< Back to Troubleshooting

SUPPORT

Is my installation working?

Predict & Profit 3.0 is designed to place few trades and to stop when it cannot prove something. A quiet bot is usually a working bot. This page shows how to tell the difference using the commands and dashboard screens you already have.

The commands below only read your local installation. They do not start, stop or change anything, and they do not contact Kalshi. They assume the standard installation folder; use your own if you chose a different one.

1. Confirm which version you run

"$HOME/predict-profit/bin/predict-profit" version

The output shows the active version and a status of ACTIVE, CONFIGURATION_REQUIRED or RECOVERY_REQUIRED. The version should match the version in the file name of the ZIP you downloaded from Gumroad.

The current Gumroad download is predict-and-profit-3.0.1.zip. Its file size, SHA-256 checksum and the exact commands to check it are under Verify your download. Check the ZIP before you extract or install it, and keep the original file.

If the version has extra text added after the number, or does not match your download, the installed code is not the release you downloaded. Read Changed installations before you upgrade or run any recovery step.

In the dashboard, Health then Build and schema identity lists the source hashes of the code serving the dashboard and both bots. Not attested means the value could not be established. Include this panel when you contact support.

2. What Doctor checks, and what it does not

"$HOME/predict-profit/bin/predict-profit" doctor

Doctor checks that the installed release matches the record made when it was installed, that the private folders and credential files are in place, that the Python environments, configuration and PostgreSQL databases agree, that no setup or upgrade was left half finished, and that each component you started is running.

Doctor does not contact Kalshi; it reports provider connectivity as not checked. It does not check whether live trading is authorized; when an activation file exists it says it does not validate live startup authorization. It does not look at safety halts, orders or fills. It compares the installed files with the record made at installation, not with the file you downloaded.

A Doctor PASS means the installation is intact. It does not mean trading is working.

3. Live or simulated?

"$HOME/predict-profit/bin/predict-profit" status

Under intent, "weather": true or "econ": true means that bot was started with --live. false means dry mode: the bot evaluates markets and records what it would have done without sending an order. The two bots can run in different modes.

The dashboard top bar shows what the running bots report: LIVE ACCOUNT, PAPER ACCOUNT, MIXED EXECUTION MODES (one bot live, one dry) or ACCOUNT MODE UNKNOWN. The Health page repeats this as Account execution mode, with each bot's process attestation.

The LIVE WEATHER badge on the Reactor and Network pages appears in every mode, so it does not tell you whether orders are live. Use the account badge.

On the Trades page, a Status of dry_run is a simulated order. Nothing was sent to Kalshi.

4. Orders, fills and settled results

Weather. The Trades page lists every order attempt. Submitted is when the bot recorded the attempt. A status of accepted means Kalshi accepted the order, not that it filled. Fill detail shows filled and remaining contracts. Settled and Result appear once the market settles, with Recorded ledger P&L and Fees for that order.

Economics. The Economics page shows totals under Capital and results: Resting orders, Working contracts, Settled trades and Recorded ledger P&L. Economics orders are not listed one by one in the dashboard.

Dashboard ledger figures come from the bots' own records. Your order, fill and settlement history on Kalshi is the authoritative record for both bots.

  • Recorded ledger P&L: realized results of settled trades after recorded fees. Open positions are not included.
  • Weather realized P&L today: settlements since your local midnight. No settlements means nothing settled, not break-even.
  • Position value: the current market value of open positions as Kalshi reports it. It is not profit.
  • Kalshi account total: cash plus position value, the figure Kalshi.com labels Portfolio.

5. Why did a scan place nothing?

Open Decisions. Each row is one Weather market the bot evaluated, with the Reason it stopped. The Reactor page shows the same reasons under Rejected candidates. Economics shows its own Recent opportunity decisions and Decision funnel.

spread_too_widevolume_too_low

Not enough liquidity in that market.

edge_score_too_lowdollar_edge_too_smallscore_too_low

The market price was too close to the model to be worth the trade.

price_too_lowprice_too_high

The contract price was outside your configured range.

no_forecast_available

No forecast guidance for that station and day. Same-day Weather markets show this routinely.

already_open_or_heldcity_exposure_capseries_exposure_capcost_exceeds_remaining

A position, open order or limit is already using that room.

New rows with reasons like these, a Next live scan countdown and Live data showing Fresh mean the bot is filtering normally. Loosening thresholds to force trades is not a fix.

6. Signs that need attention

  • Next live scan shows OVERDUE.
  • Live data shows Stale or Unknown.
  • No new Decisions rows across several scan intervals.
  • A safety panel shows TRIPPED (new entries halted) or UNKNOWN (current safety cannot be verified). Both block new risk.
  • Health shows Account authorization with reasons listed, or Unresolved submissions above 0.
  • Weather status detail or Economics status detail shows a warning.
  • Doctor reports FAIL.

Known limitation, verified in versions 3.0.0 and 3.0.1. When the Economics bot runs live and cannot confirm the shared account state, it places no orders for that scan, but its safety panel can still read SAFE with the reason Within the daily loss threshold. SAFE means the latest loss-limit check passed; the panel itself notes that account authorization is a separate check. Read Account authorization and Unresolved submissions on the Health page as well, and compare with your Kalshi order history.

7. When to stop and contact support

The fixes on the troubleshooting page for interrupted setup and interrupted upgrades are safe to follow yourself.

Stop and email support before doing anything else when:

  • A safety panel stays TRIPPED or UNKNOWN after the next check.
  • Unresolved submissions is above 0.
  • The version does not match the release you downloaded.
  • A message tells you to run a recovery command.

Do not clear halts, delete journals or marker files, edit the databases, or restore a backup over newer data. Each of these can remove the evidence needed to keep your trading history intact.

8. Using an AI assistant safely

An AI assistant is useful for explaining an error message or a log line. Remove secrets first: API key IDs, private keys, passwords, connection strings and account identifiers.

Explaining is different from changing a trading installation. Do not let an assistant edit files inside the installed release, change the databases, clear a halt, or run recovery, migration or reset commands. Changed code can alter how orders and safety checks behave, and the upgrade command refuses to start from a version it does not recognize. If an explanation points to a code or data change, email support first.

What to send support

Email support@predictandprofit.io. A complete first email usually settles the question in one reply. Include:

  1. The output of these three commands:
    "$HOME/predict-profit/bin/predict-profit" version
"$HOME/predict-profit/bin/predict-profit" status
"$HOME/predict-profit/bin/predict-profit" doctor
  2. Your operating system, including whether it runs under WSL2, and your Python and PostgreSQL versions.
  3. Full-page screenshots of the dashboard Health page and of the page where you see the problem.
  4. The exact error text and roughly when it happened.
  5. Whether anything inside the installation has been changed, by you or by an AI assistant, with a short summary of what changed.

Keep secrets out. Never send .env or other credential files, anything from shared/config, private keys, database dumps, passwords or connection strings. Read logs and command output before you share them, and remove API key IDs and account identifiers.