POLY | UMA DIGEST

tly (bypassing bot quotes) is being discussed informally.

POLYMARKET DIGEST — 25 SEPTEMBER 2026

⚠️ Bugs & Issues

Latency & API problems are severe and ongoing. Multiple users report POST order latency of 320–13,000ms+ on co-located servers. pepesong noted the issue has persisted for 2–3 weeks, with brief improvements followed by regression. The CLOB, UI, API, and WebSocket are all reported as broken simultaneously. The webpage is also loading 2–3s slower than usual after refresh.

A CLOB error with negRisk contracts (py-sdk issue #305) remains unresolved despite repeated requests for acknowledgement in both #bug-reports and #devs.

A Gamma API bug was reported by im.the.wizard: a valid after_cursor returns HTTP 403 when end_date_min is present in the query, but 200 without it — reproducible and confirmed with specific URLs.

The leaderboard is reporting incorrect data. A workaround using datadash.xyz was suggested.

A display bug was flagged showing -100 instead of +100 in one market.

Limit order panel is also bugged — shares get stuck after placing a limit order, requiring a wait before placing another.

Paris highest temperature market (Sep 25) has no NOAA data since the previous evening (~10:12 PM local time), with louisbodin noting it is not a NOAA issue but likely a station problem at Le Bourget (LFPB), an unstaffed secondary airport.

Rewards stuck at 0 reported by moonshotnft.

---

🔥 News & Announcements

Lower temperature markets reduced: ak504111 confirmed that several lowest-temperature markets for equatorial cities are being removed to consolidate liquidity, as those cities had low volume and the lowest-temperature format was less relevant there.

tylermulvi (co-founder of VoxPop, a Polymarket builder) is seeking a contact to fast-track a session-keys + Combos RFQ enablement request submitted to builder@ on Sep 18, with no response so far.

---

📋 Clarifications

"Will Netanyahu use a visual aid during his UNGA speech?" has been resolved Yes. Polymarket issued an official clarification: Netanyahu's display of a Starlink communication device during his UN General Assembly address qualifies under the rule requiring him to "raise a physical object bearing an image, map, chart, graphic, or text toward the audience or camera." The order book was cleared at 9:00 PM ET on Sep 24.

Sep 20–21 weather data outage: sanhwy reported that Synoptic has confirmed the missing observations were caused by an attack on their data ingestion, and the data has since been backfilled through their API — but too late for affected Polymarket markets. Under current rules, revisions after the first datapoint of the following date are not considered. sanhwy is asking whether Polymarket will review rules to allow waiting for official backfills or fallback sources in confirmed outage scenarios. No official response yet.

---

⚡ Disputes

Secret maker rebate tier ("rebategate" — maker edition): phoenyx69 identified multiple wallet addresses allegedly receiving 35% maker rebates, listing 7 specific profiles. The community reports this undisclosed tier has existed for at least 40 days. floaty_boi and others note that the previous taker rebategate was never publicly addressed, and this new issue compounds frustration. gr33dy.eth asked how to apply for the higher rebate tier. Polymarket staff have not responded publicly. tehnoobkiddo raised broader concerns about private incentives undermining platform fairness.

User losses from outages and exploits remain unresolved: brenmaelen states they are down $6k due to an outage, with DMs to staff ignored for days. aquilifer is seeking acknowledgement of $2k lost due to ghost fill exploits after they were supposedly fixed, stating they have also been ignored.

Alleged market manipulation: showmedamoney__ flagged the account @smash-to-net as a manipulator and called for a ban. sonderwrld corroborated, claiming the same account took money from them.

Bot front-running in climate markets: dryad0987 reports bots immediately placing limit orders 1 cent above any human limit order, discouraging participation. A community proposal to peer-match human traders direc

UMA DIGEST — 25 SEPTEMBER 2026

🔥 News & Announcements

infinity_d (UMA mod) extended liveness on the Ecuador "US boots on cartel operation" market and the Oct 31 bracket while the team reviewed details. The market ultimately settled YES after photos and reporting confirmed US personnel participated in raids alongside Ecuadorian police. pi7486 later claimed the market misresolved, arguing the evidence showed only "support" roles (excluded by rules) rather than direct participation, referencing a prior clarification where a similar market flipped from 99.5¢ YES to NO.

⚠️ Bugs & Issues

oscarlard built a full website (itf.p4proof.xyz) using ITF's HLS stream timestamps to prove a specific tennis proposal was made 3 minutes 33 seconds before the relevant score (5-7, 4-5) was reached. The tool analyzed 48 proposals across three ITF matches, discriminating 44 valid from 4 P4 proposals with zero false positives or negatives.

📋 Clarifications

The Netanyahu "visual aid at UNGA" market became contentious after he held up a Starlink terminal during his speech. raskell1 argued the visible X/SpaceX logo qualifies as a "graphic" under the rules. jenia9419 and mebeim countered that the terminal bears no readable text or image intended for the audience and does not meet the rule's specific list ("image, map, chart, graphic, or text"). A clarification was requested. The CS2 Exodia/Yaw market was flagged for extension as "highly contentious at best, incorrect at worst" — stradiveri noted it had already been clarified as 50-50.

⚡ Disputes

ITF Tennis P4 dispute (Koike vs Alarcon): oscarlard's timestamp proof showed the Over 21.5 proposal was made ~3.5 minutes before the 5-3→5-4 score change, making it P4. joes49572 (proposer) argued the original dispute was based only on inadmissible logs and assumptions, and that solid direct proof should be required before a P4 ruling. mebeim responded that the disputer's initial reasoning is immaterial — only the facts at hand matter for resolution. joes49572 acknowledged the new evidence is real but maintained the original dispute would have been invalid without it. Discussion broadened into calls for Polymarket to publish clear written rules for P4 evidence standards, rather than relying on precedent and community vibes.

