Using Correlations and Market News to Make Better Trading Decisions

Helious Desk · 8 August 2026 · 7 min read

Correlation is one of the few statistics a retail trader can actually use, and it is almost always put to the wrong job. It is bad at predicting the next move and very good at telling you how much risk you are really carrying.

1. What does a correlation actually tell you?

How closely two things have moved together over a chosen period, on a scale from plus one to minus one. Plus one is lockstep, minus one is exactly opposite, zero is no consistent relationship.

The number only looks backwards. It is an average over a window you picked, so a different window gives a different answer. And it says nothing about size, which means two markets can be highly correlated while one of them moves ten times as far.

It describes the past. Treating it as a forecast of the next hour is where most of the trouble starts.

2. What is the biggest mistake people make?

Treating it as a cause. Two markets moving together nearly always means both are responding to the same third thing, and in macro that third thing is usually interest rate expectations. Gold and the Nasdaq are not moving each other. They are both reacting to real yields.

Believe that A drives B and you will trade B off A, then get run over the day the real driver changes. Knowing the shared cause is what tells you when the relationship should hold and when it should not, and the correlation number on its own never will. Our post on how the dollar, yields and FX correlate maps out what those shared causes usually are.

3. How does correlation help me size positions?

This is the most useful thing correlation does and the one almost nobody bothers with. It shows you hidden concentration. The number of positions you hold is not the number of bets you have made.

Long Nasdaq, long gold, short the dollar, long bonds. Four tickets, four markets, and on paper it is a nicely spread book. In practice all four are the same wager that rate expectations fall. One hawkish surprise loses on every one of them at the same moment, and the account takes four times the hit the trader thought they had arranged.

The question before you add anything is whether it is a different bet or the same bet wearing different clothes. If your positions are correlated, your real risk is much bigger than your position count suggests.

4. Why do correlations break when I need them?

Because in a real crisis people sell what they can, not what they want to, and forced selling ignores fundamentals. Assets that normally have nothing to do with each other start moving together, and correlations across risk assets tend to head toward one at exactly the moment diversification was supposed to protect you.

None of that is the statistic failing. It is what stressed markets do, and it is predictable enough to plan around. Size on the assumption that correlations will rise when things go wrong, not on the comfortable numbers from a calm stretch. A volatility reading sitting in the calm part of its range is exactly when that warning is easiest to ignore.

5. Can I use a correlated market to confirm a move?

Yes, and it is one of the cleaner uses of the number. If equities jump but the market that normally moves with them does nothing, one of the two is wrong, and it is usually the one that moved alone. An index rallying while the front end of the yield curve sits perfectly still looks like positioning or thin liquidity rather than a real repricing, and those moves tend to fade.

Confirmation is a filter, not a signal. It does not put you into a trade. All it does is mark the moves that deserve less trust, which is a smaller job than picking entries and a more reliable one. The momentum score does much the same on the tape itself.

6. Does correlation tell me which market leads?

No, and that misunderstanding is expensive. Correlation is symmetric. It measures that two things moved together and says nothing about which moved first. If you want lead and lag, you have to go and measure lead and lag.

Apparent leadership is often an artefact anyway. It can come from the hours each market trades, or from one being more liquid and therefore quicker to reprice. Build a strategy on the belief that A predicts B, when all you measured is that they move together, and you have found one of the easier ways to lose money slowly enough that you do not notice.

7. What timeframe should I measure over?

The one that matches your holding period, plus at least one other for context. A 120 day correlation is close to irrelevant to a trade you intend to hold for twenty minutes, and an intraday correlation is noisy and unstable for a position you plan to hold for months. Mismatching the two is common, and it gets you a number that looks respectable and tells you nothing.

Run more than one window and compare them. If a relationship holds over ninety days but has broken down over thirty, something changed recently, and that gap usually tells you more than either number on its own.

8. What does it mean when a correlation changes?

It means the market has changed what it is worried about, and that is worth a lot more than the correlation itself. When stocks and bonds stop moving in their usual relationship, the market has typically switched from pricing growth to pricing inflation, or crossed from one regime into another. The correlation is the symptom of that switch.

The number tells you that something shifted and roughly when. It cannot tell you what. For that you need to know what happened around the date the relationship changed, which is a news problem, not a statistics problem.

9. How does market news fit with all this?

They answer different halves of the same question, and neither works well alone. Correlation looks backwards and tells you how markets have been relating to each other, and therefore where your real risk sits. News looks forward, at what is scheduled and what just happened.

Use correlation to understand the current structure of the market. The calendar tells you what is about to test that structure and the feed explains it when it changes. When a correlation breaks on the same day as a major central bank decision, you already know why instead of working it out afterwards. Whether the release was a genuine shock or just a loud headline is a question for the surprise z-score, not for the size of the move.

10. What does a simple routine look like?

Start with what your positions correlate to, so that you are counting real bets and not tickets. Then look at what is scheduled today, because a release is what most often tests a relationship, and an alert means it never arrives unannounced.

Size for stress, on the assumption that correlations rise when things go wrong. When something moves, look at the related market and see whether it agrees or whether this one went alone. And when a relationship changes, go and find out why instead of assuming the statistic broke. None of that takes more than a few minutes, and none of it is analysis.

Where the terminal fits

Helious carries a cross-asset correlation matrix in a single card, covering the US 10 year and 5 year, the S&P 500, the dollar index, gold, crude, high yield credit and the VIX, across 30, 60, 90 and 120 day windows. Four windows instead of one is what lets you catch a relationship that holds over ninety days and breaks over thirty. The card sits beside the calendar with alerts, the live curve, every release scored against its own history the second it prints and a real time news feed, so the structure and the reason are on the same screen. It is $39.99 a month with a free tier, and the methodology page shows the workings.

Where to go next

The relationships themselves are covered in dollar, yields, Treasuries and FX, and the Treasury curve for stocks and futures explains why equities sit downstream. VIX, VVIX and implied volatility is the one to read for stress. And what tests a relationship is nearly always a scheduled number, which is what the economic calendar and which indicators matter most are for.