Surprise factors

The 7-day outlook on our front page answers where is the price likely to go. This page answers a different question: what could make that answer irrelevant. Every number below was measured against our own price history before it was written down, and several of them contradict the received story.

Where we are today

Surprise riskElevated

Chance of a >10% move in 30 days22%

Base rate 8.3% — the share of months since 1960 with a move larger than 10%, using the lower figure that applies after a quiet month (8.3% vs 14.0%). Raised 2.60× because 2 factors are live: Vegetable-oil complex repricing, Strong El Niño yield shock. Over 7 days rather than 30 the equivalent rate is 1.3%. This assessment never adjusts the 7-day price forecast.

armedVegetable-oil complex repricing — Palm–soyoil spread at $475/t discount — beyond the 95th percentile of the last 34 years.
watchStrong El Niño yield shock — ONI at 1.4 — strong El Niño territory. Capped at watch: the yield effect lags 6–12 months and the 2023-24 event produced no first-year output hit.

Assessed 2026-08-14 against the 2026-08-13 close.

How often big moves actually happen

Before any factor is considered, this is simply how often palm oil moves a lot. These are the numbers the probability above is built from.

Window	Chance of a move larger than ±10%	Sample
7 trading days	1.3%	602 windows, Dec 2023–Aug 2026
1 month	14.0%	798 months, 1960–2026
1 month, after a quiet month	8.3%	266 quiet months
3 months	38.2%	796 overlapping windows

Why the alert is framed at 30 days and not 7.At a weekly horizon a >10% move is roughly a 1-in-75 event — it has happened eight times in our entire daily record, most recently the week ending 12 March 2026 (+11.12%). An alert claiming weekly shock risk would be crying wolf. At a monthly horizon the same event happens about one month in seven, which is often enough to be worth planning around.

A calm market halves the risk. It does not remove it.After a month in which price moved less than 2%, the chance of a >10% move in the following month falls from 14.0% to 8.3% — measured across 266 such months. That is precisely why this page exists during a quiet stretch.

The factors

Ordered by measured palm-specific magnitude — not by how dramatic each one sounds. The vegetable-oil complex sits above Indonesian export bans because that is what the data says.

Vegetable-oil complex repricing

Vegetable-oil complexeither direction

Palm is one leg of a substitutable complex. When soybean oil reprices — South American drought, a US crush shift, biofuel policy — palm follows, because buyers switch on relative price rather than on palm's own fundamentals.

What it has actually done: The strongest statistical link we can measure: correlation 0.642, beta 0.723, R² 0.412 across 413 monthly observations (1992–2026). Between Dec-2024 and Jun-2026 soybean oil rose 75.1% while palm fell 7.2% — a 47pp palm-specific divergence, sustained rather than spiked.
Typical time to peak: about 12 weeks
Does it unwind?: Usually not — complex repricing tends to persist rather than round-trip, which distinguishes it from policy shocks.
What we can detect: We track the palm–soyoil (BOPO) spread every run and flag it when it leaves its normal range.
What we cannot detect: Our soybean-oil figure is a MONTHLY AVERAGE from FRED, so it lags badly as a live signal. A fast move in soyoil reaches us weeks late.

Strong El Niño yield shock

Climateupward bias

Heat and drought stress palms, but the yield effect appears 6–12 months later, not immediately. That lag is what makes this a standing risk flag rather than a weekly signal.

What it has actually done: The 2015–16 event is the one clean case: Malaysian output fell 11.0% year-on-year, the worst in 26 years of USDA records by a 3.6pp margin, and price peaked 34.8% higher twelve months after the ONI peak.
Typical time to peak: about 1 year
Does it unwind?: Completely, and slowly. The 2015-16 spike peaked in Jan-2017 and was fully given back by Nov-2018.
What we can detect: We ingest the ONI index every run. Sustained readings at or above +1.0 raise this to watch.
What we cannot detect: The mechanism is verifiable in only 1 of 5 historical events — our production data starts in 2000 (USDA) and 2012 (MPOB). And 2023-24 is the cautionary case: a strong El Niño with NO first-year output hit at all (Malaysia +7.2%). ONI alone never arms this factor.

