Build Your Own
Write a formula from any variable and function in the Formula Reference, then save it. Saved signals can be edited or cloned later, so you can fork a built-in as a starting point and tune it to your universe.
- Example: takerBuyRatio >= 0.62 && imb >= 25 (aggressive buying into bid-side pressure).
- Example: abs(fundingBps) >= 3 (stretched funding in either direction).
- Example: ofi >= highest(ofi, 30) && imb >= 15 (order-flow pushing to a 30-candle extreme).
Describe It In Plain English
When AI assist is enabled, describe the behavior you want — "aggressive selling into a heavy bid book" — and VYX drafts a formula you can review and refine before saving.
The AI Keeps Working If You Leave
Building with the AI can take a minute — it reads your live board, checks a candidate against real data, and validates what it builds over history. That work runs on VYX, not in your browser, so you are free to leave. Refresh the page, close the tab, or pick the conversation back up on your phone: the build carries on and you rejoin it mid-stream, with whatever it has written so far already there.
- Every turn is saved as it happens, so a reload can never lose a reply or something the AI already built.
- Anything waiting on your Accept stays waiting — come back tomorrow, or on another device, and the decision is still there to make.
- One conversation runs at a time. If a build is still going when you open a different chat, VYX tells you and links you back to it.
Stop A Turn
While the AI is working, the send button becomes a stop button. Press it and the turn ends there — VYX cancels the request it has open, so the work genuinely stops rather than carrying on out of sight. The reply keeps whatever it had already written, marked as stopped, and anything it had already built for you stays on your board. Because stopping is recorded against the conversation and not your tab, you can stop a build from any device, including one that did not start it.
Rebuild A Chart Layer As A Signal
The chart’s built-in layers are not a closed set you can only switch on and off — the rules behind them are written in the same formula language you have. The Sweeps layer is the clearest example: it marks a candle when that bar’s dominant taker event lands in the top 6% of the market’s own recent history and is at least 2.5× its median. Written as a signal that is one line, and it produces the same marks on the same bars — now scoring your board, filtering your table, and available to your AI trader, not just drawn on the chart.
- The formula: pctRank(sweep, 120) >= 0.94 && sweep >= 2.5 * median(sweep, 120) — or start from the built-in "Sweep (chart parity)" signal and tune it.
- Loosen it by dropping the rank (0.90 marks the top 10%); tighten it by raising the median multiple.
- sweepSide tells you which way the aggression went, so you can split it into separate buy and sell reads.
Choose How It Paints
Pick a visualization mode independent of the match threshold: diverging for two-sided behavior (bid versus ask), or a high-value / low-value gradient for one-directional metrics. Trigger-style formulas can still render as a continuous metric so rare predicates show useful structure across history.
Score The Board, Or Read It On The Chart
Every signal answers two separate questions: does it rank the heatmap, and does it draw on the chart? The Score toggle controls the first — on means the signal feeds the board’s color and ranking. The chart toggles (Zone, Marker, Line, and Strip) control the second. The Line surface plots the signal’s score, and you choose where it draws under Draw in its settings: “Panel below” gives it its own indicator panel with a value axis, while “On chart” lays it over the candles on its own hidden scale, so you read its inflections directly against price. Turn Score off and switch on Strip, and the signal rides beneath the candlestick chart as its own colored row on the same timeline as price, without touching what ranks the board. That is how you keep one signal ranking the heatmap — say, a consolidation read to find where to look — while others ride the chart for direction, like imbalance or CVD. No need to blend everything into one score and lose the read you built.
Built-In Signals
Some signals ship with VYX and just work: Sessions shades the market sessions on the chart’s time axis, Macro events marks scheduled releases like CPI and FOMC by impact, and News marks the focused market’s catalysts. They live in your desk’s signal rail like any signal — with a small built-in tag — but carry settings instead of a formula: there is nothing to compose, you add one to a desk or you don’t. Tick a built-in on or off to show or hide its layer, open its gear to tune it (which sessions, which impact tiers), or remove it from the desk entirely and re-add it later from Add a signal → Built-in.
