A Michigan election trader bought and sold 128,683 shares in 53 minutes
The Michigan round trip cost 4,177 pUSD before Grant Cardone alleged political manipulation. An Election Day follow-up examines large trades in Michigan and Texas.
By MSR Decode · Published November 2, 2026 · Updated November 3, 2026 · 6 min read
One Polymarket account spent 106,044.53 pUSD buying shares that would pay out if a Democrat won Michigan’s Senate race, then sold almost all of them within 53 minutes of its first purchase. It recovered 101,867.49 pUSD after fees, leaving a 4,177.03 pUSD shortfall on the seven trades. pUSD is the token used to fund these trades.[1]
The account finished selling more than an hour before Grant Cardone posted his allegation on X that Democrats were using political funds to manipulate Polymarket odds. MSR Decode traced the Michigan purchases and sales to examine what happened before that claim.[2]
What Cardone said on X
“I am convinced Democrats are using political funds to manipulate Polymarket odds.”
His Michigan image showed Abdul El-Sayed at 79% and Mike Rogers at 22%. These percentages represented market prices read as chances of winning, not predicted vote shares. Cardone’s post did not identify the account examined here.[2]
Buying and selling the same position
The account began with no shares backing a Democratic victory in Michigan. It made five purchases over about 14 minutes, acquiring roughly 128,683 shares, then sold almost all of them in two transactions. Less than one-thousandth of a share remained.[1]
The size and speed of the reversal stand out. The account committed more than 106,000 pUSD, including fees, but had sold almost the entire position within an hour. Most of the money came back through the sales, with a loss on the round trip.
Michigan · Seven transactions
A 4,177.03 pUSD cash shortfall
The account recovered most of its spending after reversing the position.
Together, the purchases and sales counted as about 207,975 pUSD of trading before fees. That total counts both directions of the same round trip. It does not mean that 207,975 pUSD of new money entered the market or stayed behind a Democratic victory.[1]
How large trades can change the odds
The largest purchase bought about 14,119 shares at 81 cents and another 105,799 at 82 cents. The archived order book—the list of available buy and sell orders—showed the 81-cent sell offers disappearing and the 82-cent offers falling by the matching amount, allowing for rounding.[3]
When the account sold its remaining shares, it accepted buying offers at 80, 79 and 78 cents. The first recorded reductions in those offers matched the quantities sold. All seven transactions also matched trade records in the archive.[3]
Around the largest purchase, the lowest available selling price rose from 81 to 82 cents. Around the final sale, the highest buying price fell from 80 to 79 cents. These comparisons cover roughly a minute around each trade message; the supporting records give the exact timing.[3]
This shows how a large order can use up offers at one price and trade at the next. The odds people see can change without new election information. If a trader deliberately used that effect to promote a misleading impression of a race, the market could become a tool for influencing perceptions.
