Risk-on and risk-off: how to read the regime

What risk-on and risk-off are

This lesson explains how to establish in a repeatable way whether the market is in a risk-on regime, a risk-off regime or a mixed one, and how to translate that reading into position size. It is written for a trader who already has an entry plan with a stop, but no way of connecting it to the cross-market context.

Risk-on is a regime in which capital seeks risk. Equities, growth companies and high-beta assets are bought more readily. Risk-off works the other way round: capital seeks safety, and pressure appears on the same basket of assets, only in the opposite direction. This is a simplification. It works as a context filter, but not as a law of physics or a forecast. The basket of high-beta assets also includes the largest cryptocurrencies, although their correlation with the regime is not constant — the division of roles between Bitcoin and Ethereum is described in the lesson Bitcoin and Ethereum.

Two caveats matter here. First, gold sometimes acts as an inflation hedge and sometimes as a risk-off asset — what decides its behaviour in a given episode is the context, not the label. Second, a global risk-off reaches the Warsaw exchange with a delay or through banks and exporting companies. A portfolio valued in PLN (Polish złoty) feels the USD/PLN rate on top of that, and UCITS ETFs on global indices breathe the same air as the underlying market.

Four cross-asset confirmations in a fixed order

A regime is not announced outright — you have to read it from the behaviour of several markets at once. To make that reading repeatable, check four confirmations always in the same order. The order is not part of the analysis, it is part of the measurement: only then can entries from different weeks be compared.

One down day does not change the regime. A regime changes when confirmations appear in several independent markets at the same time and last longer than a single session. That is why you do not judge the regime “by eye” — write down four zeros and ones, and calculate the rest from them. Reading a regime on an index only makes sense when you know what that basket contains — the construction of indices and of commodity futures is described in the lesson Indices and commodities.

Protocol: regime agreement in points (0–4)

All thresholds in this protocol are illustrative figures. Match them to your instrument and confirm them in your broker’s materials before you treat them as your own.

p_i                  ∈ {0, 1}   # confirmation i: 1 = in line with the hypothesis, 0 = not in line
Agreement            = p1 + p2 + p3 + p4          # score from 0 to 4
correlation_threshold = 0.7                       # from this level positions count into one basket
Exposure             = Σ |notional_risk_j|        # sum over positions with correlation ≥ correlation_threshold

The steps, in this order:

1. Write down your hypothesis before you look at charts. Risk-on, risk-off or mixed. Without a hypothesis there is nothing to confirm, and an assessment made after the fact always looks wise.
2. Check the four confirmations from the table above, in order from 1 to 4, and record each as 0 or 1. Do not change the order while you are reading.
3. Calculate Agreement and record it together with the date and the reference instruments you based it on.
4. Assign the score to a band and only then choose size. Agreement 4 → full size within the plan; Agreement 3 → half size; Agreement 2 or less → no new positions. The band thresholds are illustrative figures, but they have to be fixed, not adjusted to your mood.
5. Calculate Exposure across positions with a correlation equal to or above the threshold. If the sum exceeds your daily R limit, reduce size or close part of the position. The limit itself is described in the lesson on daily limits.
6. After the session, add the result in R and check consistency. Compare what the Agreement value predicted with how the market actually behaved. After a dozen sessions you will see whether your confirmations have measurement value or merely sound good.

A worked example — illustrative numbers. Capital of $50,000, risk of 1 per cent per position, that is $500 per trade. Three longs: a WIG20 company, a UCITS ETF on a broad index and a CFD on EUR/USD. If the correlation of each pair is 0.7, then Exposure is $1,500. In a risk-off episode those three positions usually lose at the same time, so the basket’s result is often three times worse than the number of positions would suggest. The 1–2 per cent per trade rule protects you from a single mistake, but not from a basket like that.