Ecuador cartel operation market: Proposer (armanh01) and girsanov_27 argued US personnel visibly participated in raids with photos, videos and press reports as evidence. pi7486 countered that all reporting described the US role as "support" and "technology provision," which is explicitly excluded by the market's rules, citing a prior clarification (July patrol case). rex416 requested an extension to allow a dispute; armanh01 objected since rex did not dispute himself. The market settled YES without a dispute being filed. pi7486 subsequently claimed the market misresolved.

POLYMARKET DIGEST — 24 SEPTEMBER 2026

🔥 News & Announcements

0x3577 noted that weather market rules need improvement, stating "we are having too many loopholes who are making people lose money from wrong results." They expect the data source to change as NWS is "failing every week," and hinted a rule update is coming soon.

⚠️ Bugs & Issues

A significant CLOB degradation occurred during the day. The order status API went down, causing bots to incorrectly treat filled orders as cancelled — orders were appearing, disappearing, and reappearing in a loop. walem confirmed the order status API was down but noted it did not immediately affect order placement, adding that "the last few times CLOB went down shortly thereafter." polywatchdog confirmed: "CLOB is degraded. Cancel and orders endpoints not consistent." chickenwing_14999 pinged staff noting the Trades API was broken again.

artex6666 raised serious complaints about cancel latency: averaging 250ms on P90 cancels, and getting fills after cancel responses — requesting the taker delay be restored to 500ms or a proper FIFO system be implemented. The issue was described as "catastrophic" over the last two days.

A separate recurring bug causes orders to be filled after cancellation (sometimes 30+ minutes later), reported by artex6666 and berth6833. honeybear0x noted this bug has been unfixed "for weeks."

whaleeddy reported their bot address was banned while trading BTC 5-minute markets, receiving HTTP 400 errors. artex6666 suggested contacting support.

A user (bullbear_95834) noted the GET order endpoint started taking 4–5 seconds to return order details after placement, a delay not present before.

Other reported issues: positions not showing after purchase (silvaselva), inability to open the Venezuela/Israel market (jimmybuzzard), and the stats page disappearing (nz377).

brenmaelen publicly called out staff for ignoring DMs regarding $6,000 lost due to an outage, with suhailkakar responding to look into it.

📋 Clarifications

Crypto Up/Down settlement logic: artex6666 clarified (from personal code and observation) that Polymarket uses the Chainlink report with the latest validFromTimestamp ≤ T (expiration), not the first message received after T. A message arriving after T can still be ignored if its validFromTimestamp predates expiration. artex6666 noted this is based on their own code/observations rather than official documentation.

Taker delays for sports markets: berth6833 shared that the taker delay is 1 second across most sports markets, but 0 seconds for MLB and NBA.

API rate limits: blunkoed clarified that rate limit tiers are automatically updated based on trading volume. CLOB rate limits are per wallet; Cloudflare limits are per IP.

⚡ Disputes

Panama City weather market — 32°C vs 31°C: A dispute is ongoing over whether Panama's temperature reached 32°C. sakula1 argues that Panama's civil aviation authority website showed a METAR reading of 32°C for over an hour before being corrected, and that some bots traded on this early reading. modali1 counters that the NWS (the official resolution source) only ever showed 31°C, and that the order book was being cleared 15–30 minutes before the official METAR timestamp — suggesting foreknowledge or a data anomaly rather than a legitimate 32°C reading. The situation is complicated by late METAR issuance (sometimes 35+ minutes past the hour) and alleged DDoS-like bot traffic hitting the Panama aviation site around METAR release times. No official resolution has been announced. user_v123 noted a 33°C SPECI report also appeared later.

UMA DIGEST — 24 SEPTEMBER 2026

🔥 News & Announcements

alexchunus flagged that some proposers have been resolving TV show/YouTube mention markets as NO based on transcripts alone, sometimes within minutes of release — suggesting either insider access or AI-assisted listening. He recommends watching the videos to catch mistakes.

alexchunus also proposed that non-sports/non-price markets should have a fixed 9-hour liveness window with no extensions, which exquisite_wolf_79651 supported. bromley86 suggested an 8-hour default window to avoid variable-length settlements.

Tennis rewards have apparently dropped to 0.8, with no more racing mentioned.

⚠️ Bugs & Issues

whoooo8888 reported being unable to post in dispute threads despite slow mode showing no previous messages. infinity_d acknowledged the issue as odd.

A potential bot arithmetic error was identified in tennis proposals: the bot proposed over 8.5 s2 and over 21.5 match totals simultaneously, and over 9.5 s2 and over 22.5 match totals simultaneously — suggesting the game and match totals thresholds were being triggered at the wrong time.

⚡ Disputes

Saudi Oil Pipeline market (Polymarket)
Proposed as NO. sachosho noted the resolution source requires official Saudi government confirmation, which has not been issued. gum5683 argued it should resolve YES based on Reuters reporting and confirmed funds are being pooled for a dispute. sachosho maintained that Reuters does not qualify under the stated resolution criteria.

"Will Trump say ballroom?" ribbon-cutting ceremony market
bromley86 proposed NO and the market was contested by nefffooor, who claims to have heard Trump say "ballroom" on the White House livestream at ~1:37 of the linked VOD, confirmed with good headphones. bromley86 countered that the word is inaudible over helicopter noise, and noted that Kalshi resolved the same market NO. infinity_d (resolver) was asked to hold off on settlement. The dispute remains open pending further review, with a ~$2,000 block order in the book. mebeim argued Polymarket should clarify all such markets as NO immediately.

