Polybot release notes

Polybot 2.36.6 CURRENT

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.

SHA-256 checksums: Windows Apple Silicon

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.

SHA-256 checksums: Windows Apple Silicon

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.

SHA-256 checksums: Windows Apple Silicon

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.

SHA-256 checksums: Windows Apple Silicon

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.

SHA-256 checksums: Windows Apple Silicon

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.

SHA-256 checksums: Windows Apple Silicon

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.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.35.11

ROM PolyBot 2.35.11 keeps delayed, replayed, and invalid-time Polymarket US trade prints out of the Large Trade scanner. A trade must pass the stream's freshness and duplicate checks before it reaches the scanner, and the scanner checks its event time again before recording a signal. This prevents old prints from appearing as new activity after a reconnect or delayed scan; it does not increase the number of real trading opportunities.

The Overview now names Practice, paused, and live readiness separately. It shows the buying power and evidence still needed for live trading even when Practice is available. A scan skipped because trading is paused or blocked also clears the previous cycle's candidate and placement counts, so the dashboard does not present them as current activity.

The referral text in onboarding and API setup now points to Polymarket US for the current offer amount and qualifying terms instead of promising a fixed bonus. The referral code remains ROMANR and ROM may receive a referral reward.

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

SHA-256 checksums: Windows Apple Silicon

Polybot 2.35.10

ROM PolyBot 2.35.10 makes Practice usable with a connected Polymarket US API and no money in the account. Practice uses a simulated bankroll and can start while live buying power or live execution readiness is unavailable. Live trading remains gated by account funds and qualified evidence.

This release improves the evidence used to judge strategy changes. Busy market snapshots no longer evict recorded signals and settlements during the 60-day evidence window or when the recorder queue fills. Older recorded signals can be matched to public Polymarket US settlement results without placing an order. Practice's correlated-position limit now counts simulated positions separately from live positions, and its fee description reflects the applicable US taker fee.

The desktop window also has a visible recovery screen for renderer errors and a bounded reload attempt after a renderer-process or navigation failure. These changes do not establish that the strategy is profitable or increase the number of qualifying market signals by themselves.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.35.9

ROM PolyBot 2.35.9 clarifies why live trading cannot start. The Strategy checklist now shows whether qualified signal evidence and USD buying power are available, and the Evidence gate reports how many settled event samples it has recorded. A rejected Polymarket US API key appears as an actionable replacement-key message in the connection test and Overview rather than leaving a stale no-funds preflight on screen.

This release does not change entry rules or make an unverified strategy profitable. Live orders still require a valid Polymarket US API key, available buying power, and qualified evidence. Practice remains available with a simulated balance while evidence accumulates.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.35.8

ROM PolyBot 2.35.8 makes live-trading readiness explicit when the Polymarket US API is connected but the account has no USD buying power. The trading-status gate now distinguishes an empty account from an unreadable balance and clears automatically when a successful balance poll sees funds. The Overview and title bar show the no-funds state instead of implying that live orders are ready.

No strategy threshold, sizing formula, or order-routing behavior changed. A funded account and valid API credentials are still required to validate live order placement; these checks do not establish profitability.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.35.7

ROM PolyBot 2.35.7 fixes Settings → Check for updates. The checker now compares the installed version with the newest ROM PolyBot installer for the current platform. The shared repository's generic latest release can point to ROM Trader or another app and previously showed an incorrect version.

Trading logic and risk settings are unchanged. Live trading still requires valid Polymarket US API credentials and qualified strategy evidence.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.35.6

ROM PolyBot 2.35.6 makes quiet trading sessions easier to diagnose. The Opportunity funnel shows how many markets had a fresh, warmed momentum window and the strongest observed directional trade cluster beside the live entry threshold. Markets absent from the refreshed catalog age out of the scanner universe instead of consuming watch slots indefinitely.

This release does not lower the $500 live cluster threshold or claim improved returns. Live trading still requires a Polymarket US API key accepted by the exchange and qualified, fee-adjusted strategy evidence. Start in Practice and compare settled outcomes before changing risk or entry settings.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.35.5

ROM PolyBot 2.35.5 fixes two causes of missed momentum opportunities. The scanner and market stream now follow the same non-expired market universe as it refreshes. The trade tape also accepts exchange timestamps up to two seconds ahead of the local receipt clock; affected trades are scored at receipt time, while larger skew and stale trades remain rejected.

The Opportunity funnel now shows whether trade prints were accepted by the scanner or rejected and names the leading rejection reason. Existing live-entry evidence, order-size and risk rules remain in force. This update does not establish that a strategy is profitable, and live trading still requires a valid Polymarket US API key.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.35.4

• Explains when Polymarket US rejects a saved API key and pauses authenticated streams until a valid key is connected.
• Shows missing qualified strategy evidence before live mode can be started. Practice remains available to collect settled outcomes.
• Refreshes the live market catalog immediately after reconnecting the API.

If the API page says Polymarket US does not recognize your key, create a fresh Key ID and Secret Key in the Polymarket US developer portal and save them in the app.

SHA-256 checksums: Windows

Polybot 2.35.3

• Re-validates Polymarket US credentials while the app runs and pauses live trading when the exchange rejects them.
• Stops authenticated streams while the account is disconnected instead of reconnecting endlessly.
• Bounds the live market subscription universe to the 500 markets the scanner can evaluate.
• Keeps the public market catalog refreshing so the engine can recover after reconnecting.