How to calculate total exposure to a correlated basket

Three longs in different markets are not three independent opportunities if they react to the same driver. Before you add a fourth position, answer three questions:

1. Do these positions share a common source? A broad market index, the dollar and general sentiment are often the same driver under different names.
2. What is the correlation of the pairs you hold? Check it on data from recent weeks, not from memory.
3. What is the sum of notional risk in the basket? Compare it with your daily R limit, not with the number of positions.

You write the answer into the worksheet next to the Agreement value. Correlation itself is not constant: it rises in risk-off episodes and falls in calm weeks. You will find more on how that translates into a whole portfolio in the lesson on portfolio risk.

When the regime says “do not trade”

Four conditions. Meeting any one of them means no new position or a reduction of one already open — not “a smaller risk to give it a try”.

1. Agreement is 2 or less while your plan assumes full size. If fewer than three of the four markets confirm the hypothesis, you do not have a regime, only noise. In that case, half size or flat.
2. Three or more long positions have a correlation equal to or above the threshold, and the combined result exceeds your daily R limit. That is one position on the basket. Adding to it increases risk rather than improving your price.
3. The regime shifts during the session and you want to add exposure. A change confirmed by at least three of the four points within one session is a signal to reduce. Averaging down a loss at that moment is a bet against your own measurement.
4. You cannot point to a reference instrument for your hypothesis. If you do not know which chart you will check the regime on, you are not measuring, you are guessing. One specific instrument per confirmation is a necessary condition.

Worksheet: a table to fill in

Copy the table into a spreadsheet and complete it after every session. The columns are deliberately empty: value comes from what you measure on your own data, not from what you read here.

You fill in the p1–p4 and Agreement columns before the session and the Result column after it ends. After a dozen rows you will compare days on which Agreement was 4 with days on which it was 0–2. That is the only way to check whether your confirmations measure anything.

Common mistakes in reading the regime

• Treating the frame as an entry signal instead of a context filter. Risk-off does not tell you when to enter, it only tells you how carefully to manage exposure.
• Assigning an asset permanently to one category. Gold is sometimes an inflation hedge and sometimes a risk-off asset; the context of the episode decides.
• Ignoring correlation when setting size. Several long positions in different markets during a risk-off episode are often one large risk, not several independent ones.
• Confusing a one-day pullback with a regime change. React to a confirmed pattern across several markets, not to a single red candle.
• Changing the hypothesis during the session. A new reading made after the fact is not a measurement, it is fitting the narrative to the result.

Measurable exercise

For five consecutive sessions, record in the evening the four confirmations, the Agreement value and the result in R, using the table from this lesson. Do not change the definitions of the confirmations along the way.

After five sessions, compare two sets: days on which Agreement was 3–4, and days on which it was 0–2. Check whether the result in R differs between them, and whether the number of positions opened on low-agreement days falls. This is not a test of how well the strategy works, it is a test of measurement consistency — if after a week you calculate Agreement differently than you did on day one, the problem is in the definition, not in the market. Keep the metrics from your journal according to the rules in the lesson on the journal and review.

Check yourself

Common questions

Does risk-on/risk-off replace technical analysis?

No. It is a context filter. A setup without context is often a trap, and market mood alone, without an entry plan, is not enough for a trade.

Why do the zloty and the Warsaw exchange react to Fed decisions?

Global liquidity and the dollar rate affect emerging markets and local risk assets. Domestic data and NBP decisions still matter, but they do not operate in isolation. The context of central bank decisions is covered in a separate lesson in the fundamental track.

What about crypto correlation?

Crypto often behaves like a high-beta asset, but individual episodes differ from one another. Treat it as a hypothesis to test in your journal, not as a fixed rule.

Why calculate the Agreement when the regime is visible on the chart?

Because an assessment "by eye" changes with the mood, while a number does not. The Agreement breaks a general impression into four checkable conditions, so you can compare sessions and instruments with each other. Only from recorded numbers can you work out whether your confirmations were accurate or merely cautious — otherwise you have nothing to compare. It is a decision filter, not a forecast: match the thresholds to your instrument and confirm them on your own data.