ROM PolyBot 2.36.6

ROM PolyBot 2.36.6 fixes candidate starvation in Large Trade and Momentum selection. Previously, the bot fetched the newest 100 rows and then removed already-used signals. Those rows could hide fresh, unused candidates, causing a cycle to find nothing to evaluate. Used IDs are now excluded before the query's 50-candidate limit, with stable newest-first ordering.

No entry thresholds were lowered. Freshness, resolution, allowed signal types, qualified live edge, fees, quote quality, available funds, sizing and risk limits still apply. Tests reproduce the missed-candidate case, preserve the exclusions and verify large seen-ID histories without SQLite parameter-limit failures. No live orders were sent during validation.

This release includes the 2.36.5 visual refresh. Finding more candidates does not prove profitable trading or guarantee fills; the existing evidence qualification remains required for live entries.

ROM PolyBot 2.36.5

ROM PolyBot 2.36.5 refreshes the Terminal and Trade Visualizer with detailed blue radar graphics, larger charts, clearer panel spacing, readable wrapping logs and a responsive layout. Hourly profit and loss bars use separate colors with a visible zero baseline.

The compact navigation now keeps every screen accessible, including the Visualizer. Animation controls and reduced-motion preferences are respected. Decorative terminal particles no longer randomly imply winning or losing trades; flow animation is explicitly separated from real fills, and the displayed book-price gap is not presented as validated profit.

Trading logic, sizing and execution rules are unchanged. Type checks, the production build, UI audit and wide/narrow graphics checks passed. Installers remain unsigned; the Apple Silicon disk image is not notarized.

ROM PolyBot 2.36.4

ROM PolyBot 2.36.4 addresses two entry bottlenecks: fresh trade bursts can trigger debounced signal scans before the normal periodic scan, and qualified live candidates are selected by conservative calibrated edge rather than being discarded by an additional heuristic confidence cutoff.

Large Trade and Momentum scans can respond to new validated US exchange receipts once every 10 seconds, with configured periodic fallbacks preserved. Replay uses the same calibrated selection rules as live trading. Strategy settings now explain which flow-score filters apply to Practice and how live net edge is evaluated.

Fee, quote, depth, market minimum, sizing, drawdown and evidence qualification controls remain in place. Tests cover burst capture, low-score qualified candidates, stale or missing evidence and preserved user filters. No live orders were sent during validation. More timely candidate evaluation does not guarantee entries or profits; live trading still requires qualified evidence and available buying power.

ROM PolyBot 2.36.3

ROM PolyBot 2.36.3 adds a persistent, bounded decision history to the Overview. It shows the markets currently watched, recent signal activity, the most common reasons candidates stopped, and the stage and reason for each recent candidate. Practice fills, submitted live orders, submission failures, and later position status are labeled separately, so a submitted order is never presented as a confirmed fill.

The journal records diagnostic market identifiers and gate explanations, not API credentials or order payloads. Repeated identical cycle-wide blockers are saved once per minute; individual candidate decisions are saved as they occur. The history survives a restart and keeps the latest 25,000 decisions for up to seven days.

This release changes observability only. It does not relax entry or risk gates, place a live order during validation, or establish that the strategy is profitable. Practice was validated with zero live buying power.

ROM PolyBot 2.36.2

ROM PolyBot 2.36.2 adds in-app updates. After this version is installed, the app checks for new stable releases automatically, downloads the matching installer in the background, and verifies its SHA-256 checksum against the release before offering it. Settings shows the running version, download progress, and clear retry or cancel controls. A user action is required to install, so an update cannot silently stop a running strategy.

On Windows, the app opens the verified installer and closes for installation. On Apple Silicon, it opens the verified DMG for the user to replace the app in Applications. The current Mac release is not signed for seamless in-place replacement. Users on 2.36.1 or earlier need to install 2.36.2 once through the existing release download; later updates can be obtained inside the app.

This release changes update delivery only. It does not change trading rules, place a live order, or establish profitability.

ROM PolyBot 2.36.1

ROM PolyBot 2.36.1 adds a clear answer to “Why am I not seeing trades?” on the Overview page. It shows the first current obstacle, such as a paused strategy, disconnected Polymarket US API, unavailable buying power, missing qualified live evidence, a stopped market feed, or an old decision cycle. The action beside the explanation opens the relevant page.

The panel separates market trades and signals recorded over the last 24 hours from candidates and entries in the last confirmed scan. An entry created by the app is not presented as a filled order. Practice remains available with simulated funds when the live account has no buying power; live evidence, execution, and risk requirements are unchanged.

This update changes diagnostics only. It does not loosen trading filters, increase trade frequency, place an order during setup, or establish profitability. No live order was placed during validation.

ROM PolyBot 2.36.0

ROM PolyBot 2.36.0 makes the 15-minute crypto engine trade live only what your own recorded windows show to be profitable after fees. Once switched on, the engine used to place real orders for whatever its settings said, whether or not those settings had ever made money. Live crypto entries now wait until replaying your exact settings over the last 14 days of recorded windows shows at least 50 trades over 5 days, a profit that survives a pessimistic loss rate, and a profit in both halves of the period. That replay assumes no fill better than the ask one recording earlier, so a quote that had only just appeared cannot make a strategy look profitable. The recorder logs every window whether or not the engine trades, so the evidence builds without risk. The Crypto page says why entries are waiting; "Require proven edge" in the advanced risk settings turns the rule off. The default Deep Favorite settings measured about −0.7¢ per contract after fees in earlier replays, so expect the engine to wait with them.

Once your settings qualify, a coin whose own replayed record clearly loses is left out of live entries, and the Crypto page names it. A new "Proven edge (Kelly)" bet size is available: it stakes a share of the Kelly amount your evidence supports, a quarter by default, and bets nothing until the settings qualify. It is off unless you choose it.

The crypto model now expects large sudden moves as often as crypto actually makes them. It used a normal curve, which was most overconfident exactly where the engine buys, at 97–99¢. Three standard deviations from the strike, the old model said 99.87% and the new one says 99.41%. "Tail model" in the model settings switches back to the old curve. The model's auto-pause also compares recent high-confidence calls with what they cost, not only with a fixed hit rate, and model entries now cap their order price so the minimum edge you set still holds at the price paid; when the book has moved past that, the entry is skipped and shown as "edge gone".

Backtests now follow the strategy that actually trades. Crypto backtests exit where your stop-loss, take-profit and sell-into-strength settings would, instead of holding every trade to settlement. Main-engine backtests apply the same qualified-edge rule live trading uses, instead of the raw signal score. The Backtest page adds "Compare presets on my data", which runs your settings and each preset through the evidence gate side by side.

With custom entry rules on, the main engine checks your entry-cost rule again at the price it actually pays, so a book that fell after the signal no longer buys below your rule. The private account stream now reports its connection health.

User scripts now run in a restricted Python. They can use the calculation modules the Scripts guide lists, such as math, statistics, datetime and json, but not the file system, the network or the app's stored credentials. A saved script that imports anything else is switched off with the reason shown next to it.

The app now runs on Electron 44, which closes the security advisories in the Electron version it shipped with. Updates are now published from github.com/romanstma-cpu/rom-polybot.

No live order was placed during validation. Live execution and profitability still require verification with a funded, eligible Polymarket US account.