If the API page says Not connected, generate a fresh Key ID and Secret Key in the Polymarket US developer portal, then save them in ROM PolyBot.

SHA-256 checksums: Windows

Polybot 2.35.2

This release repairs the Polymarket US market intake path and makes feed failures visible before live trading.

• Keeps valid tradable markets when the US reference API omits rolling-volume fields.
• Limits live WebSocket subscriptions to the 500-market scanner universe.
• Shows rejected subscriptions and empty market feeds as a blocking readiness issue.
• Preserves quote, spread, evidence, position-size, and risk controls.

Available for Windows 10/11 and Apple Silicon Macs running macOS 15 or later.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.35.1

Version and update visibility

• Settings now displays the exact version of the executable currently running.
• Check for updates compares that version with the latest public ROM PolyBot release.
• When an update exists, Settings links directly to the official GitHub release.
• The version panel explains that it identifies the active executable when Windows contains more than one installation.

Large evidence stores

• The candidate-funnel report now reads a bounded signal sample and the latest runtime blocker independently.
• An overnight database with heavy market traffic can no longer overflow this Evidence report.
• Calibration continues using its existing bounded, per-evidence loader.

Small bankroll verification

• The current engine was verified against the user's saved database with a $3.15 balance.
• New entries are blocked only below one cent; normal sizing, exchange minimums, available buying power, calibration, and account-risk controls still apply.

Validation

• 2,003 Python tests pass.
• TypeScript checking, the strict UI audit, capacity probe, and production builds pass.
• The packaged backend passes its native dependency and signing self-test.

Polybot 2.35.0

Trade-frequency diagnosis

• Adds an event-deduplicated candidate funnel to the Evidence screen.
• Shows how many recorded Large Trade and Momentum candidates cleared the signal gate.
• Groups rejected candidates by confidence, edge, category, resolution horizon, entry price, signal type, data quality, and custom rules.
• Shows the latest engine-wide blocker, including missing qualified calibration evidence, account-risk stops, unavailable balances, and recovery pauses.
• Keeps the report observational: it cannot lower a threshold, size a position, or place an order.

Repeated scans of the same event count once per signal source, preventing a busy market from inflating the apparent sample. Existing live admission, calibration, position-sizing, risk, and execution behavior are unchanged.

Validation

• 2,002 Python tests pass.
• TypeScript type checking passes.
• Strict UI audit and Electron locator-drift checks pass.
• Capacity probe and production builds pass.

Polybot 2.34.0

Micro-bankroll trading update

• Supports live percent sizing with balances below $5 by reserving the actual fee-adjusted cost of one whole contract.
• A $4 bankroll can place a one-contract order when it fits the user’s cash reserve, exposure, hard cap, market depth, and the market’s minimum quantity.
• Removes the arbitrary $5 minimum from portfolio replay.
• Practice bankrolls now support values down to $0.50 for realistic low-balance testing.
• Clarifies the automatic small-balance behavior in the sizing UI.

This does not force an order that violates an exchange minimum or a saved risk limit. Validated by the complete release gate and packaged-backend self-test.

Polybot 2.33.0

Strategy evidence upgrade

• Adds fixed hierarchical calibration pools so valid same-category evidence can qualify when exact price/score groups are too sparse.
• Always prefers the most specific qualified group.
• Never shares evidence across signal source, side, category, or scoring version.
• Every fallback independently passes chronological holdout, recency, accuracy, fee stress, stability, and conservative-return checks.

Validated with the full Python suite, TypeScript checks, UI audit, capacity probe, production build, and packaged-backend self-test.

Polybot 2.32.0

Strategy preset correction: replaces stale World Cup and conflicting category assumptions with evidence-gated Edge Stack selection across available categories. Crypto Large Trades is now clearly labeled research-only. Live entries still require chronological holdout evidence after fees.

Polybot 2.31.1

Bugfix release: prevents Polymarket US buying power from having pending orders deducted twice, and fixes non-monotonic large-trade size scoring. Includes regression tests, the Windows installer, and SHA-256 checksum.

Polybot 2.31.0

ROM Polybot 2.31.0 renames Whale signals to Large Trade Signals, adds clearer public-order-flow guidance, and surfaces category filters that are blocking recent opportunities. Includes the Windows installer and SHA-256 checksum.

Polybot 2.30.0

Current fees and stronger ML evidence

• Prices Polymarket US trades with the exchange-wide fee schedule effective September 17, 2026 while preserving the correct rates for historical backtests.
• Requires evidence from at least 20 independent resolution days before the ML model can qualify for promotion.
• Adds day-clustered confidence floors for Brier-score improvement, log-loss improvement, and fee-adjusted shadow return.
• Prevents correlated same-day outcomes from looking like independent proof.
• Expands the Evidence workspace with independent-day counts, confidence floors, and deterministic resample details.
• Keeps the model outside live order approval, routing, and sizing.

The models remain in shadow mode and cannot change live order decisions. Improved live returns have not been demonstrated.

The Windows installer and Apple Silicon app are unsigned. Verify downloads using the published SHA-256 checksums before opening them.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.29.0

Machine-learning evidence lab

• Records versioned pre-trade market features, including spread, depth, imbalance, microprice, quote age, and signal context.
• Adds a leakage-resistant shadow ranker with chronological holdout, an embargo, event deduplication, regularization, and conservative probability shrinkage.
• Adds shadow execution models for fill probability and 120-second adverse selection.
• Adds an append-only forward prediction ledger that scores only later-settled outcomes against the market baseline.
• Shows model readiness and evidence quality in the Evidence workspace.

