Research Desk
Fundamentals, explained rather than predicted
Four analysts publish one brief each market morning. The desk tells you what it has measured — what central banks actually said, what the crowd already holds, how far past releases actually moved the market — and defines every term as it goes. It does not tell you what to trade, and it never will.
We tried to build the version that predicts. We ran the studies, and they failed: following currency strength ties a coin flip, aggregated economic surprise loses to its own placebo over 8.5 years, and speculative positioning ranks next-month returns at an information coefficient of 0.03. So the desk does the thing the evidence supports — it explains. The research is published in full, including the parts that did not work.
The morning brief, in plain English
Every technical term is defined the first time it appears. "Statement tone" arrives with the sentence that explains it is a score from −100 to +100 counting hawkish and dovish phrases in a central bank's own words — not a forecast of what the bank will do next. Underneath sits Your book: the currencies you actually hold, matched against the desk's stances, the last measured tone and any high-impact event in the next three days. The analysts never see your account; the platform does that join itself.
- The short version is one click away. Experienced readers keep the terse brief.
- Disagreements are computed, not narrated. When two analysts hold opposite stances on a currency, the platform detects it and the brief must present it as unresolved.
- Silence is a real answer. An analyst with no data says so in one clause rather than filling the gap.
What changed since yesterday
A daily publication that is rewritten from scratch every morning gives you no way to tell what is new. This panel is computed by the platform, not written by an analyst: stance changes, newly scored central-bank statements, expectations opened or settled, and anything high-impact due within 24 hours. If you are new to this, it is the most useful thing on the page — knowing what is different does not require already knowing what is normal.
Biggest moves today — stated as co-occurrence, not cause
The most-asked question in retail trading is "why is EURUSD down?". The desk answers it the only honest way: here is what moved, and here is what else happened in the same session. It will tell you a release and a move happened together. It is explicitly forbidden from telling you one caused the other, and it says plainly when nothing coincided at all.
Sudden moves, explained as they happen
The same discipline now runs live. When any pair moves several times its normal hourly range inside ten minutes, the platform checks the calendar and the news wire for anything in the same window — within about a minute of the move completing, before you have had time to ask. What it finds lands on the wire for everyone, and if you are holding the pair, it reaches you directly as a notification. Two rules keep it honest: the note is assembled by code from stored headlines, so nothing can be invented — and when nothing coincides, it says so, because fast moves often print before any headline exists and "no visible catalyst yet" is itself worth knowing. One note per pair per few hours, so a wild session never becomes noise. It runs on the platform's own price feed, so it works even when your engine is off.
What past releases did
Before a big release, the desk reports how far the market actually travelled in the hour after the same release previously — typical, largest and smallest. This is measured history and says nothing about direction. It answers "is today likely to be violent?", which is a risk question, not a forecast. Where a release has too little history to characterise, the desk stays quiet.
What the crowd already holds
Each week the desk reads the CFTC's Commitments of Traders report and reports what large speculators hold in each currency. The number that matters is not the position — it is the percentile against that currency's own history, going back to 2018. "Net short" is a direction. "The 2nd percentile of 400 weekly readings" tells you almost no week in eight years has looked like this, which is the part you can actually use.
Why we can show this safely. We measured whether positioning predicts currency direction across 2,468 weekly reports and both CFTC report families. It does not — the information coefficient is at or below 0.03 at every horizon we tested, and the one formulation that technically survived did so with its sign inverted against its own hypothesis. Because it has no predictive power, describing it cannot smuggle a forecast in through the back door.
Central banks, as a trajectory
A single tone score is close to unreadable — 100 means nothing unless you know the last one was also 100. The desk shows the sequence, so 88 → 100 → 100 reads as a bank that firmed and held, and 40 → 22 reads as one that softened. Next scheduled decisions sit beside it, because between meetings this is a sparse signal and the calendar is the honest way to say so.
What SentryQ users hold
Counts of accounts, never sizes, never balances, never a single person's book. A pair appears only once enough accounts hold it that no individual can be identified from the number — and below that floor it is not shown at all, not even as a small number. The rule lives in the database query, not in a setting someone can turn off.
A desk that keeps its own score
When an analyst states an expectation, the platform records it and later marks it right or wrong by reading its own price history. The analysts get no say in whether they were right, and the hit rate is published next to the claims. A desk that cannot be wrong in public is not worth reading.
- Stances describe, they never instruct. Hovering any stance shows the plain definition.
- Conviction is confidence, not probability — and it is labelled as such.
- Outcomes are resolved from bar history, and a thesis with missing data is voided rather than quietly counted as a win.
Ask the desk — and be told when it will not answer
A question box sits at the bottom of every tab, so asking about what you are reading never costs you your place. It answers only from what the desk has measured. Ask it what a basis point is and it tells you. Ask it what to buy, where to put a stop, or what a market will do next, and it declines in one sentence — then gives you the closest descriptive fact it holds, so you get an answer rather than a wall.
That behaviour is not a hope, it is a gate. Before release we run the questions a real user actually types, including direct requests for trade advice and attempts to talk the desk out of its own rules. The current build passes 12 of 12 with zero violations, and every refusal still returned a real measured number.
The glossary is deliberately not generated. These definitions do not change from day to day, so asking a model to rewrite them each morning would be spend on a fixed string and a fresh chance to get it subtly wrong. They are written once, and the same wording is used wherever a term is explained.
Read the desk on your own account
The Research Desk ships with the SentryQ desktop app and reads your own positions from your own broker account. Nothing leaves your machine except the snapshot the platform needs to show you your book.