How to use a macro forecast table, and why I built one

Every month a handful of US data releases move rates, the dollar and equities in a matter of seconds. I built a free table that puts my forecast for each one next to what the prediction market expects, so you can see in one glance where a surprise is most likely.

By Chirag Mirani · Published Saturday, November 7, 2026

What is the forecast table?

It covers six releases that matter most for near-term market moves:

Why did I build it?

I have spent about twenty years in quantitative finance and macro investing. Before every release I was doing the same thing: checking the calendar, checking what the market expected, checking my own numbers, and trying to hold all of it in my head at once. That is exactly when behavioral biases creep in. You anchor on the last print, or on the loudest headline.

So I put it all on one page. The point is not to replace your own view. It is to give you one honest, consistent second opinion on the next print, every time.

How do you read it?

1. Find the next release. The top row is the next event, with its day, date and time in US Eastern.
2. Compare the two forecasts. Kalshi consensus is what traders with money at stake expect. The house forecast is what my models expect from the data.
3. Check the surprise-risk flag. "Elevated surprise risk" means the model sees a real chance the print lands away from consensus. "Review required" means my forecast and Kalshi are far enough apart that I check official guidance and the news before trusting either.
4. Check the last actual. It shows where the series is coming from, with a link to the official release.
5. Score it afterwards. Once the number is out, the track record page settles both forecasts against the actual.

A simple example

Say core CPI is due on Tuesday at 8:30 AM Eastern. Kalshi expects 0.20% for the month and my model says 0.23%. My forecast sits above the market, so the model leans toward a hotter-than-expected print. If you are positioned for a soft number, that is a reason to check your risk before the release, not after.

Who is it for?

• Traders and portfolio managers who want a quick read on where the surprise risk sits before a print.
• Prediction-market traders on Kalshi who want an independent number to compare against the market price.
• Analysts, students and journalists who want a clean calendar with dates, times, consensus and official sources in one place.
• Developers and AI tools that need the data in a feed. The full table is at dashboard_data.json.

How was it created?

The inputs are public: the Bureau of Labor Statistics, the Bureau of Economic Analysis, the Department of Labor, ADP, Federal Reserve Economic Data (FRED), and Kalshi for market consensus. Each release has its own model, and they talk to each other. For example, a fresh jobless claims number feeds the payrolls and unemployment forecasts, so those rerun when claims change. For core PCE I build a bridge from the CPI components, because most of PCE is already known once CPI is out.

Every evening the models rerun only where their inputs have moved, and publish their forecasts. Twice a day on weekdays, the site pulls the latest Kalshi prices and the latest official numbers, and rebuilds the page. I don't publish the model code, but the outputs are all public and scored in the open.

How are the results?

Every forecast is snapshotted before the release and scored after it. There is no backfilling. The record starts on the day each snapshot was taken, so what you see is what was actually on the page beforehand.

As of Saturday, November 7, 2026. The track record page always has the current numbers.

To be straight about it: Kalshi is a tough benchmark, and on simple closeness it is roughly a coin flip between us so far. Where I think the table earns its place is on direction. It has been right more often than not on which side of consensus the print lands, and that is the part that matters when you are deciding whether to lean into a release or step aside. Seventeen releases is a small sample, and I will keep scoring it in public as it grows.

Where can it go wrong?

The models learn from the past, so they cannot see genuinely new information: a hurricane, a strike, a government shutdown, a change in how a statistic is measured. When the table misses badly, I treat that as a signal in itself. Usually it means something new is driving the data that the market hasn't fully priced either.

It is a guide, not a trading system, and it is not investment advice. Use it as one perspective next to your own.

Frequently asked questions

Go to the table

The next release, its time, my forecast and the market's are all on one page.

Built and maintained by Chirag Mirani. Nothing here is investment advice.