The models run in shadow mode and cannot change live order decisions. They must earn promotion through forward evidence first. Improved live returns have not been demonstrated.

The Windows installer is unsigned. Verify it using the published SHA-256 checksum before opening it.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.28.1

This Windows release makes the desktop workspace easier to scan and easier for a new user to complete setup.

• Trade Readiness puts the four controls that affect the next entry first.
• Supporting account safeguards remain available without crowding the main view.
• The Strategy page gives an unconnected user a direct path to Polymarket US API setup.

Trading logic, saved settings, order routing, and risk controls are unchanged.

SHA-256 checksums: Windows

Polybot 2.28.0

ROM Polybot 2.28.0 completes a production-hardening pass without loosening trading or risk rules.

• Execution and market-data requests have separate capacity lanes.
• Trade Readiness now includes database, disk, stream, circuit and API health.
• Critical readiness failures prevent live mode from starting.
• Trace IDs connect RPC and strategy-cycle logs during diagnosis.
• The watchdog recovers even when dependency cleanup stalls.
• Public builds must pass the complete test suite, a 3× peak-capacity probe, packaged backend self-tests and platform-specific launch checks.

Trading starts paused. Use Practice first and review your limits and evidence before enabling live orders. Improved live returns have not been demonstrated.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.27.0

ROM Polybot 2.27.0 adds entry last-look protection.

• Fetches a fresh market quote immediately before every live main-strategy submission.
• Skips an entry if its required price worsened after sizing.
• Skips a crossing entry if the displayed liquidity required for its size disappeared.
• Accepts price improvement without raising the original order limit.

This release reduces stale-price and vanished-liquidity entry risk. It does not guarantee fills or improved returns. The Windows installer is unsigned and the Apple Silicon installer is not Apple-notarized. Verify each download against its published SHA-256 file.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.26.0

ROM Polybot 2.26.0 adds post-fill execution measurement and a conservative adverse-selection guard.

• Live main-strategy fills record side-aware movement after 30, 120, and 300 seconds.
• Comparable entries can pause only after 30 independent market-day observations across at least 10 days and 8 markets show clearly unfavorable movement.
• The Evidence screen displays guard collection progress and any active pause.

This release does not claim guaranteed or improved returns. The Windows installer is unsigned and the Apple Silicon installer is not Apple-notarized. Verify each download against its published SHA-256 file.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.25.0

ROM Polybot 2.25.0 adds a qualified-edge safety gate for live main-strategy orders.

• Live Whale and Momentum entries require a recent signal group that passed chronological holdout accuracy and after-cost return checks.
• The gate applies to percent, fixed-contract, and Kelly sizing.
• Practice continues collecting otherwise eligible candidates so evidence can grow while live entries wait.
• Qualified candidates are ranked by conservative after-cost edge.
• Evidence and Strategy screens explain the live-entry requirement and current qualification progress.

This release does not claim guaranteed or improved returns. Signal groups can stop qualifying when evidence becomes stale or market behavior changes. Existing position limits, exposure limits, cash reserve, execution checks, and emergency controls still apply.

The Windows installer is unsigned and the Apple Silicon installer is not Apple-notarized. Verify downloads against the published SHA-256 files.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.24.0

Safer live execution

• Shows a degraded price-stream state when a connected, subscribed WebSocket stops delivering data. Old stream books are never used; the app uses a fresh REST quote when available.
• Keeps the existing circuit breaker for repeated quote or order failures, with a clear status in Trade Readiness and Main Engine.
• Detects significant computer clock drift before presenting an API connection as failed.
• Paces Polymarket US requests and retries safe read requests with bounded backoff.

Better risk and routing

• Measures drawdown from transfer-adjusted account equity so deposits and withdrawals do not look like trading profit or loss.
• Adds safety alerts for paused execution, order recovery, authentication, and drawdown conditions.
• Makes the entry spread limit and chase limit configurable, while keeping depth, fee, and risk checks in force.

No strategy can guarantee profit. Start in Practice mode and review all live-risk settings before enabling live orders.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.23.0

Evidence-prioritized market selection

• When several live Whale or Momentum candidates compete for limited account capacity, Polybot now considers sources with qualified positive settled-practice evidence first.
• New, unqualified, or conservatively non-positive sources remain at 25% starter priority and size while starter gating is enabled.
• Practice ordering stays unchanged so challenger results remain comparable, and Kelly mode retains its calibrated net-capital-return ranking.
• The rolling evidence plan is cached for five minutes so candidate selection does not repeat its bootstrap calculation every scan.

The selector requires the existing independent-event thresholds across both 30-day and 90-day windows. Missing evidence does not claim an edge. Historical practice results do not guarantee live fills or future profit.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.22.0

Clearer trading workspace

• The Overview screen now presents the strategy as a readable four-stage pipeline: signals, live quote, risk checks and order routing.
• Refined navigation, responsive spacing, typography and terminal styling make the app easier to scan without adding decorative market claims.
• The Strategy page now explains real-account sizing and the separate practice balance in plain language.

Balance-aware entries

• Polybot refreshes available funds before every scan and stops using the old $5 account floor.
• Small percent-sized accounts can allocate up to $0.50 when a whole contract fits.
• Existing cash reserve, position, portfolio, related-outcome, fee, order-book depth and exchange-minimum limits remain authoritative.