Where the money came from
During the Michigan trading, the account received another 1,000 pUSD and then 98,999 pUSD. The larger payment arrived just 21 seconds before its largest purchase. We traced those payments through an account that converted incoming tokens into pUSD, but the trail did not identify the original funder.[4]
Election Day follow-up: another large Michigan position
On November 3 at 2:08 a.m. Eastern, a different account bought about 86,197 shares backing a Democratic Senate victory for 70,350.13 pUSD, including fees. The price before fees was 81 cents per share. At 6:54 a.m. Eastern, it sold about 39,963 shares at 80 cents and received 31,714.42 pUSD after fees. Just 98 seconds later, at 6:56 a.m. Eastern, it bought another 1,790 shares for 1,461.02 pUSD, including fees, at 81 cents before fees.[6]
The account had also bought about 6,329 shares for 5,042.00 pUSD, including fees, on November 2 at 9:42 p.m. Eastern. It retained about 54,353 shares after these four trades, so it had only partly reversed its position.[6]
The blockchain records confirm the trades, and the archive matches the three November 3 transactions. At the sampled points around those trades, the best buying and selling offers remained 80 and 81 cents. Unlike the original round trip, this comparison did not show a change in the best available prices. A large trade can leave those prices unchanged when enough orders remain available.[6]
The original account examined above had no new activity in the public feed between November 2 at 1 p.m. and November 3 at 1:09 p.m. Eastern. The new account’s trading is a separate finding; it does not extend the original account’s 4,177.03 pUSD loss calculation.[6]
Texas follow-up: large bets in both directions
Cardone’s same post also questioned the Texas Senate market. The November 3 follow-up found large purchases backing both a Republican victory and a Republican defeat.[7]
At 12:23 p.m. Eastern on November 3, another account bought about 313,470 NO shares in the Republican-win contract for 231,304.32 pUSD, including fees. These shares pay out if a Republican does not win. The order remained available at 73 cents and received smaller fills afterward.[7]
A different account backed a Republican victory. At 7:44 p.m. Eastern on November 2, it bought about 343,699 YES shares for 99,845.57 pUSD, including fees. That order filled offers at 27, 28 and 29 cents. Archived quotes rose around the purchases, showing how a large order can consume offers at different prices.[7]
Gross purchases: 105,278.659997 pUSD; buy fees: 765.865400. Gross sales: 102,696.643800 pUSD; sale fees: 829.149280. Spending 106,044.525397 minus receipts 101,867.494520 = 4,177.030877 pUSD. Gross turnover sums purchases and sales before fees. The position went from zero to 0.000742 YES share. The cash calculation excludes gas, other assets and markets, outside hedges and payments. Native scans cover 09:00–14:00 UTC. This is a bounded reconstruction, not a lifetime account result.
Research scope: the election was unresolved at the November 2, 2026, 18:00 UTC cutoff. No responses were sought for this case file. The first purchase was at 11:34:24 UTC and the final sale at 12:27:04 UTC, an interval of 52 minutes and 40 seconds. Five buys acquired 128,682.920742 shares and two sales disposed of 128,682.92. Gross turnover was 207,975.303797 pUSD. pUSD amounts are native collateral units, not independently checked dollar redemption amounts. The record establishes one near-complete round trip; it does not establish repeated cycles, self-trading or coordinated trading between accounts under common control.
2. Original post and Michigan market
Original X.com post by Grant Cardone; Michigan Senate election market. Direct X retrieval during the original research returned HTTP 403. A retained FxTwitter response supplies the quoted statement and attributes the post time to November 2 at 13:32:47 UTC. The post’s Michigan image showed 79% and 22%; the capture time is unknown. Original image URLs and hashes were retained, but the image files were not recovered for this case file. The original investigation examined Michigan. The dated November 3 follow-up also examines the Texas Senate markets in the same post. The two party contracts allow other-party outcomes and are not exact complements.
Post extract and source-response hash. The private research retains the complete mirror response; the excerpt is not an independent authentication of X’s record.
3. Execution levels and archived orders
Pendulum Flow v3.1 archive; analysis and method boundaries; matched millisecond records; replay results. Largest buy: 14,119.39 shares at 81 cents and 105,799.419633 at 82 cents. The 82-cent reduction differs by 0.000367 share, consistent with displayed two-decimal depth. Final sale: 30,455.67 shares at 80 cents, 36,485.28 at 79 cents and 16,741.97 at 78 cents. First reductions match those three quantities exactly.
Quote comparisons use one second before and 60 seconds after trade-message receipt. Depth updates do not carry transaction hashes. Block timestamps, venue timestamps and collector receipt times differ; the book evidence establishes correspondence, not a causal estimate of the entire price move. Completeness, cancellations and other traders’ activity limit interpretation. Supporting analysis files retain their original research context; not every referenced private workpaper is included in this selected evidence set.
