# Vector Desk — quick start

Updated 21 September 2026. Open [Vector Desk](https://arwyn.party/vectordesk/). Public research needs no account or wallet.

## Your first five minutes

Vector Desk tests whether unusually busy periods of trading are followed by different price movements. It also tests whether fixed trading rules would make money after estimated costs. We have not established a dependable trading advantage.

You can help without being a trader. Nothing here requires buying crypto or connecting a wallet.

1. Open **How it works** for a short explanation and an example. A **candle** summarises prices for a period, such as one hour. A **vector candle** is one our rule flags as unusually active. A **paper trade** is practice with no real purchase or sale.
2. On **Overview**, check the report date and read the finding. “Too early” means the study cannot answer yet. Research data is collected automatically; you do not need to run tests or make practice trades to keep it going.
3. Use **Help improve Vector Desk** on Overview. Tell us what confused you, what failed, or what would make the site useful. Write a short message and press **Send feedback**. An email is optional if you want a reply. No account is needed.

For example: “I thought 92% meant I would win 92% of trades. What does that number mean?” This is useful feedback. The number describes how often a price was revisited in a sample; it does not predict your chance of making money.

Your message and any reply email are stored in Vector Desk's Cloudflare database for the project owner. They are not posted publicly or added to the trading study. A daily code derived from your network address limits repeated submissions; the raw address is not saved with feedback. Please leave out credentials and personal trading records. If sending fails, your text stays in the form. You can retry or email [Vectordesk@frogclub.io](mailto:Vectordesk@frogclub.io).

Your browser-local practice journal is **not** sent with feedback or pooled into the shared research. Simply visiting the site does not contribute trades to the experiment. You can contribute by reporting confusing wording, reproducible errors or a useful idea. Reference-method corrections are welcome: include the source and explain the difference, without personal account data. New research ideas are recorded for a future review; the current study's rules stay fixed.

## A simple example

If a candle opens at 100 and closes at 110, the halfway price is 105. That is its **body midpoint**. A later touch of 105 within the study's time limit counts as a midpoint revisit. It does not show that a trade would have been profitable: that depends on the entry, exit and costs.

**How it works → Plain-English glossary** explains the other terms. You can leave the exact strategy rules and optional practice tools until later.

## Reading the detailed evidence — optional

1. Start on **Overview**. Read the report timestamp, freshness notice and next collection target. A delayed or offline report remains a dated snapshot; returning to the page checks for updates.
2. Read the **Latest hourly setup snapshot** for UBTC, UETH or HYPE. **Waiting** means the complete pattern was absent on that closed candle. **Pattern forming** still needs confirmation. **Pattern confirmed** is a pattern check; current quotes, costs and position limits are checked separately.
3. Compare the **vector / comparison** figures. The comparison permits any volume and can include vector candles. **Settled** outcomes are separated from **Provisional** sample-end exits. Displayed returns include those provisional marks and modeled costs. Arms may contain the same trades, so their counts are not independent evidence.
4. Read **Vectors alongside ordinary candles**. This is the saved July–September historical development sample, for green candles revisited downward within 24 hours. Compare all three markets and both outcome rows. Results are mixed; overlapping observations do not establish a dependable recovery advantage or after-cost returns. It is separate from the current forward studies.
5. Explore **Historical vector-candle ranges**. Amounts estimate the whole original candle's traded value, not value in the remaining fragment, outstanding orders or market-maker inventory. Source-candle time is separate from map observation time. Ranges shrink as price revisits them while retaining their original value for ranking. Overview shows the three largest on each side; **Market desk** provides the fuller map and its controls.
6. Expand **Are vectors recovered in both directions?** for green-down and red-up comparisons. Each completed observation needs 24 hours of follow-up. Young events remain pending. Recovery frequency is not a trading return; the simulated trading strategies remain spot longs.

UBTC and UETH are venue representations of BTC and ETH, quoted against USDC on Hyperliquid. These are that venue's candles, not a combined view of every exchange.