A $0.50 balance does not guarantee an eligible order. The current market price, whole-contract minimum, available depth, fees and saved risk limits must all permit it. Strategy scores are heuristics and improved returns have not been demonstrated.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.21.0

Safer live sizing

• Whale and Momentum sources now begin at 25% live size until they have a settled, independent practice record across both the recent and longer evidence windows.
• A source with a non-positive conservative settled-practice result remains at 25% size. Qualified positive evidence restores normal size; the optional Evidence Allocation setting can still promote it above normal size.

Better execution fit

• Qualifying live entries now receive a market-quality adjustment based on observable spread, movement since signal, estimated execution fees, and displayed depth.
• The adjustment only reduces size on weaker but still valid books. Existing quote, depth, fee, exposure, and loss-limit protections remain in force.

These controls are execution and risk-management safeguards. They do not predict returns or guarantee profitability.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.20.0

Public-launch safety update

• Live trading now requires an explicit acknowledgment of real-order and loss risk after the user reviews saved limits.
• Emergency stop pauses the main strategy first and then requests cancellation of its pending orders. It does not liquidate existing holdings; exits remain a separate, reviewed action.
• First-use guidance now follows Connect API → Set limits → Run practice before live orders are enabled.
• The Logs page can export a sanitized support report. API keys, secret keys, passphrases, bearer tokens, authorization values, and Discord webhook URLs are redacted before export. Review a report before sharing it.

Validation

• TypeScript type checks, production build, and packaged backend self-test passed.
• Public live-review, Emergency stop, and liquidation safety tests passed with no live credentials or real orders.
• Full Python suite passed: 1,753 passed, 140 skipped.

Scope

Emergency stop controls the main whale/momentum strategy only. Separate crypto, copy, and script engines remain separately armed and must be stopped from their own controls.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.19.0

Maker-first execution

The main strategy now includes maker-join. It improves or joins the best bid without crossing the current offer, sends the Polymarket US participateDontInitiate flag, and is therefore rejected by the exchange if it would immediately take liquidity. New installations select this route by default. Existing saved order-style choices remain unchanged.

Maker entries are rechecked immediately before submission and an unfilled remainder is canceled after 12 seconds by default. Partial fills remain fully accounted for while cancellation is confirmed. The app records confirmed maker fees or rebates, but it does not assume a rebate when sizing a trade.

Practice mode cannot know maker queue position, so its one-shot fills continue to use conservative taker pricing and dated fees. Portfolio replay requires a later recorded book to trade through a passive quote and still assumes no maker rebate.

Faster market data

The main order path now consumes a full authenticated US WebSocket book when it is at most two seconds old and falls back to the REST book after reconnects or during warm-up. The new-install trade scan cadence is five seconds. The private WebSocket still drives order/fill reconciliation, with bounded REST polling as a disconnect and maker-expiry fallback.

Validation

• Full Python backend suite passed, including maker pricing, post-only payload, WebSocket freshness, final quote recheck, conservative replay, accounting, risk, and recovery tests.
• TypeScript checks, configuration-boundary tests, production frontend build, UI audit, packaged backend self-test, packaged Windows startup/settings smoke, calibration UI, and portfolio-replay UI passed.
• The native Apple Silicon build runs the backend self-test and packaged app smoke test in GitHub Actions, and its disk image is verified before release.
• The Windows installer SHA-256 is published separately. Mac hashes are copied from the native release artifacts.

Profitability status and rollout

This release does not demonstrate improved historical or live returns. Maker-first routing should reduce entry cost on filled orders, but it can also lower fill rate and create adverse-selection risk.

Before increasing size, compare crossing with maker TTLs of 6, 12, and 20 seconds under 250/500/1000 ms latency and 25/50/100% depth stress. Track net return on capital, maximum drawdown, fill rate, cancellation rate, capital hours, and 5/30/120-second post-fill markout. Require at least 200 terminal maker attempts across 20 trading days, 50 fills, positive results in both time halves, and a positive day-block-bootstrap lower bound. Then complete a separate 14-day one-contract forward sample.

The detailed priority roadmap and exact pilot configuration are included in the desktop source at docs/QUANT-ROADMAP-2.19.0.md.

Platform notes

ROM Polybot requires Polymarket US API credentials. Windows is 64-bit. Mac Apple Silicon builds require macOS 15 or later and are not Apple-notarized; follow Apple's documented opening guidance and verify the published hashes.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.18.0

Strengthens the evidence-based Kelly strategy. Percent and contract sizing modes retain their existing signal selection, and this release does not switch sizing modes or enable live trading.

Strategy changes

• Prediction accuracy alone no longer qualifies a score group. The trades that its training estimate would select must also pass net-return holdout checks.
• The settlement simulation charges two cents of adverse entry movement, conservative scheduled fees and another cent of uncertainty. It does not assume maker rebates or claim that a quoted order would actually fill.
• Qualification needs at least 40 selected holdout trades across seven UTC days, a positive lower bootstrap estimate using day blocks, and positive net P&L in both chronological halves. Existing training, embargo, event deduplication, complete-holdout, recency and prediction-quality requirements remain.
• Live Kelly scans and historical replay prioritize qualified candidates by conservative stressed return on capital instead of raw confidence minus price. Fresh quotes, depth, fees, exposure limits and sizing are still checked at entry.

