Why the Order Book Empties Before NFP

Helious Desk · 4 August 2026 · 7 min read

Every first Friday of the month the book empties a few minutes before 8:30am ET. Depth that was there a minute ago is gone. Most traders take it personally and assume someone is coming for their stop. Nobody is. The orders belong to market makers protecting their own quotes from a number nobody can price sensibly, and the orders go back in once the number is public.

1. What does it mean when the order book empties?

The resting limit orders on the ladder have been cancelled. A minute before the release the book looks healthy, size at every level. By the time the number lands most of that size has gone and the bid and ask sit further apart. Nothing traded and nothing broke. The people who placed those orders took them out, and they fully intend to put them back.

What the thin book then does to price is in what happens to the DOM during a news release. This post is about why it empties at all, and why payrolls is the worst of them.

2. Why does it happen specifically before NFP?

Payrolls is scheduled to the second, so everyone knows when the risk arrives and can step aside before it does. It moves rate expectations hard and directly, which is what makes the price move big enough to bother about. And it is not one number. Payrolls, the unemployment rate, average hourly earnings and revisions to the prior two months all land in the same second, and they can point in different directions.

That last part is what separates payrolls from everything else. Even after the number prints, fair value is still genuinely uncertain, because the market has to work out which of the four to trade. On CPI one number dominates and confidence comes back faster. On payrolls it does not, so the book stays thin for longer. The payrolls explainer covers how the release is put together.

3. Who is actually pulling their orders?

Mostly automated market makers and liquidity providers, who supply most of the resting size in liquid futures. Their business is quoting both sides continuously and earning a small spread across a huge number of trades, so they have no view on payrolls and no interest in you in particular. Banks and funds with working orders pull too. But the collapse in depth you can see is mainly the automated quoting standing aside for ninety seconds.

4. Why pull the orders instead of just widening the spread?

Because the loss a market maker is trying to avoid is bigger than any spread they could charge for it.

A limit order is a free option

Leave a resting order in the book and you have given everyone else the right to trade against your price whenever it suits them. Normally that is a fine bargain, since most of the people who take it are trading for reasons that have nothing to do with your price, and you collect the spread. Around a scheduled release it stops being a fine bargain. The only people exercising the option are the ones who already know the number is about to move against you, so you get filled at exactly the moment you are about to be wrong, every time.

Widening does not fix it

The obvious answer is to quote wider and charge for the risk, and it does not work. The gap on a genuine payrolls surprise can be many times the widest spread anyone would sensibly show. What you earn on the spread is small and capped. What you lose being run over is neither. Quote wide enough to actually cover it and nobody trades with you anyway.

So they cancel. Pricing something that cannot be priced sensibly costs more than standing aside, and measured in basis points the maths is not close.

5. Is this manipulation or stop hunting?

No. Plenty of people believe otherwise, and it makes them trade worse. Nobody is removing liquidity to hunt your stop. The behaviour is defensive rather than predatory, the people doing it are protecting their own quotes, and they do it at the same scheduled moment every month wherever anyone's stops happen to be sitting. Most of them are automated systems that do not know where your individual order is and would not care if they did.

It feels targeted because the effect on you is real. Your stop does fill badly. But that is the absence of anyone willing to take the other side, not the presence of someone trying to reach you, and the difference matters, because one of those you can plan around and the other just makes you angry. The release times are published in advance on the economic calendar, which is not how a conspiracy usually operates.

6. What does the timeline look like around 8:30?

It runs roughly the same way every month. From a few minutes before, depth starts thinning as working orders are withdrawn, which is easy to miss unless you are watching for it. In the final thirty seconds the withdrawal accelerates and the spread visibly widens.

Then the number lands. It reaches machines first, and price moves through whatever levels are left, which is why it appears to jump rather than travel. In the seconds after, the first move is often partly retraced as the detail gets read, and over the following minutes liquidity rebuilds as participants regain confidence about fair value. An alert a few minutes ahead is the cheapest way to stop that sequence surprising you.

7. Why can I not just be faster?

Because the opening part of the reaction is a machine contest measured in fractions of a second, decided by co-located systems and direct data feeds rather than by anyone reading anything. A human seeing a headline, forming a view and clicking is not competing in that window and never will be.

Better to accept that than fight it. Being fastest is not available to you. Being right about whether the number was a genuine surprise is, and so is waiting until the book has refilled before you do anything about it. That is where a discretionary trader adds something a machine is not already doing better.

8. Does the same thing happen in other markets?

Yes, and it happens across the whole complex at once rather than one market at a time. Treasury futures thin out, and the front end of the curve usually reprices first because it is the most direct read on the policy path. Major currency pairs thin the same way. Equity index futures thin roughly in proportion to how deep they normally are, so the E-mini generally holds up better than the Nasdaq contract, which is thinner and travels further for the same order size.

Watching the 2s10s spread afterwards tells you whether the market repriced the Fed or repriced growth. A quiet release such as weekly jobless claims barely disturbs any of it, which is part of why it is a gentler place to learn.

9. What does this mean for how I trade payrolls?

Your stop is not what you think it is during the print. A standard stop becomes a market order and fills at whatever price exists in a thin book, so being flat is the only real protection against a gap. Your normal size does not fit either, because the same position moves price much further when depth is missing, and the range that is normal for payrolls is not the range your risk settings assume. And the first move is the least trustworthy part of the whole event. It happened in the thinnest conditions of the day, on the least complete information.

Waiting for the spread to normalise costs a few ticks and removes most of the risk. If you are on a funded account, this is not optional: one gapped stop on a trailing drawdown ends it, which we went through in which news to trade to pass prop firm evals. And if payrolls simply is not worth the trouble for your account, NFP or unemployment claims makes the case for the quieter alternative.

10. How do I tell whether the number actually mattered?

Not from the ladder, which shows you the reaction and never the cause. A payrolls print that landed exactly on expectations and one that genuinely shocked the market produce the same thin book and the same fast first move. On the ladder they look identical, and only one of them holds.

To tell them apart, compare the release to what was forecast and scale that gap against how much that series normally misses by. That is what a surprise z-score does, and it answers a question the ladder cannot. Then check whether the front end of the curve repriced. If rate expectations did not move, the index move was positioning rather than repricing, and it usually does not hold.

Where the terminal fits

Helious is not a charting or order flow tool and it will not replace your ladder. It sits next to the ladder. The calendar means 8:30 never arrives unannounced, and every release is scored against its own history the second it prints. A live squawk reads you the number instead of making you hunt for it, the curve reaction sits beside it, and a momentum score tells you whether the tape is confirming what you just heard. It is $39.99 a month with a free tier, and the methodology page shows the workings.

Where to go next

What the thin book does to price is in the DOM during a news release. The release itself sits on the payrolls page, and the explainer covers how it is put together. The economic calendar has the dates, and alerts mean the release never arrives while you are looking the other way.

For judging the surprise, the surprise z-score page explains the scaling and the release guide walks through reading a print. Two more posts worth your time are predicting stock moves from job reports and news on futures prop firms.