actual figures appear on the board within minutes of a release rather than waiting for the weekly rebuild, so a number that printed this morning is on the page this morning. Nothing is written by hand between rebuilds, which is why every statistic on the site moves as the record grows.
These are records of what happened, not forecasts. Sample sizes run from twenty three to thirty six observations per release, which is enough to see a direction and not enough to settle an argument, and every page shows its own count. Consensus forecasts are not published by the agencies and are read from a calendar, so they are the weakest input here. Prices are index CFDs rather than the CME contracts. Nothing on this site is investment advice, and I am not a licensed adviser.
Corrections are genuinely welcome, particularly if you have measured any of this yourself and got a different answer. [email protected]
Releases: Core PCE Price Index Consumer Price Index FOMC Rate Decision GDP JOLTS Job Openings Nonfarm Payrolls Producer Price Index Retail Sales