The bootstrap estimate is a historical filter, not a guaranteed confidence bound under changing or dependent markets. The simulated check assumes holding one contract per selected event to settlement; it does not validate early exits, portfolio drawdown or actual live fills.

Validation

• 1,743 backend tests passed and 139 skipped.
• Tests reject gains erased by costs, recent deterioration, sparse samples, stale/future models and an accuracy-improving model whose selected trades lose.
• TypeScript, the production build, packaged Windows startup, calibration UI checks and portfolio replay UI checks passed without real orders.
• Fee assumptions were checked against https://docs.polymarket.us/fees.
• Native Apple Silicon and Intel builds passed backend self-tests, packaged startup/settings checks and disk-image verification.

Before using real money

Collect main-strategy signals and subsequent settlements. Keep parameters fixed while evaluating later periods, then compare portfolio replay across multiple periods, including losing periods and adverse execution scenarios. Track fill rate, net returns and drawdown in a separate forward sample before increasing size. This release has not demonstrated improved live returns. Kelly remains blocked for groups without qualified evidence.

Mac packages are not Apple-notarized. Review https://support.apple.com/102445 and verify the published checksums before installing.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.17.4

Improves accounting while an entry-order cancellation awaits confirmation.

• Newly confirmed whole-contract fills update the position quantity, average price, fees and cost immediately, even while cancellation remains pending.
• The order retains its recovery status. Recording a fill does not imply that the remaining quantity was canceled or permit new entries.
• Incomplete accounting, fractional fills and quantities above the order size retain an accounting block and the existing reservation. Cancellation-pending status no longer overrides that accounting protection.
• Repeated updates do not duplicate fills, and older snapshots cannot reduce the confirmed quantity. Positions with recorded exits keep their existing protection against entry snapshots restoring sold quantities.

Verification: 1,737 backend tests passed and 139 skipped, including additional fills during cancellation, idempotent updates and incomplete accounting. All exchange responses in these tests were mocked; no real orders were placed.

Windows and both native Mac packages passed backend self-tests and packaged startup/settings checks. The Mac disk images passed native verification.

Mac builds are not Apple-notarized. Review https://support.apple.com/102445 and verify the published checksums before installing. No claim of improved live profitability is made.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.17.3

Fixes entry-order expiry after a partial fill. Previously, automatic expiry only canceled orders with zero fills, leaving a partially filled order's remaining quantity resting beyond the configured deadline.

Expiry now requests cancellation when an open journaled entry has any unfilled quantity. Confirmed fills remain in the position. If an order finishes filling during cancellation, exchange evidence determines the final quantity and cost. An unconfirmed cancellation stays pending with funds reserved.

The configured expiry duration is unchanged. Disabling expiry still disables automatic expiry, and orders before their deadline remain open.

Verification: 1,731 backend tests passed and 139 skipped, including cancellation races, unconfirmed cancellation, disabled expiry and not-yet-expired orders. The Windows production build, backend self-test and packaged startup passed. Tests used mocked exchange responses; no real orders were placed.

Native Apple Silicon and Intel packages also passed backend self-tests, packaged startup/settings checks and disk-image verification.

Mac builds are not Apple-notarized. Review https://support.apple.com/102445 and verify the published checksums before installing. This fix is not evidence of improved live profitability.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.17.2

Execution feedback now filters orders in the database instead of loading every order from the last 30 days for each candidate. New indexes select the relevant market, source and route. The Evidence report still covers the full period.

The fee and poor-fill thresholds are unchanged. Regression tests verify matching results, price-range boundaries, and preservation of existing orders when the indexes are added. Equal timestamps now have a deterministic order.

In a local, in-memory synthetic test with 20,000 orders and 20 matches, the median read took about 0.1 ms versus 85 ms for the full-history read. This measures the database step only; it does not establish faster exchange execution or returns.

Backend validation: 1,726 tests passed and 139 skipped. The focused execution tests also passed after the final ordering change.

Windows, Apple Silicon and Intel packages passed backend self-tests and packaged startup/settings checks. Both Mac disk images passed native verification.

No real orders were placed for testing. Mac builds are not Apple-notarized; review https://support.apple.com/102445 and verify the published checksums.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.17.1

Fixes a settings-validation mismatch that rejected supported order styles and could prevent full saved configurations from being restored.

• Accepts the app's existing limit_cross, limit_mid and market price styles.
• Continues rejecting time-in-force values such as FOK as invalid price styles.
• Adds a regression check against the entire shipped default configuration, including verification that no settings are silently dropped.
• Packaged startup checks now exercise order-style updates, full configuration restore and invalid-value rejection while trading remains disabled.

TypeScript checks, config regression tests, the production build, frozen-backend self-test and packaged Windows checks passed. This release does not change strategy signals, routing implementation or risk limits.

Native Apple Silicon and Intel builds also passed backend self-tests, packaged settings-restore and startup checks, and disk-image verification.

Includes the execution feedback introduced in 2.17.0. No live profitability claim is made. Mac builds are not Apple-notarized; consult https://support.apple.com/102445 and verify the published checksums before installing.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.17.0

This release adds execution feedback for live main-strategy entries.