4. Funding and technical control
Final funding and control assessment. Historical owner at five sampled block endpoints: 0x9bdf52642943d9b5b665de67c39c50c31d1e1dd2. The immediate funding address was neither owner nor an authorized session signer at those endpoints. This does not identify the seven order signers or prove different beneficial controllers. The upstream Relay label is an explorer attribution, not independently verified operator or customer identification. The previously unverified implementation-name attribution for the conversion account remains withdrawn.
5. Evidence selection and corrections
Selected source-file provenance and SHA-256 hashes. The original Michigan-only edition derived from the November 2 combined report. The November 3 follow-up adds separate Michigan and Texas observations. The complete original report and research remain retained privately. The earlier 43.9% Michigan concentration figure applied only to a large-trade filter; the broader returned-record calculation was 35.48%. Neither measures fresh funding or certified whole-market volume. No concentration claim is needed to establish this seven-trade reversal.
All seven archived trade messages match the blockchain transactions by transaction identifier, outcome, direction and size. The quote comparisons run from one second before each trade-message receipt to 60 seconds afterward. Later changes exceeded the account’s trades. The funding investigation checked the contract-designated owner at five block endpoints and did not identify an originating customer deposit.
6. November 3 Michigan follow-up
Selected native records, historical balances and archived quote observations. Comparison account: 0x8bd3a7b9a1b58ebc3f243d2eb05327163ea3f69a. Native trades, in Eastern time: November 2 at 21:42:33, BUY 6,329.113923 YES for 4,999.999999 pUSD before 41.999990 fees; November 3 at 02:08:33, BUY 86,196.913578 for 69,819.499998 before 530.628190 fees; 06:54:31, SELL 39,962.730000 for 31,970.184000 before 255.761470 fees; and 06:56:09, BUY 1,790.123457 for 1,450.000000 before 11.020000 fees. Actual pUSD debits and credits agree with these amounts. The large purchase cost 70,350.128188 including fees; the sale returned 31,714.422530 after fees. A native balance check returned zero before the first purchase and 54,353.420958 YES after the fourth trade.
Quote comparisons use the last retained bid/ask observations at collector receive-time cuts one minute and one second before, and one second, one minute and five minutes after each matched November 3 message. Every selected cut returned 0.80/0.81. These are sampled records, not continuous executable quotes or proof of no transient price effect. Native block, venue and collector clocks differ; archive continuity and complete mirror validation remain unverified.
The original-account follow-up queried the public activity index from November 2, 18:00 UTC through November 3, 18:09:20 UTC. The response was terminal and empty; retained response. This is not an all-source inactivity finding. The original seven-trade calculation remains unchanged.
The Republican-YES buyer was 0x60bf706d142f4620e28a2701cfafb5565a85942d. Two transactions at November 3, 00:44:01 UTC (November 2, 19:44:01 Eastern) share its signed order hash and acquired 343,698.626552 YES shares. Gross consideration was 97,059.750000 pUSD; fees were 2,785.818750; actual payment was 99,845.568750. Matched maker legs normalize to YES prices of 0.27, 0.28 and 0.29. Counterparty order cash is not the buyer’s expenditure; the buyer’s own fill and transfer records establish its cost.
The archive matches both Republican-YES transaction hashes and quantities. Last observed YES bid/ask at the one-second-before trade-message-receipt cuts was 0.26/0.27; one and five minutes afterward it was 0.27/0.28. The earlier quote record was about five seconds old at the pre-message cut. These sampled observations include other activity and do not isolate the order’s full causal effect or prove deliberate distortion. The hour-17 archive was unavailable when checked, so no historical-book attribution is offered for the large Republican-NO order. Source file hashes match captured publisher manifests; this does not certify complete archive coverage.
The two party contracts allow other-party outcomes and are not exact complements. The selected purchases are distinct from overall positions, overall profit, new external funding or an election effect. All pUSD amounts retain the native collateral denomination.
Ask MSR to investigate
Tell us which wallet, market, payout or profit claim you want checked. MSR accepts commissioned investigations using the evidence standards shown in this report.