Tennis P4 dispute — Total Games O/U 21.5
oscarlard disputed the proposal as premature (P4), claiming the bot proposed the over 21.5 market when the score was 3-5 in the second set, over four minutes before the score reached 4-5. joes49572 argued the dispute lacks the required standard of proof — no real-time screenshots or video with a visible clock — and that relying on VOD timestamps and UMA proposal timings is purely circumstantial. oscarlard subsequently posted VOD and ITF API evidence: the ITF API records match completion at 19:59:11 UTC, and the VOD corroborates a 4+ minute early proposal. mebeim suggested the approach is workable if corroborated across multiple matches. joes49572 maintained that the dispute should not have been filed without solid prior evidence and that this sets a bad precedent. The dispute remains active pending UMA vote.

New Hampshire Democratic Senate Primary margin market
snowhite1359 questioned whether the market was proposed early (P4), noting that the NH Secretary of State results page carries a disclaimer that tallies are subject to change. im.the.wizard argued the results are effectively finalized since the SoS linked them, and noted the proposer has a strong track record. No dispute filed yet.

📋 Clarifications

A CS2 market (Exodia vs. Yawara, Map 1) was flagged as potentially needing a 50/50 default resolution after HLTV listed the result as a default win. infinity_d directed the user to the appropriate channels.

A cricket market (CRI CYM vs BER) was flagged for urgent resolution to prevent user losses, as two matches between the same teams exist and the correct one had already concluded. infinity_d redirected to the relevant channels.

POLYMARKET DIGEST — 23 SEPTEMBER 2026

🔥 News & Announcements

The Polymarket team is actively investigating weather data source issues. ak504111 confirmed internal discussions are underway about potentially reverting to Wunderground as a temporary fix, while a longer-term solution is being developed. The team also confirmed they are looking into hourly temperature markets similar to Kalshi's new offering.

maskache07 proposed reverting to Wunderground immediately and forming a small private working group of experienced traders to design a permanent solution before any new rules are pushed.

⚠️ Bugs & Issues

Critical trade routing bug: olegs_13 reported losing $3,000 after buying shares in one market only to have the order execute in a completely different market. This has allegedly happened 3–4 times.

Redemption failure: jcsnc99 reported ~$1,577 in resolved winning positions stuck unredeemed since Sep 14 on one wallet, with "Claim all" reverting mid-process when mixing negRisk and standard markets in a batch.

Share balance not recognized: A user in #climate reported being unable to exit a ~$1,600 position, with the UI returning "Not enough shares" despite confirmed holdings. ak504111 escalated and requested a screen recording.

Weather data outages: weather.gov (via Synopticdata) has been experiencing repeated data gaps, affecting multiple city markets including Lucknow, Manila, Cape Town, and Panama. The Panama market saw no 15:00 temperature reading, with the outage confirmed on both aviationweather and Wunderground as well.

X Tracker undercounting: sabari_stocks flagged that the White House X post tracker missed several posts from Sunday's "Summer 250 Photo Dump," reporting 204 original posts for the week of Sep 15–22 and asking for a recount before resolution.

CLOB API false-positive ban: polywatchdog reported receiving a 400 "address banned" error from the CLOB API despite respecting all rate limits, and believes it is a false positive. tsmultra asked for a DM with the proxy address.

Order placement latency: Multiple users reported laggy order placement (~10 seconds). skulldoom97 stated latencies should be better, and chickenwing_14999 confirmed improvements on their end, though artex6666 reported severe latency spikes again later in the day.

WebSocket delivery skew: oleh931816 documented WS market update delays of up to 686ms (p99 >100ms) on AWS eu-west-2, with individual markets stalling while others on the same connection continue normally. 830lincoln confirmed experiencing the same, with delays up to 10+ seconds, noting the issue returned after being fixed about a year ago.

CLOB signature latency disparity: th0rped0 measured that accounts created after May 4, 2026 are locked into the deposit wallet flow (type 3 signatures), which is consistently 6–8ms slower than type 2, and ~10ms slower than legacy type 1 accounts. blunkoed suggested waiting for CLOB 2.0, expected end of year.

📋 Clarifications

Regarding the weather data source debate, notaforecaster_32454 clarified that weather.gov pulls from the Synopticdata API, while Wunderground uses weather.com's API, and suggested https://tgftp.nws.noaa.gov/ as the closest raw source. However, user_v123 noted tgftp only retains 24 hours of history and does not include SPECIs in cycle files, making dispute resolution problematic. The aviationweather.gov API was raised as an alternative with archiving and SPECI support. user_v123 also suggested considering a switch to Celsius to avoid temperature conversion rounding issues inherent to Fahrenheit/METAR T-group parsing.

ak504111 clarified that the rules already allow fallback to Wunderground when WRH data is fully missing, and indicated the team will add clarity on what constitutes "missing data" to potentially cover partial outages as well.

UMA DIGEST — 23 SEPTEMBER 2026

⚡ Disputes

A dispute is active around a Kuusamo market (appears to be a CS/esports match). The game was played on stream with results confirmed on HLTV, but the market was subsequently changed to a 50/50 resolution, prompting strong pushback from users. mebeim described the situation as "absolute insanity."

looppah raised the procedural question of whether the market will resolve based on the PM's clarification or if UMA voters can override it. mebeim clarified that resolution will follow the clarification, noting that UMA has only overruled a clarification on one or two occasions where it was "blatantly wrong." nopsuke suggested the 50/50 outcome reflects the unpredictable nature of clarifications rather than common sense.