Marker: What A Handle Counts
A signal’s Marker surface drops handles on the chart, and you choose what one handle means. “Level starts” — the default — puts one handle where each level begins, which keeps the chart sparse and pairs with the Zone bands. “Every fire” puts a handle on every bar the signal fired, using the exact test the heatmap uses to light a cell — switch to it when you want the cells you scan and the handles you read to be the same set.
- Every fire: a lit cell on the map and a handle on the chart are one-to-one. Raise the board’s minimum score and both thin out together. This is the mode to pick when markers and the heatmap disagreeing has been bothering you.
- Level starts (default): pairs exactly with the Zone bands, because they are the same zones — but it merges runs of fires into one handle and caps how many draw, so it will not match the heatmap. That is by design, not a setting to fix.
- Switch between them per signal in the Plot control, under Mark.
Prove It
Every signal has a Proof panel that replays it over recent history — using the exact engine the heatmap runs — and reports, honestly, whether it earns its place: when it fires, does the next move run bigger than the market (attention), and is its side right (direction)? It is allowed to say "no edge." There is no look-ahead, and it stays cautious when the data is thin. The verdict is deliberately hard to fool:
- Net of trading costs: a real but tiny edge that fees and spread would eat is told plainly ("+6bps gross, −3bps after costs") — not scored green.
- Honest sample size: markets that move together are counted as the single move they are, so a two-minute market-wide swing can’t masquerade as thousands of independent wins. The verdict says "≈41 independent bars," not "5,142."
- Survives a re-test: an edge that holds on the first half of the window but flips on the second is demoted, and a marginal pass among many tests is discounted for the fishing.
- Measured per event: it reads the move after a fire STARTS, so a signal that stays lit for an hour counts as one event, not sixty.
- Regime-aware: it can surface an edge that only lives in a regime — "works only when the tape is already fast" — as its own, still-tradeable object.
Live Edge Health
Proof grades a signal against the past; the edge-health dot grades it against the present. Once a signal has a validated verdict, its Match Log carries a small live dot: green while recent real fires keep paying inside the validated band, amber when the live edge decays or flips sign, grey until there is enough live evidence to judge. Hover it for the numbers — "live 12 fires: +3.1bps vs validated +5.2±2.1bps." It is the fastest way to catch an edge going stale before it costs you.
Robustness — Plateau Or Lucky Peak
The Robustness tab sweeps a signal across a grid of its own settings — how strong a reading counts as a fire, and how long a window it compares against — and paints the result as a heatmap. A broad bright region means the edge holds across nearby settings (robust); a single bright square alone in a dark field means you have tuned to the past (overfit). Click any square to see its full read. It is the fastest way to tell a real signal from a lucky one before you trade it.
Notifications Come From What You Already Built
There is no separate alert-formula builder on VYX. Your signals score the board, your patterns watch it, and price alerts guard your levels — notifications attach to those, so there is never a second condition to author or keep in sync. Price alerts (below) chime, email, and push; and each Trade Desk has a Matches toggle in the chart’s Alerts panel — turn it on and VYX tells you when that desk produces a new match, including its armed patterns, tab closed or not.
- One toggle per desk: what counts as a match is whatever that desk’s lit signals and patterns already say — nothing extra to configure.
- Your AI trader can reach you too: turn on Notify in its alerts control and its authored plans, closed trades (honestly labeled sim or live), and daily review notes arrive by email and push — not just in the decision log.
- Turn on Push in the chart’s Alerts panel for a lock-screen notification — a fast nudge on top of the always-on email. On iPhone, add VYX to your Home Screen first; enable it on each device you want notified. If the toggle will not turn on, the push notifications guide has the fix for each browser.
- A market that keeps re-firing won’t spam you — repeat fires on the same market are muted for a cooldown window, and email volume is capped per day.
Price Alerts Fire In Real Time
Price alerts — the levels you set with one gesture on the chart — are watched server-side in near-real-time, not at candle close. Cross the level and VYX fires within seconds: a chime and banner if a tab is open (a tab resting hidden in the background conserves your machine and may chime up to half a minute later), an email (and a push notification, if you turned it on) if not. They sync with your account, so a level set on your desktop is armed on your phone too. One-shot by design: a fired alert stays on the chart as a marker until you clear it.
More docs
Try it on the live map
Open the desk and put this to work across 300+ Hyperliquid markets — no install, no account.
Open VYX