Changes

• The Evidence page shows confirmed fill quantity, recorded submission timing, signal-to-fill price changes, and fees per contract over the last 30 days.
• New entries record their signal price and submission quote alongside the durable order journal. Repeated exchange updates do not duplicate fills.
• Entry margins now include the fee reserve. With sufficient live evidence, higher observed fees can reduce order size or prevent an entry. Kelly's existing fee allowance is not charged twice.
• Consistently poor fills can pause a market/source/route/price group after at least 20 completed orders across 10 days. Evidence expires after 30 days. Open, uncertain and rejected submissions do not count as completed no-fills.
• Crossing/resting describes the quote at submission, not an exchange-confirmed maker/taker role. Order routing and position exits remain unchanged.

Verification

• 1,724 backend tests passed; 139 skipped.
• All 17 end-to-end suites passed in isolated profiles without real orders.
• TypeScript, config validation, the 25-screen UI audit, production build, frozen-backend self-test and packaged Windows startup passed.
• Native Apple Silicon and Intel packages passed their frozen-backend self-test, packaged startup check and disk-image verification.

Limits

Only newly instrumented live main-strategy entries build this execution record. Practice fills are excluded. Sparse history leaves learned adjustments inactive; the scheduled fee reserve still applies. These changes have not demonstrated improved live profitability, and no real orders were placed for validation.

The Mac packages are not Apple-notarized. Review Apple's opening guidance at https://support.apple.com/102445 and verify downloads using the published hashes.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.16.0

ROM Polybot 2.16.0 turns recorded practice evidence into an optional, conservative live-sizing adjustment for the main strategy.

Changed

• Strategy now includes Evidence-based allocation for Whale and Momentum signals. It is off by default and applies only to new live main-strategy entries when enabled.
• Each source must qualify in both a 30-day window and a 90-day window. Repeated positions from the same event are combined so correlated rows do not inflate the sample.
• The allocator uses the lower 10th-percentile bootstrapped return from practice outcomes. Missing evidence leaves sizing unchanged. Qualified weak evidence reduces live size to 0.25x; stronger evidence can receive up to 1.50x.
• Existing per-position, cash-reserve, portfolio-exposure, related-outcome, and market-depth limits remain final ceilings after allocation.
• Practice sizing never uses the allocator, preserving an unbiased comparison record. The live-start review states whether evidence allocation is on.

Verification

• 1,844 backend tests collected: 1,705 passed, 139 skipped, and 0 failed.
• All 16 Electron end-to-end suites passed against isolated profiles. The new suite verifies the allocation report, opt-in persistence, narrow layout, and that using the control does not enable live trading.
• TypeScript checks, config-boundary validation, the production build, strict interface audits, the frozen-backend self-test, the Windows installer build, and packaged Windows startup passed.
• Native Apple Silicon and Intel packages passed the frozen-backend self-test, packaged startup smoke test, and disk-image verification in GitHub Actions.

Before you trade

Practice evidence is simulated and may not match live fill quality. Allocation multipliers do not predict returns or guarantee profitability. Trading starts paused, evidence allocation starts off, and this release has not been validated with the publisher's live Polymarket US credentials.

The macOS builds aren't Apple-notarized. Follow https://support.apple.com/102445 when macOS identifies the developer as unknown. Don't bypass a damage or malware warning. Verify the downloaded DMG against POLYBOT-MAC-CHECKSUMS-2.16.0.txt.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.15.0

ROM Polybot 2.15.0 adds an evidence-aware ranking for strategies tested in practice mode.

Changed

• The Evidence page now compares settled practice results for the built-in Whale and Momentum strategies and each custom script.
• A strategy stays visibly Collecting until it has at least 30 settled fills across 20 distinct markets and 7 observation days. Small samples do not receive a rank.
• Qualified strategies are ranked with recorded return on capital at risk, sample-size shrinkage, and realized drawdown. The screen also shows net P&L, win/loss counts, average P&L, and the exact evidence still needed.
• Rankings use practice records only and cannot enable or place live orders.

Verification

• 1,838 backend tests collected: 1,699 passed, 139 skipped, and 0 failed.
• All 15 Electron end-to-end suites passed against isolated profiles. The new ranking suite covers qualified and collecting states at desktop and narrow widths. None of the suites placed live orders.
• TypeScript checks, the production build, the strict interface audit, the frozen-backend self-test, the Windows installer build, and packaged Windows startup passed.
• Native Apple Silicon and Intel packages passed the frozen-backend self-test, packaged startup smoke test, and disk-image verification in GitHub Actions.

Before you trade

Practice rankings describe recorded simulated outcomes. They do not prove that the same orders would fill live or that a strategy will remain profitable. Trading still starts paused, and this release has not been validated with the publisher's live Polymarket US credentials.

The macOS builds aren't Apple-notarized. Follow https://support.apple.com/102445 when macOS identifies the developer as unknown. Don't bypass a damage or malware warning. Verify the downloaded DMG against POLYBOT-MAC-CHECKSUMS-2.15.0.txt.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.14.4

ROM Polybot 2.14.4 tightens custom-strategy risk controls and finishes the cross-platform interface work for Windows and macOS.

Changed

• Custom scripts now use the account-wide related-outcome exposure cap for crypto, market, and signal entries. When a group has limited capacity left, Polybot reduces the order size; when the group is full, Polybot refuses the order. This control applies when Maximum related-outcome exposure is set above 0%.
• Start at login now works on macOS. The setting and credential-storage guidance use operating-system-neutral wording on Windows and Mac.
• Checkboxes and sliders meet the app's minimum interaction-size target. The rendered interface audit reports no undersized controls, overflow, contrast failures, unnamed controls, or renderer errors.