POLYMARKET DIGEST — 22 SEPTEMBER 2026

⚠️ Bugs & Issues

A major and prolonged CLOB latency degradation has affected Polymarket for over 12+ hours across Sept 21–22. Users reported order placement times of 400–5000ms avg, p95 up to 13,000ms, and p99 reaching 30 seconds at peak. Multiple users also reported 500 Internal Server Error: context deadline exceeded, 425 Too Early errors, failed limit orders, and cancel timeouts exceeding 5 seconds. The GET /positions (v2) endpoint was also falsely returning 429 errors for nearly 24 hours. skulldoom97 acknowledged the degradation and confirmed no CLOB restart is planned; suhailkakar stated a fix was being deployed. Separately, fill notifications on the user WebSocket stream were also delayed, which suhailkakar confirmed the team is aware of.

Additional UI bugs reported: the Profit/Loss graph shows incorrect data for 1Y, YTD, and All-time views; the Favourites tab fails to display; and a general "Something went wrong" error blocks withdrawals and trades due to a failing relayer/session signing mechanism. exo_pm noted a today/tomorrow average price sort is being added, while alphabetical sort is not yet restored — the Search bar with a smoother API query is suggested as a workaround. A user also reported that the WebSocket server closes the entire connection when a subscribed market resolves, rather than just dropping that market's subscription.

⚡ Disputes

"Rebate Gate" — Users including sonderwrld, lochnload2718, and blunkoed continued pressing Polymarket over previously undisclosed 80%+ fee rebates given to certain participants. suhailkakar acknowledged the issue and promised an update "hopefully by this week," but users noted the delay has already stretched 8–10 weeks. blunkoed raised a broader trust concern: even if Polymarket claims the experiment stopped, there is no way to verify similar practices are not ongoing.

Market manipulation in 5-minute crypto markets — Multiple users reported ongoing wash trading and price manipulation in BTC 5-minute markets. sonderwrld alleged accounts honey-spot, potabmw, and stagbeetle were coordinating, with one manipulator reportedly making ~$40K in a single day. suhailkakar confirmed honey-spot was banned and stated the team is analyzing trading patterns for re-detection if they return on new accounts. He also confirmed the TWAP formula is controlled entirely by Chainlink and cannot be shared with Polymarket. Community members argued that account banning is largely ineffective without KYC, as manipulators can simply create new wallets. sonderwrld suggested blocking suspected accounts from converting pUSD as a harder deterrent.

📋 Clarifications

A recurring issue has been identified in weather markets, specifically for Tel Aviv and Manila (RPLL). User sanhwy documented that NOAA's WRH data pipeline experienced gaps on both September 20 and 21, with confirmed missing METAR observations during critical resolution windows. NOAA confirmed by email that "data appeared available on equipment but did not reach Synoptic — a communication error likely occurred." sanhwy noted that Polymarket's Aug 30 rule change narrowed the Weather Underground fallback to only trigger when NOAA data for the entire date is unavailable, creating a loophole where partial gaps — affecting specific observation hours — determine outcomes without triggering the fallback. sanhwy flagged that newly active wallets took large positions during these data-gap windows and called for an on-chain timing review, though no manipulation has been proven. The current market rules, as banisters noted, do not allow for resolution changes based on these gaps.

UMA DIGEST — 22 SEPTEMBER 2026

⚠️ Bugs & Issues

Tennis market rewards have been lowered due to the same bug previously seen in table tennis. Several users noted this may not have been communicated clearly enough, as proposers were still rushing to submit immediately. Discussion suggests that at the reduced reward (~$0.8 vs. prior ~$3.50), instant proposals are borderline unprofitable after gas costs.

📋 Clarifications

Trump "Alaska" market (Sep 14–20): b3gott3n argued there was no proof the video was filmed within the market's timeframe and requested a dispute window extension. The market resolved Yes before action could be taken. aenews stated the video was pre-recorded on the 16th, not during the market window. nefffooor countered that solid proof of a date existed. abhiz69 noted it likely wasn't a misresolve given a recently added rule clause specifying that content recorded during the contract period and released within the timeframe qualifies. mebeim acknowledged the clause exists but called it a poor rule for situations like this.

Russian parliamentary election ("which party gains most seats"): The market was proposed before full vote counts were in (~90% counted at proposal time). aenews, who had a large position (~$500k+) in the market, noted the ambiguity between "gains most seats" vs. "wins most seats" and that proposers could have avoided dispute risk by waiting ~40 more minutes for official CEC results. The market ultimately appears to have resolved without a dispute.

Finkelstein vs. Gabelic Match O/U 21.5: A proposer initially submitted all three game total lines (21.5, 22.5, 23.5) as 50/50 due to a walkover. The 22.5 and 23.5 proposals were disputed and reproposed as unders. joes49572 noted the 21.5 line is genuinely ambiguous (max possible games = 22–23 depending on tiebreak), while 22.5 and 23.5 unders are mathematically determinable.

⚡ Disputes

Tennis resolution source debate: An ongoing argument continues over whether ITF's official site should be the required resolution source. joes49572 argues ITF's site lags 30–60 seconds behind other data sources, making it unsuitable for timely proposals, while still acceptable as a verification source. woodmark520 and others pushed back, noting that third-party sites derive data from ITF anyway. No rule change was announced.

17 PM local time via DSM report and confirmed in the final CLI report issued the following morning. A preliminary CLI showing 71°F earlier in the day was not the final resolution figure. hans32312 confirmed Polymarket US resolves based on CLI reports.