Indonesian export policy shock

Supply policyeither direction

Indonesia supplies roughly 59% of world palm oil, so its export levy, DMO and permit rules reprice the physical market directly. But the direction is genuinely ambiguous: an export tax can lift the world price by restricting supply, or depress it by trapping stock domestically.

What it has actually done: Smaller and less directional than the headlines suggest. Controlling for soybean oil, only ONE of six measured episodes carried a real palm-specific move (Oct–Dec 2024, +21.6pp). The April-2022 total export ban produced 9.68pp of palm-specific UNDER-performance in its announcement month. Across all six, mean excess is −2.70pp (t = −1.29).
Typical time to peak: about 11 weeks
Does it unwind?: Yes, and violently. The 2024 episode round-tripped in 8 months. The 2022 sequence ended in a 29.58% single-month fall — the third-largest in 798 months of data.
What we can detect: Headline scanning for levy, DMO, export-ban and reference-price terms across our Indonesian policy feeds, plus the published levy schedule.
What we cannot detect: We read headlines, not article bodies, and all our queries are in English — Indonesian policy is briefed to Indonesian press first. We also cannot prove lead time from our own records.

Biodiesel mandate step

Demand policyeither direction

Indonesia's B-series mandate diverts palm into fuel. A step from B40 to B50 absorbs several million tonnes a year against ~50Mt of world trade — but the surprise is never the announcement, it is whether the step is actually funded and enforced.

What it has actually done: Highly variable, and mostly nothing. Three of five measured mandate steps produced no move outside one standard deviation. The B40 run-up carried +45.4pp of palm-specific excess, but our data CANNOT separate that from a simultaneous 34.8% fall in Malaysian production.
Typical time to peak: about 26 weeks
Does it unwind?: The B40 run-up gave back 39.3pp within six months of taking effect.
What we can detect: Mandate effective dates are in our database, so this is a countdown rather than a surprise — we raise it to watch inside 60 days of a step.
What we cannot detect: Funding and enforcement are the actual variable and we cannot observe either. Our own mandate table still reads B40 six weeks after B50's effective date — the slip is visible in the data before it appears in commentary.

Energy shock and shipping chokepoints

Macro & logisticsupward bias

Crude reaches palm through biodiesel blending economics, and a chokepoint closure (Hormuz, Red Sea, Suez, Malacca) raises freight on every cargo. Both are real channels — they are just far weaker than they feel.

What it has actually done: Weak on its own: palm-on-crude R² is 0.030 over the full sample, and close to zero before 2010. In the 2026 Hormuz episode Brent roughly doubled while palm moved about a sixth as much.
Does it unwind?: Energy spikes typically unwind faster than they build.
What we can detect: Brent is ingested daily and a large 7-day move raises this to watch. Chokepoint disruption reaches us only through headlines.
What we cannot detect: Crude is an amplifier, never a standalone cause — it is deliberately incapable of arming this factor by itself. We hold no freight-rate data at all.

Ringgit / rupiah shock

Macro & logisticseither direction

A currency move changes the US-dollar price without anything happening in the physical market. This is the single largest source of palm 'moves' in USD terms that are not moves at all.

What it has actually done: Correlation of monthly USD/MYR change with USD palm returns is −0.314 across 319 months. Not a shock generator — a relabelling.
Does it unwind?: Follows the currency, not the commodity.
What we can detect: USD/MYR and USD/IDR are live-ingested every run.

Speculative positioning unwind

Macro & logisticseither direction

A crowded position is fuel, not direction. An extreme net long is vulnerable to liquidation; an extreme net short to a covering squeeze. It amplifies whatever else happens.

What it has actually done: Not measurable on our sample — the CFTC series we hold is too short to establish that extreme positioning preceded large palm moves. Listed because the mechanism is well established, with no number attached.
Does it unwind?: Typically fast, days rather than weeks.
What we can detect: CFTC managed-money net positioning, weekly and released with a three-day lag. It is a soybean-oil report, not palm.
What we cannot detect: Palm has no CFTC report. This is a complex-wide sentiment read.

Acute flooding in the palm belt

Climateupward biasmonitor only

Monsoon flooding in Sabah, Sarawak, Riau or Sumatra interrupts harvesting and mill logistics for one to three weeks.