Verification

• 1,834 backend tests collected: 1,695 passed, 139 skipped, and 0 failed.
• All 14 Electron end-to-end suites passed against isolated profiles. The suites didn't place live orders.
• TypeScript checks, the production build, the strict interface audit, the frozen-backend self-test, the Windows installer build, and packaged Windows startup passed.
• Native Apple Silicon and Intel packages passed the frozen-backend self-test, packaged startup smoke test, and disk-image verification in GitHub Actions.

Before you trade

Start in practice mode and review your strategy and risk limits before you enable live trading. This release has not been validated with the publisher's live Polymarket US credentials and does not promise profitability.

The macOS builds aren't Apple-notarized. Follow https://support.apple.com/102445 when macOS identifies the developer as unknown. Don't bypass a damage or malware warning. Verify the downloaded DMG against POLYBOT-MAC-CHECKSUMS-2.14.4.txt.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.14.3

Windows and macOS downloads include slightly larger compact controls and more space around expandable settings. Trading rules and saved risk limits are unchanged.

Mac downloads are separate for Apple Silicon and Intel. Both target macOS 15 or later and bundle Python; you do not need a separate Python installation. Open the disk image and drag ROM PolyBot to Applications.

These Mac builds are not Apple-notarized. Review https://support.apple.com/102445 before opening an unidentified-developer app. Do not ignore damage or malware warnings. Verify the download using POLYBOT-MAC-CHECKSUMS-2.14.3.txt.

Validation: TypeScript and Windows production build passed; isolated layout and live-start checks passed. Each Mac build must pass the packaged backend self-test, disk-image verification, and packaged renderer/backend startup test before its link is published. No live trading was performed.

SHA-256 checksums: Windows Apple Silicon

Polybot 2.14.2

Three releases rolled into one: risk enforcement (2.14.0–2.14.1) and a full interface pass (2.14.2).

The correlated-exposure cap now binds every engine. The account-wide group-exposure query counted all three engines' fills, but only the main strategy ever read the remaining allowance. crypto15m and copy_trader could open past a limit their own positions were helping to reach. Both engines now refuse an entry into a full group and size one down to what the group has left. 25 new tests, 23 confirmed to fail against the previous source.

Interface measured, then fixed. Every page was driven in the packaged app at 1440px and 420px and measured for contrast, type scale, spacing, hit targets, accessible names, and overflow. Findings:

• Sub-11px text raised to 11px across 18 files (171 strings).
• Terminal contrast ramp fixed — two steps were at 2.8:1 and 1.9:1; now four steps that each clear 4.5:1.
• loss and indigo fill colours gained text-weight shades at 106 call sites.
• Type, spacing and radius scales unified across the app.
• Duplicate timestamps stripped from the trade log — the backend echoed a full timestamp inside the message that was already shown on the left.
• Guide rewritten to a capped measure with the content a new operator needs.
• Four unlabelled controls fixed via a shared Field component; Scripts toggle enlarged to 44×24.

Three backend failure modes closed. cancel_pending journaling now uses the same synchronous=FULL and BEGIN IMMEDIATE discipline as every other pre-network write. Attaching an exchange id already owned by another intent raises RecoveryRequired instead of a raw IntegrityError. require_entry_depth is now validated against the clamp list.

1,831 backend tests (0 failures), TypeScript check clean, 13/13 e2e suites green, 58 locators with no string drift. The design-system probe ships as npm run check:ui. No real trades were placed and no profitability improvement is claimed. The Windows installer is unsigned.

SHA-256 checksums: Windows

Polybot 2.13.1

Patch release: accessibility hardening, a performance improvement for the positions list, and a startup-crash fix.

Startup crash fixed. A React hook call had been left at module scope during a refactor, which crashed the renderer on launch with Cannot read properties of null (reading 'useContext'). The hook now lives inside its component, and the app boots normally again. Caught by the arming e2e suite, which validates the packaged app from a clean profile.

Keyboard navigation for every dialog. The four remaining full-screen overlays — the Boss Fight milestone arena, the Scripts arm and AI Context Pack dialogs, the full-reset confirmation, and the shareable-stats card — now trap focus (Tab/Shift+Tab cycle, focus restore on close, background scroll locked) and expose proper dialog semantics (role="dialog", aria-modal, descriptive labels). Keyboard users can no longer tab out of a modal into the page behind it.

Faster positions list. The Positions page now renders only the rows in view (with a small overscan) inside a fixed-height scroll container, instead of mounting the full list at once. Filters, sorting, grouping, counts, and styling are unchanged; the sticky header stays visible while scrolling. Large portfolios scroll smoothly.

Config-parity guard. A new bidirectional drift test asserts that every key in the backend default config appears in the frontend type and the IPC validator — and the reverse — so the three sources of truth cannot silently diverge. It passes today with zero drift.

1,538 backend tests, TypeScript checking, the production build, and all nine Electron UI suites pass. The arming suite now retries its first click for up to six seconds to tolerate slow cold starts instead of flaking. No real trades were placed and no profitability improvement is claimed. The Windows installer is unsigned.

SHA-256 checksums: Windows

Polybot 2.13.0

Hardening and interface honesty: fewer ways for the terminal to overstate what it knows or did.

Kill switch tells the truth. Flattening now reports the actual result of the exit request. A completed request no longer implies every position was sold — exits depend on live quotes and displayed depth, so some holdings can remain open — and the confirmation explains that execution may leave positions open and does not pause the strategy. Failed requests show their reason instead of a success message.