⚡ Disputes

The RPLL (Manila) Sep 20 highest temperature market is the focal active dispute. NOAA's WRH viewer skipped the 04:00Z–07:00Z window; the 33°C daily high occurred at 04:00Z and is absent from the viewer. PAGASA official data and Weather Underground both record 33°C. NOAA support confirmed the observation exists in other NWS systems and in the 04Z METAR cycle archive, and said it may backfill the viewer. orcanomics noted the dispute timer was extended (which hans32312 clarified happens to almost all markets and does not indicate a formal dispute was filed). No official Polymarket team resolution or clarification has been posted as of the last messages.

POLYMARKET DIGEST — 21 SEPTEMBER 2026

🔥 News & Announcements

A significant NOAA/weather.gov data outage on Sep 20 caused multi-hour gaps in temperature data for numerous international stations (RPLL, ZBAA, ZGGG, LEMD, LTAC, EPWA, ZSQD, WSSS and others), with the daily temperature extreme falling inside the gap for several of them. The outage appears to be a systemic ingestion failure in the WRH Time Series Viewer rather than a total website outage — NOAA/NWS confirmed the missing observations exist in other official NWS systems and in the 04Z METAR cycle archive. A user flagged wallet 0x13f995ad154da3e078ac2a1dd923bf2b265dd352 for unusual trading activity timed around when the data gap became publicly visible; no wrongdoing has been proven. sanhwy noted that Burwick Law is publicly investigating potential claims from Polymarket traders involving disputed market resolutions.

sanhwy called on Polymarket to restore the older, clearer rule wording: "If NOAA data is unavailable or missing for the observation period, Weather Underground will be used as the secondary resolution source" — arguing the current wording does not cover partial data gaps. As of early Sep 21, the same METAR feed outage appears to be recurring on weather.gov.

A separate incident: a bot trader (gs7422) reported a loss after Polymarket's own embedded native feed prematurely marked the Vitória SC vs. Moreirense FC match as Full-Time with a stale 0-1 score, before a late equalizer made it 1-1. The bot bought "No draw" based on that feed and lost. phoenyx69 pointed out the feed's own disclaimer states it "should not be used as the basis for any trading decision," and gs7422 acknowledged he had not seen that disclaimer.


⚠️ Bugs & Issues

Multiple technical problems were reported across Sep 20–21:

• PnL charts: eljefe_30457 reported charts only updating every ~8 hours instead of hourly, with incorrect historical data.
• CLOB "not enough balance" bug: honeybear0x reported that the CLOB treats already-settled and redeemed positions as still matched, blocking use of the full wallet balance for new orders. GitHub issues filed at py-clob-client-v2 #112 and py-clob-client #342. No fix confirmed yet.
• Order maker queue full: berth6833 reported hitting the order maker queue capacity during simultaneous NFL markets, noting that BUY and SELL appear to have separate queues ("order queue full for maker <wallet>-BUY").
• API latency and errors: multiple users (bzhksrg, toto6328, superair233, satatardy, chrisv3537) reported intermittent "Unauthorized/Invalid api key" errors, doubled latency, 429 rate-limit errors on GET positions, and frequent request exceptions.
• Limit orders not executing: adrya407 reported limit orders from the UI neither filling nor appearing as resting orders.
• Cloudflare WAF issue: cylex_2 reported a WAF block also affecting the Gamma API, likely misreading 0x... IDs as SQL-style hex literals. exo_pm (team) confirmed they will check the WAF rule.
• LP rewards not showing: audunlkleveland reported LP rewards not displaying.
• Portfolio sorting removed: nicoco_89 requested the return of alphabetical and "resolving soonest" sorting in portfolios.


📋 Clarifications

The core unresolved question for Sep 20 weather markets is whether a partial NOAA/WRH data gap — where some rows exist but the daily extreme falls in the missing interval — triggers the Weather Underground fallback rule. The current rule only switches to WU if all NOAA data for the observation date is unavailable by 11:59 PM ET the following day. modali1 and lightningfella consistently argued that as long as any NOAA data point exists for the day, NOAA data is used exclusively, citing prior team clarifications. sanhwy countered that a known ingestion failure in the WRH viewer — where the observation already existed in NOAA's own systems — should not be treated as the authoritative NOAA reading for that interval, and requested an official Polymarket team response before resolution.

hermie_28217_08100 clarified the Chicago (KMDW) Polymarket US market: the 72°F high was recorded at 5:

UMA DIGEST — 21 SEPTEMBER 2026

⚠️ Bugs & Issues

An earlier issue with 15-minute liveness proposals being delayed (effectively treated as 2-hour liveness) was reported by im.the.wizard. The problem appeared to resolve itself by midday.

⚡ Disputes

Berlin State Election (Polymarket) — A proposal resolving the Berlin state election winner market was disputed as premature. At the time of proposal, only exit polls were available — no vote counting and no official media reporting had taken place. raskell1 and armanh01 disputed the proposal, with the consensus being a clear p4 (too early). The proposer lost their bond (~$6,000), with disputants sharing the rewards. im.the.wizard noted that "results don't need to be 100% official but there must be sufficient counting," which was not the case here.

Mecklenburg-Vorpommern Parliamentary Election (Polymarket) — raskell1 flagged a similar market where preliminary results have been published but are not yet final, asking how UMA typically treats preliminary results. viniipaquete suggested the same proposer may attempt further seat-placement markets and warned those would also be disputable.

📋 Clarifications

On election markets generally: exit polls alone are not sufficient for resolution. There must be meaningful official vote counting before a proposal can stand.