What it has actually done: Deliberately unquantified. A one-to-three week disruption cannot clear the noise floor of our monthly series, whose 90th-percentile absolute move is 11.89%. Attaching a number here would be inventing one.
Does it unwind?: Typically complete once harvesting resumes.
What we can detect: Our 7-day rainfall outlook flags heavy-rain regions as a logistics note.
What we cannot detect: We hold no flood dataset — only a rainfall forecast. Severity, area flooded and mill closures are all invisible to us.

Conflict or export-tax escalation between producers

Macro & logisticsupward biasmonitor only

Armed conflict involving a producer or consumer nation, or a tit-for-tat export-tax escalation between Indonesia and Malaysia, would reprice supply security itself.

What it has actually done: The nearest measured analogue is the 2022 Black Sea disruption, which removed roughly half of world sunflower-oil exports. Even there the palm move was dominated by the complex rather than by the conflict.
Does it unwind?: Unknowable.
What we can detect: Headlines only, through our geopolitical news queries.
What we cannot detect: Genuinely unforecastable. Listed so that its absence from the probability is a stated choice rather than an oversight.

What the history actually shows

Four episodes worth understanding, because in three of them the obvious explanation is wrong. Throughout, “palm-specific” means the move left over after subtracting what soybean oil did in the same month — without that control, almost every palm shock looks larger than it was.

April 2022 — the export ban that didn’t do what everyone remembers

Indonesia, which supplies roughly 59% of world palm oil, banned exports entirely for 25 days. On a monthly-average basis the world price ended the announcement month 5.29% lower, while soybean oil rose 4.39% — 9.68 percentage points of palm-specific under-performance during the largest supply interruption in the industry’s history.

The collapse came afterwards and was enormous: July 2022 fell 29.58%, the third-largest single month in 798 months of data.

Measurement limit: our daily palm series starts in December 2023, so the intra-month path in 2022 is unmeasurable. Reports of a record intra-month futures print exist; we do not assert them because they are not in our data.

January–March 2022 — a +39.9% move that was mostly not about palm

Indonesia’s domestic market obligation is usually credited with this run. Measured against soybean oil month by month, the palm-specific component is +2.17 percentage points of the +39.9% headline. Roughly 95% of the move was the vegetable-oil complex repricing — and Russia invaded Ukraine in the middle of the window, removing about half of world sunflower-oil exports.

Anyone quoting “+32% from the Indonesian DMO” overstates the policy by roughly six times.

October–December 2024 — the one policy episode that survives the control

Indonesia moved its export levy to a 7.5% ad-valorem regime. Palm carried +21.6 percentage points of palm-specific gain while soybean oil went sideways. It then round-tripped completely: the December peak gave way to a 23.7% fall by May 2025, finishing below where it started.

Three Indonesian actions landed in that window — the levy, the B40 mandate confirmation and a genuine output shortfall. Monthly data cannot separate them, so this is 21.6 points attributable to a cluster, not to the levy alone.

2015–16 El Niño — the one clean mechanism case

Malaysian output fell 11.0% year-on-year, the worst in 26 years of USDA records by a 3.6-point margin, and price peaked 34.8% higher twelve months after the ONI peak. Then a full round trip: the January 2017 high was entirely given back by November 2018.

The mechanism is verifiable in only one of five strong El Niño events, because our production data starts in 2000. And 2023–24 is the counter-example: a strong El Niño with no first-year output hit at all. This is why the ONI index alone never raises this factor above “watch”.

What this page will not tell you

Which way. The probability is of a move either way. Across six measured Indonesian policy episodes the mean palm-specific move was negative (−2.70 points), so a model that turned bullish on a levy headline would have been wrong four times out of six.
That it will happen this week. These are 1–3 month events. Nothing on this page adjusts the 7-day forecast, by design.
How much warning we had. We do not yet store headline history, so we cannot prove from our own records that any past episode was telegraphed in advance.
Anything about floods, freight rates or enforcement. We hold no flood dataset, no freight data, and no way to observe whether a biodiesel mandate is actually funded. Those factors are listed as monitor-only rather than quietly omitted.
An accuracy score — yet. Each assessment is stored and will be scored once its 30-day window closes. No accuracy claim will be published until at least twelve independent windows have closed, the same rule we apply to the 30-day range.

For how the main forecast is scored, see our scorecard, which publishes its skill figure including when that figure is negative.