Overview reads like a control panel. A latest-decision-cycle readout shows what the last scanning cycle concluded, and three aligned metrics summarize balance, equity and exposure. Missing data stays unknown rather than reading as zero. Metrics and command cards stack cleanly in narrow windows. MainActivity now shows the leading-entry filter counts that the backend actually scored.

Navigation respects your input. Global shortcuts (Ctrl+1..9) ignore text fields, composition input and any open dialog, so typing is never hijacked. The onboarding copy matches the maintained UI tests.

Fewer edge-case crashes. Technical indicators reject NaN and infinite values instead of letting them poison EMA, VWAP and momentum math; the CLOB parser ignores malformed or out-of-range book levels; copy-trader sizing uses exact integer cents instead of float division, which was off by one contract at clean penny prices.

1,535 backend tests, TypeScript checking, the production build, the bundled-backend selftest, and nine Electron UI suites pass, including a strict terminal-review suite that checks the 600px layout, reduced-motion behaviour, shortcut guards and both rejected and zero-fill exit feedback. No real trades were placed.

No profitability improvement is claimed: these changes make reported execution honest and remove crash and loss modes. Profitability and live authenticated trading remain unverified. The Windows installer is unsigned.

SHA-256 checksums: Windows

Polybot 2.12.0

Four strategy upgrades, focused on removing false signals and unbounded loss modes.

Momentum timing (2.11). Flow is measured from bounded per-market trade windows with deduplicated receipts, warm-up and a freshness requirement. Rolling 24-hour volume differences and untimed cross-market trade lists no longer produce signals, so old prints can no longer look like fresh buying or selling pressure. An unavailable window reports as warming up or stale, never as zero flow.

Account-wide risk. A related-outcome exposure cap bounds markets that share a series or event, which per-event caps miss. A peak-equity high-water mark pauses new entries after a confirmed drawdown, and a partial recovery still measures against the true peak. A daily stop supported by realized, settled results now blocks immediately instead of waiting out a persistence delay. Both new limits default to off.

Executable depth at entry. Quotes now carry the resting size behind each price. Entry sizes against displayed depth, shrinks to what the book actually shows, skips markets that cannot support the minimum tradable size, and charges the depth-weighted cost against the signal margin before the entry test.

Exit discipline. Removed a 1-cent sell fallback that could dump a position whenever a bid was unavailable. Exits now require a live quote, concede at most a configured budget below the touch, and are bounded by displayed bid depth. A missing quote holds the position and says so.

1,091 backend tests, TypeScript checking, the production build, the bundled-backend selftest and six Electron UI suites passed. No real trades were placed.

Strategy scoring is otherwise unchanged, and no profitability improvement is claimed: these changes reduce loss modes and make reported execution honest. Because momentum measurement changed, previously collected momentum evidence is not comparable to new evidence and is calibrated separately. Profitability and live authenticated trading remain unverified. The Windows installer is unsigned.

SHA-256 checksums: Windows

Polybot 2.8.0

• Saves order intent before submission; uncertain responses retain their risk reservation.
• Blocks duplicate submissions and confirms cancellations against exchange order state.
• Records execution quantities, prices and fees, with reconciliation after disconnects.
• Reconciles main-strategy partial sales and late fills without counting proceeds twice.
• Missing execution prices, fees or unsupported fractional fills require recovery instead of invented accounting.
• 809 backend tests, desktop checks and the bundled-backend selftest passed. No real trades were placed.

Strategy scoring is unchanged. Profitability and live authenticated trading remain unverified. The Windows installer is unsigned.

SHA-256 checksums: Windows

Polybot 2.7.0

• Refined terminal overview with a clearer setup path, aligned account metrics and mode-aware practice balances.
• Added an in-app live-start review showing saved risk limits, keyboard navigation and connection-loss protection.
• Reject signals that expire while waiting for execution data, and reject entry quotes older than five seconds after retrieval begins.
• Include the simulated execution-cost allowance when sizing practice entries and checking minimum order sizes against the position budget.

Execution safeguards do not establish a profitable strategy. The local trade ledger had no completed trade sample for evaluating returns. Live authenticated trading remains unverified. Practice results are simulations and can differ from actual fills and fees.

Windows installer is unsigned. Verify its SHA-256 checksum before installation. New installations start with live trading disabled.

SHA-256 checksums: Windows

Polybot 2.6.0

• Shared main-strategy status across Overview, sidebar, header and Strategy. Enabled configuration is labelled separately from confirmed execution.
• Failed or timed-out activity reads no longer leave stale scanning visible. Offline takes precedence. Scanning requires a recent decision cycle.
• Starting setup groups mode controls and a readiness checklist immediately above saved risk limits.
• Removed duplicate live/practice switches in Advanced so all UI starts respect saved-limit and connection guards. Pause remains available for an enabled mode.
• Typecheck, production build and isolated strategy-clarity Electron checks passed. No real trading performed.

The strategy algorithm and backend bundle are unchanged from 2.5.

SHA-256 checksums: Windows

Polybot 2.5.0

No release notes were published for this version.

SHA-256 checksums: Windows

Polybot 2.4.0

No release notes were published for this version.

SHA-256 checksums: Windows

Polybot 2.3.0

No release notes were published for this version.

SHA-256 checksums: Windows

Polybot 2.2.0

No release notes were published for this version.

SHA-256 checksums: Windows