POLYMARKET DIGEST — 20 SEPTEMBER 2026

⚠️ Bugs & Issues

Polymarket suffered a major multi-hour outage on September 19. The CLOB went down around 16:40 UTC, with widespread reports of orders not placing, not cancelling, and not appearing in the book. skulldoom97 (staff) put the platform into cancel-only mode at ~17:58 UTC and promised a 5-minute heads-up before re-enabling. Trading was restored at ~18:55 UTC per pmkeaton, but a second outage immediately followed, attributed by pmkeaton to a Polygon network incident (https://status.polygon.technology/incidents/01M2XMCZJ4EXFC5Q2AH7XQ0NDD). Trading resumed again around 20:10 UTC. tsmultra and iflexson (staff) confirmed the team was aware throughout.

pmkeaton and suhailkakar confirmed refunds will be issued for losses incurred during the outage. A user who bought before the outage and could not sell during it asked whether their market losses qualify — the answer was implicitly yes, per pmkeaton's statement.

Additional ongoing issues reported:
— PnL chart showing incorrect data with fake peaks/dips (multiple users, still unresolved)
— Positions/stats data stale; 0xdanzu confirmed stats are being normalized as a new data API rolls out
— Rewards page not loading fully (tuesday3329)
— Order fills appearing on-chain with ~20-minute delay during the incident (berth6833)
— Throttled taker orders filling into already-closed windows (mmmmmmmmmmmmmmmmmm_)
— API book endpoint (bids/asks sort order) not reliably sorted despite documentation claiming best-price-first; fargason_ advised not to rely on guaranteed ordering
— Obsidian tier status reportedly being granted to accounts with only ~200k weighted volume, far below the stated 10M threshold (votobias_12985, phoenyx69)

Community members and verrylowsharpe repeated a longstanding request for automatic order cancellation when the exchange goes down, noting that 5-hour maintenance notices do not account for all timezones. brka also called for better in-platform banners and notifications during outages, tagging the team directly.

Users noted the outage coincided with the 0-fee sports weekend (college football, soccer), with speculation that the volume spike caused the CLOB overload. shy_blue_rock mentioned the matching engine is being rewritten; ravdes21 linked a post from dev Josh Stevens apparently referencing the rewrite.

The LEC Summer Split 2026 MVP market resolved as "Other" despite Caliste winning the award — 0xdanzu confirmed this was a deployment mistake (wrong end date set) and that the market team will issue refunds.

🔥 News & Announcements

0-fee trading is active this weekend on sports markets (college football, NFL tomorrow, soccer). The Russia military action market was flagged by viniipaquete as potentially missing its no-fee flag.

⚡ Disputes

An ongoing "rebate gate" controversy continues to dominate the devs channel. Multiple users (lochnload2718, lebrawnjaymez, blunkoed, vilius.u) have been repeatedly posting disclaimers warning that select market makers have reportedly received rebates as high as 80%, with no public disclosure of eligibility criteria. iflexson asked them to stop, describing it as spam; users pushed back saying it is not spam. pmkeaton has not publicly addressed the substance of the rebate claims. A WSJ article linked by blunkoed (https://www.wsj.com/business/polymarkets-fraud-regulators-coplan-5f418ab0) appears related to regulatory/fraud scrutiny of Polymarket. A separate concern was raised about KYC-verified colocation in London receiving worse latency than non-KYC servers in Amsterdam (avataroflol).

UMA DIGEST — 20 SEPTEMBER 2026

⚡ Disputes

CFB market (Flint vs FLATL 1Q Total 7.5) was flagged over whether a field goal occurred in Q1 or Q2. alexchunus and others argued the scoreboard showed 10 points in Q1, while 830lincoln countered that ESPN's play-by-play misattributed a Q2 FG to Q1, citing school tweets. dyanreel initially cited NCAA play-by-play as confirming a Q1 FG, but later retracted after reviewing game video, confirming it was a Q2 goal. No dispute was ultimately filed — the matter appears dropped.

⚠️ Bugs & Issues

im.the.wizard reported hundreds of proposals stuck unsettled for 3+ hours, including 15-minute liveness ones, while other proposers like Flame had fewer than 20 pending. cricket_666 confirmed experiencing the same delay on 15-min liveness proposals, suggesting a platform-wide settlement glitch.

POLYMARKET DIGEST — 19 SEPTEMBER 2026

🔥 News & Announcements

ZEC markets (5m/15m) are intentionally hidden from UI — used for TWAP testing only; will be re-announced when re-enabled (confirmed by staff 0xdanzu).

Session Key API (POST /v1/session-signers/authorizations) returns 403 for valid Builder API keys — session keys are in limited beta, access is restricted (confirmed by staff tsmultra).

⚠️ Bugs & Issues

CLOB performance (critical, ongoing): Cancel/order latency severely degraded — cancel ACK delays up to 26 000ms, normal order RTT jumped from 50ms → 200ms+ even on colo. ~40% of fills were cancel-then-filled-anyway. Staff (skulldoom97, 0xdanzu) acknowledged; fix requires a deploy. Issue worsening for 21+ hours.

WebSocket disconnects (SERVER_CLOSE code=1013, "slow consumer"): PM sends event bursts faster than clients can consume → buffer fills → disconnect. Workaround: increase RCVBUF, use internal queue.

Cloudflare WAF / data-api v2: Specific combinations of condition IDs in GET /v2/resolutions?condition= return HTTP 403. Deterministic by content, not rate/IP. Forwarded to team.

Login bug: Email + TOTP flow returns "Request Cancelled" error.

Fee UI broken on multiple markets.

Limit orders with decimal prices require page reload to place; error appears otherwise.

PnL stats showing inconsistent values day-to-day with zero trades executed.

