1,500 Trump Posts Analyzed: How Presidential Rhetoric Moves Oil Prices

The Dataset

Since early 2026, we have monitored every post published on Truth Social by President Trump. Our system captures each post within 3-5 seconds of publication, runs it through an LLM scoring pipeline, and records the WTI crude oil price at the moment of capture. Follow-up price checks happen at 5m, 15m, 1h, and 24h to measure actual market impact.

Of the 1,500+ posts analyzed, approximately 85% score zero — meaning the model determines they have no meaningful oil market relevance. These include posts about domestic politics, personal commentary, endorsements, and general rhetoric that does not touch energy, trade, or geopolitical themes tied to oil supply.

The remaining 15% receive a non-zero score on our -5 to +5 scale and become tradeable signals. These are the posts this analysis focuses on.

Which Themes Move Oil

Not all oil-relevant posts are equal. We categorize each scored post by theme, and the data shows clear hierarchies in market impact:

Iran-related posts produce the most consistent and largest moves. When the President posts about military strikes, new sanctions on Iranian crude exports, or direct threats toward Iranian infrastructure, WTI reacts within minutes. The average absolute move on high-scoring Iran posts exceeds 1.5% in the first 15 minutes.

Hormuz-related posts are the most volatile. The April 2026 Strait of Hormuz blockade sequence produced the largest single-post moves in our dataset, with individual posts triggering 3-5% swings as markets priced in supply disruption risk for 20% of global oil transit.

De-escalation posts — statements suggesting diplomatic openings, ceasefire discussions, or the lifting of sanctions — produce bearish moves. These are less frequent but equally tradeable. A single post suggesting ships were resuming transit through Hormuz dropped WTI by over 8% as the risk premium unwound.

Signal Accuracy by Time Window

We measure accuracy as directional correctness: did the price move in the direction our model predicted? A +3 BULLISH signal is correct if WTI is higher at the measurement point than at signal time.

The 15-minute window shows the highest accuracy at 72%. This makes intuitive sense: presidential posts create an immediate information shock that the market prices in quickly. By 15 minutes, the directional move is typically established but has not yet been diluted by other market forces.

At 1 hour, accuracy drops to 64% as other news, institutional flows, and algorithmic rebalancing introduce noise. At 24 hours, accuracy is 55% — still above random, but the signal is significantly weaker as macro factors dominate.

The key insight for traders: these signals are most useful as short-duration trades. Enter on signal, target 15-60 minutes, and exit. Holding overnight on a presidential tweet signal alone is a weaker proposition.

Score Magnitude Matters

Not all signals are created equal. Our scoring scale runs from -5 to +5, and the data shows a clear relationship between score magnitude and both move size and accuracy:

Signals scored ±5 — the highest conviction level — hit at 82% accuracy with average moves exceeding 2%. These are rare (roughly 3% of all scored posts) but highly actionable. Our auto-trader is configured to only execute on signals with an absolute score of 3 or higher, filtering out the lower-conviction noise.

The Speed Advantage

Time-to-signal is critical. Our data shows that the majority of the price move from a high-scoring post occurs in the first 5-10 minutes. By the time the post circulates through news desks, trading chat rooms, and financial media, the initial reaction is already priced in.

Our pipeline captures posts within 3-5 seconds, scores them in under 2 seconds, and pushes to connected traders via SSE in under 1 second. Total latency from post to signal: under 8 seconds. This gives programmatic traders a structural edge over manual monitoring.

For traders on Hyperliquid, this matters even more. The xyz:CL perpetual trades 24/7, so when a post drops outside CME hours — weekends, holidays, the daily 4-5pm CT maintenance break — you can still trade the reaction immediately while CME traders wait for the market to reopen.

What Does Not Move Oil

Equally important is understanding what the model correctly filters out. The 85% of posts that score zero include:

• Domestic policy posts (immigration, education, healthcare) with no energy or trade angle
• Personal attacks on political opponents without policy implications
• Retweets of supportive media coverage
• Posts about past military achievements without new threats or escalation
• General economic commentary without specific tariff or trade action

A post saying “the economy is doing great” does not move oil. A post saying “we are imposing 25% tariffs on all Canadian energy imports effective Monday” does. The model distinguishes between these with high reliability because it evaluates specific policy language, not general sentiment.

The Hormuz Case Study

The April 2026 Strait of Hormuz crisis produced the most concentrated cluster of high-scoring signals in our dataset. Over a 48-hour period, multiple posts scored ±4 and ±5 as the situation escalated and then partially de-escalated.

WTI moved from approximately $95 to above $103 during the escalation phase, then dropped back below $93 on de-escalation signals. Traders connected to our stream captured both legs of this move. The system correctly scored the initial blockade announcement as +5 BULLISH and the subsequent transit resumption signals as -5 BEARISH.

This sequence demonstrated two things: first, the model handles rapidly evolving situations where the same theme flips direction within hours. Second, the speed of signal delivery matters most during these high-volatility events, when price moves are fastest and largest.

Implications for Traders

The data points to a clear trading framework:

• Filter aggressively. Only trade signals with absolute scores of 3+. The lower scores have marginal accuracy and small moves that do not justify the spread and fees.
• Trade the 15-minute window. This is where accuracy peaks. Enter on signal, set a 15-60 minute exit target.
• Size by score.A ±5 signal has 82% accuracy and 2%+ average moves. A ±2 signal is barely above random at 60%. Allocate accordingly.
• Use 24/7 venues. Posts do not follow market hours. If you can only trade during CME sessions, you miss the weekend and off-hours signals that often produce the largest moves.
• Watch for theme clusters. When multiple posts on the same theme fire within hours, the move compounds. The Hormuz sequence moved WTI 8%+ across a series of related posts.

Accessing the Signals

All signals are available in real time through three channels:

• SSE Stream— lowest latency, connect programmatically with a single line of code. Free during beta.
• Telegram Bot — @oilapibot. Use /signals for recent alerts, /stats for live accuracy, /price for current WTI from Hyperliquid.
• Dashboard — usoil.ai. Live signal feed, price charts, forecast models, and accuracy tracking.