## Practice and keep a record

Open a market from Overview or **Market desk**. Choose a market, trading pace and setup, then enter **Planned entry**, **Stop price** and **Target price** in the position planner. Review the fee, slippage and risk assumptions. For a long plan, stop must be below entry and target above it. **Record paper plan** becomes available only when the displayed checks permit it.

Switching market, pace or history clears the previous plan's prices. **Refresh market** reloads the same context and retains a staged plan. Check the selected market before recording.

In **Paper journal**, manually enter an exit when you want to record a hypothetical outcome. This journal is saved only in the current browser and website origin. It does not update automatically, sync across devices or become a broker fill. Use **Export journal** before clearing browser data or moving devices. Keep the JSON file privately. Use **Import journal** to select an Export journal backup (up to 2 MiB and 2,000 plans). Review the number of new plans and skipped duplicates, then confirm or cancel. Matching IDs keep the existing record, even if the backup has a newer exit. Invalid files leave the journal unchanged. Imported open plans count toward the normal journal limits; restoring a backup does not connect the cloud recorder.

Use **Export this report** on Overview to save the exact shared snapshot displayed. If downloading is blocked, the same JSON appears in a copyable text box. This public report export is separate from your personal journal.

## What runs automatically — 21 September 2026 addendum

| Feature | What happens | What you need to do |
| --- | --- | --- |
| Shared research | Targets one completed-candle collection every four hours, with hourly catch-up checks. Scheduling may be delayed. It runs with your browser closed. | Read timestamps and warnings; you do not need to press research buttons. |
| Public page | Checks the saved report every five minutes while visible, and on return or reconnection. | Distinguish a refreshed display from a newly collected report. |
| Market desk | Requests current public candles when opened without loaded data, when context changes, or when **Refresh market** is pressed. | Refresh when you need new market data; this is not a continuously streaming chart. |
| Owner's private recorder | Targets minute checks while an owner recording session is active, including with the browser closed. Saves observations and modeled outcomes privately. | The owner uses **Owner sign-in**, then **Paper journal** for controls and **Export full paper record**. Visitors do not have personal cloud recording. |
| Manual paper journal | Saves plans and user-entered exits locally. | Record outcomes yourself; export a backup or restore one with **Import journal**. |

The owner recorder is separate from the fixed research comparisons. It uses observed quotes and, for some exits, reconstructed minute candles; neither is a real broker fill. **Pause new entries** continues monitoring an existing paper position. Gaps and ambiguous outcomes remain visible rather than being invented as fills.

## Dates, older reports and next steps

The fixed trading window is **20 September 2026 00:00 UTC to 19 November 2026 00:00 UTC**, end exclusive. The descriptive two-way recovery study starts **21 September 2026 00:00 UTC**. Its follow-up rules still apply. Reaching the end date does not by itself establish profitability or complete missing evidence.

**Advanced → Fixed experiments** explains the comparisons and offers the [frozen protocol](protocol.json), [historical experiment report](fixed-experiments.md) and [historical results](fixed-experiments.json). Those historical files describe the data and operations at their recorded dates. In particular, the original protocol's statement that there is no background collection predates the automatic collection described above. The frozen rules and files are retained; this dated addendum describes current operations. Use Overview for the latest collected report.

**Strategy lab** runs optional tests on loaded history. **Advanced → Decision replay** explains historical decisions. Repeatedly inspecting or tuning against a holdout makes it development data; these tools do not create a new independent test by being run again.

Read **How it works** and the [method audit](method-audit.md) for definitions and known limitations. The exact intended reference indicator and full strategy equivalence remain unverified. Use **Help shape the beta** on Overview to copy a feedback template; nothing is sent automatically. No wallets are connected and no real orders are placed.

The [current scope and next decisions](scope.md) are visible from Overview. Participant Study B is deferred to a 19 November scope review; no address-classification study is running behind owner sign-in. The next priorities are unattended reliability and independent beta sessions. Existing notebook, recorder and replay features remain available while feature expansion is paused.