Position tab not accurately reflecting shares held/sold.

Leaderboard numbers reported as incorrect.

Page loading failures and excessive load times reported by multiple users.

Docs broken link: docs.polymarket.com/concepts/order-lifecycle returns 404.

Support response times averaging 1–2 months.

📋 Clarifications

Real Cartagena FC vs. Bogota FC (originally 2026-08-11): ~40 markets redirected to event col2-car-bog-2026-08-20 for resolution.

FK Auda Riga vs. Ogre United (originally 2026-08-03): ~50 markets redirected to event lva1-aud-ogr-2026-09-16 for resolution.

⚡ Disputes

"Rebate Gate" — select users reportedly received 80% taker fee rebates ~4 weeks ago (around Aug 19), creating structural disadvantage for others. User singlur reports taker fill rate dropped from 80% to ~zero overnight on the same date. Mod tsmultra threatened timeouts for users spreading this message, which itself drew community criticism.

XAU/USD resolution dispute: User shariful1000 disputes "Will XAU/USD dip to 4250 by Sep 14" resolution — same price level triggered YES on one market but NO on another; argues 1-minute candle low below 4250 should count as valid YES.

Platform health warning circulating in community, citing: private incentives to select participants, international Polymarket version being quietly sunset in favor of US version, CLOB unlikely to be fixed before next year, and Polymarket losing market share to Kalshi.

UMA DIGEST — 19 SEPTEMBER 2026

⚠️ Bugs & Issues

A bot is incorrectly proposing MLB overs on markets where no challenge occurred and the run was never on the board. alexchunus identified the root cause: inverted logic — the bot checks 7.5 > 7 instead of 7 > 7.5, causing proposals on 7.5 and 8.5 totals where actual scores were 7 and 8. 830lincoln noted a bad data feed may also be a factor. Compared to a prior incident where geowiz code confused Kentucky and Western Kentucky.

⚡ Disputes

Ongoing disagreement over a neg-risk market resolution. abhiz69 argues the neg-risk excuse is invalid — every bracket can resolve N, and cites multiple examples. nefffooor counters that Infinity voting P2 seems wrong and the market should remain P4 until clarified, suggesting Poly could resolve to a low-priced bracket to minimize refunds. abhiz69 suspects Poly would only clarify to N if it resulted in a smaller refund obligation.

POLYMARKET DIGEST — 18 SEPTEMBER 2026

⚠️ Bugs & Issues

Redeem broken — multiple users reported inability to redeem winning positions for 12+ hours. ak504111 confirmed it was being investigated; resolved around 11:20 UTC.

Data API v2 — redeem endpoint not yet implemented in the new API. 0xdanzu confirmed it's being reintroduced, ETA unknown.

LP rewards returning 0% — adam22234 reported earning 0 rewards across multiple markets despite meeting criteria. Unresolved.

CLOB DeadlineExceeded errors on FAK orders (context deadline exceeded 5552ms) — reported by miraacook, confirmed as an intermittent known issue by berth6833.

Stale CLOB order book data — sharkhunterr reported certain markets showing stale odds via CLOB while the UI updates normally.

Referral links — imstillrich reported referral links showing as "expired" despite being used successfully for signups, with no bonus credited.

🔥 News & Announcements

Zero fees on NFL & College Football this weekend (Sept 19–20). sharkhunterr confirmed the promotion. hipp74 noted taker fees still appear on third-party sites — unclear if it's a rebate or a direct fee waiver not yet reflected in the Gamma API.

Taker delay on crypto markets reduced to 150ms (changelog stated 50ms, but blunkoed corrected it to 150ms), down from 250ms. Change took effect August 17.

Builder channel restructured — displaysonx announced builder discussion is consolidating into a new channel; old channel preserved as read-only archive.

Polymarket CEO shayne_coplan posted on X acknowledging degraded onchain product performance, stating the team is "working behind the scenes to unify the products." Community reaction was skeptical — berth6833: "after 6 months he finally heard about the frustration."

⚡ Disputes

80% rebate controversy ongoing. Multiple dev-channel regulars (lochnload2718, blunkoed, tehnoobkiddo) are appending disclaimers to every message, warning that select market makers reportedly receive up to 80% fee rebates with no public disclosure of eligibility. richlb_predict noted the 80% rebate was reportedly paused; blunkoed questioned whether it was truly paused or simply moved off-chain. lochnload2718 warned that if community criticism is suppressed on Discord, they will escalate to Twitter.

Co-location KYC — blunkoed clarified that even after KYC approval, co-location doesn't provide meaningful latency improvement because orders still route through Cloudflare in a separate DC, with an additional wallet check adding latency. Dublin reportedly performs comparably or better. KYC approval times range from 1 week to 6 months.

UMA DIGEST — 18 SEPTEMBER 2026

⚡ Disputes

NBA Free Agency markets (DeMar DeRozan & Rui Hachimura) — Two Polymarket NBA free agency markets are under dispute while in archived/pending-clarification status. The core issue: both players have joined new teams, but the market structure creates ambiguity about which team is the correct resolution. laurleo notes it "could genuinely go either way" depending on whether title or description string takes precedence. abhiz69 and armanh01 criticized proposing archived markets as inviting adverse selection. Polymarket team apparently didn't extend the dispute window, which abhiz69 called "shocking." alexchunus raised a broader concern — a precedent from a prior "negrisk" market where an early wrong proposal was voted P2 (instead of P4 pending clarification), effectively allowing a misresolve to slip through. nefffooor echoed frustration: "Just let the wrong bracket slip through on negrisk and then you can do whatever you want and UMA will support it." mebeim noted UMA's hands are tied if Polymarket doesn't clarify in time.

ITF tennis set total market — mrchowchowdamn flagged a potential misresolve on an ITF match (Lizuno vs Jansse, Sep 14): Set 1 O/U 9.5 was proposed Under, but the set ended 6-4 (10 games total = Over). laurleo and fernando5495 confirmed it likely misresolved and could have been disputed, with laurleo noting "6-4 is obviously over 9.5."

POLYMARKET DIGEST — 17 SEPTEMBER 2026

🔥 News & Announcements

Pyth feed deprecations (Sept 16, 8 PM ET) — three commodity feeds retired:
• WTI Crude Oil (WTIV6/USD, WTIX6/USD): ~30+ September & week-of-Sept-14 markets migrated to CME Group official daily high/low.
• Natural Gas (NGDV6/USD): ~12 weekly NG markets (week of Sept 14) similarly migrated to CME Group NG settlements.

⚠️ Bugs & Issues

• Positions not updating (4+ days): filled orders stuck in UI, not appearing in positions panel. No team response.
• Portfolio value flickering: cash & portfolio balance swinging 40–50%, not reflecting real positions.
• Sort by price missing from positions panel — feature removed or broken, impacts rebalancing.
• Leaderboard broken: monthly weather leaderboard leader has zero weather trades.
• Vance rate stale: UI still shows 3.25% despite Fed hike to 3.50%.
• KYC delay: at least one user waiting 1 month for approval.
• Scam alert: fake "Polymod" account phishing users via DM — confirmed impersonator, not staff.

📋 Clarifications

• El Niño markets (7 markets): "Regional Oceanic Index" corrected to Relative Oceanic Index (RONI) across all 2026–27 El Niño markets.
• Moscow air traffic market: resolution criteria tightened — "temporary restrictions" qualify as suspension; airports "accepting/dispatching flights in agreement with authorities" do not independently qualify.

⚡ Disputes

"Rebate gate" — ongoing (~24+ days, no resolution): Polymarket secretly granted 80% taker fee rebates to select traders. Staff (keaton, Suhail) previously denied preferential treatment publicly. Discovered on-chain by community. pmkeaton now admits no proper response given but calls it "not my area of expertise"; Suhail deflecting. Community alleging potential CFTC/MiFID II violations. Users threatening to exit; retail sentiment deteriorating.

LP reward bots flagged: wallets 0x7f9c6ce5... and 0x581ee4cf... suspected of cancel-after-fill exploitation — fixed timestamps, tiny fills, rewards withdrawn ~10 min post-distribution. Taker delay variable also raised from 50 → 150 over past 3 weeks, possibly related.

POLYMARKET DIGEST — 16 SEPTEMBER 2026

⚠️ Bugs & Issues

Multiple ongoing platform problems reported:
— Archived/deleted markets appearing in Active Portfolio with ghost values; toggle to hide them broken (resets itself)
— Open orders not displaying correctly: placed orders invisible, cancelled orders persist
— Apple Pay/card deposits not working
— Profile trading volume mismatches leaderboard stats
— API batch order endpoint timing out; CLOB API blocking Gnosis Safe addresses with "maker address not allowed" error
— WebSocket price_change feed dropping cancellation deltas silently, causing phantom orderbook levels — no sequence number exists to detect gaps client-side. berth6833 confirmed this is a known issue and has requested seq_id support in the next engine
— Order latency spikes reported: p99 hitting 1200ms on 5m crypto markets during volatile moves; sonderwrld shared stats showing POST p90 at 102ms, CANCEL p90 at 68ms
— RTDS streams dropping out regularly

📋 Clarifications

US–Iran ceasefire markets (Sep 4, 7, 11, 18, 30): Rules clarified — pre-creation military actions are not qualifying events; the 14-day ceasefire window starts the day after market creation if no post-creation action has occurred.

Bab el-Mandeb & Strait of Hormuz transit markets: Resolution source formally added as IMF PortWatch transit calls data; unrounded values take precedence over rounded ones.

AI Chatbot Arena 1520 score markets (Sep 30 & Dec 31): Polymarket announced order books were cleared at 7:30 PM ET Sep 15, then issued a resolution clarification: on Sep 2, claude-fable-5.1-max displayed a score of 1523, meeting the 1520 threshold. Rules impose no minimum vote count and do not disqualify scores later revised — markets resolving YES.

⚡ Disputes

Rebate Gate remains the dominant ongoing conflict. Community members (sonderwrld, vader17, kryme69, vilius.u, lochnload2718 and others) are demanding a public post-mortem on secret 80% maker rebate deals given to select accounts — described as tens of thousands of dollars in unfairly distributed advantages. suhailkakar (Polymarket) engaged briefly, stating "a bunch of things are getting finalized" on rebates, acknowledged the taker delay increase as a response to feedback, but dismissed critics as "3–5 people." Community rejected this characterization. vilius.u urged a full honest post-mortem. lochnload2718 called for taking the campaign to Twitter under #POLYREBATEGATE. suhailkakar then went offline, further inflaming the channel.

pinner flagged two addresses (0x7f9c6ce5785b87f31b5e397c8985d7d894829236, 0x581ee4cf297c2a057aed53d928f60f1adf981526) as suspected cancel-after-fill bots generating programmatic fills at fixed timestamps with tiny notionals to avoid detection, withdrawing LP rewards 10 minutes after distribution.

Community also pressing for a combined US/international order book — kryme69 raised it directly; suhailkakar said it is not current priority (focus is DeFi web experience & trading performance). A tweet from @caronpolymarket hinted it may be coming, but no official confirmation.