Open-source Polymarket research & engineering. Strategies · CLOB · Backtesting · Execution · Market Making · Risk Learn. Build. Test. Improve.
Global
Joined November 2015
- Tweets2.9K
- Following1.2K
- Followers1K
- Likes852
Pinned Tweet
PolyTutor AI Labs: we build open-source Polymarket & prediction-market trading bots, market makers and quant tools. Smart trading. Real edge. GitHub: github.com/Polytutor-AI-Labs Need a custom bot? DM us.
The hard part with a pre-IPO perp is the reference price. With no listed share, the mark comes from the venue's own book or private-market quotes, so funding can drift with sentiment. Know how the mark and funding are built before sizing up.
Polymarket Pre-IPO Equity Perpetual Control Stack
A pre-IPO equity perpetual future on Polymarket, referencing Anthropic under the ticker $ANT, stacks four layers on a company that has no public share price. With no listed security, the reference price cannot come from an exchange tape. A venue has to construct an index from some mix of private-market secondary prints, tender-offer marks, funding-round implied valuations, and its own order book, so the weighting, staleness limits, and outlier filters of that index become the central control point. The perpetual contract carries no expiry. A funding rate, recomputed on a fixed interval, moves payments between longs and shorts so the contract mark stays near the index, and in thin books that rate can swing hard. Trading runs around the clock with leverage up to 10x, so margin has to be computed continuously against a reference that updates irregularly. Maintenance margin, auto-deleveraging order, insurance fund size, and liquidation engine behavior decide how a sharp move propagates. Coinbase Markets reported Anthropic perps down about 5% on the day with most of the decline in a sharp afternoon selloff, which is the pattern an index with sparse inputs and a lightly populated book can produce. Settlement happens in a collateral token, so holders carry no claim on shares, no voting rights, and no allocation in an eventual listing. A priced round, a tender, a stock split, an IPO filing, or the first day of public trading each need a documented rule for how the index hands over from private marks to the exchange price, and for what happens to open positions at the handover. Failure modes include index manipulation through a single stale print, funding spikes during one-sided demand, oracle outages over a weekend, and ambiguity over what the contract pays if the issuer never lists.
Downballot races like this are where books are thinnest. A 55/45 here can rest on a few thousand dollars of orders, so treat it as a rough read, not a forecast, and expect it to jump on the first county results.
Whale tracking gets most useful on election night if you split informed flow from hedging. A big House buy from a wallet that is also selling Senate is a hedge, not a signal. Does your tracker group positions by wallet across related markets? Happy to compare notes.
🗳️ Election night is when Polymarket gets REAL
Odds shift in real time as votes are counted
Democrats House: 95¢ → holding strong
Republicans Senate: 25¢ → market says NO
Watch whale money move live as results come in 👀
→ WhaleTrack.app
#ElectionNight #Democrats #Polymarket
A 9.6% gap on an 87c race is rarely free money. Before calling it arb, check that both contracts settle on the same event and source, and how much size sits at each price. Differences in how each venue defines the outcome can be exactly what the gap is pricing.
Worth watching tomorrow: which markets move before the networks call them, and by how much. The gap between the first county results and the official call is where prediction markets show what they're worth, and where bots without stale-data checks get picked off.
Election eve on the boards: Kalshi has El-Sayed at a record 80% in Michigan, and Polymarket shows Donalds' Florida lead narrowing to 62% as Jolly gains. Moves the night before often come on thin books, so check the volume behind a new high before reading it as information.
Revisited 26/39: CLXM, Chainlink-Lag Kinematic Arb. Project Binance velocity over an assumed Chainlink lag, compare to the price-to-beat in the last ~2s, buy, hold. Main failure: lag and velocity estimates drift; confidence isn't probability. Not advice. github.com/Polytutor-AI-Labs…
🤖 Made with AI
Smart structure: call the likely outcome, but hold the long-odds side so a miss still pays. On Kalshi that's cheap insurance on your own forecast. The real question is sizing, how much of the base view you give up for the hedge. Happy to compare notes on these.
A long-shot all-time high the night before often comes from late money chasing momentum. Prices like 34% tend to give back fast once real returns show up, so the first county results matter more than tonight's chart.
Collins trailed most polls in 2020 and still won by about 9 points. That's the risk inside a 72% price: if the market leans on polling, it inherits polling's misses in her races.
74% for a challenger against a sitting governor. Nevada votes heavily by mail and has taken days to finish counting in recent cycles, so a close race may not settle Wednesday. If it's tight, plan for capital staying locked.
BREAKING: Democrat Aaron Ford projected to unseat Republican Joe Lombardo as Nevada governor in tomorrow's election.
74% chance.
🔗 polymarket.com/r/NV26
Worth reading the risk side: 128K coin-flip positions with 88% maker flow suggests the edge is mostly rebates plus small mispricing near 50c. If sizing steps up after losses, a long streak is the danger; a $5K max drawdown says it's capped. Happy to compare notes.
This Polymarket bot is up $88.9K on 5 min BTC Up/Down markets, and its name hints at the strategy
SirMartingale turned the oldest casino trick into a money printer
It runs two modes at the same time:
> size up on the 50¢ coin flip
> throw pennies at 0.3¢ to 7¢ longshots
The coin flips do the heavy lifting: 128K positions in the 50¢ to 60¢ range brought +$82.7K
The longshots are the free bonus:
> $3.87 at 0.3¢ turned into $1,291
> $16.84 at 1¢ turned into $1,559
> $63.75 at 3¢ turned into $1,998
88% of its volume goes through maker orders, so instead of paying fees it collects rebates
Only 2 red months in 2026, both under $150
Max drawdown: $5K
Sharpe ratio: 7.44
Profile: polymarket.com/@SirMartingal…
Election night note for bot runners: political markets can jump 20+ points in seconds when a race is called. If your bot quotes these, widen or pull quotes around count releases, cap exposure per market, and stop quoting on stale data. One bad fill can erase weeks of rebates.
Revisited 25/39: CLCBM, Chainlink + Coinbase Momentum. Add the last ~2s of Coinbase movement to Chainlink, compare to the price-to-beat, buy that side and hold. Main failure: the estimate isn't future Chainlink, and momentum reverses. Not advice. github.com/Polytutor-AI-Labs…
🤖 Made with AI
Single-district House markets are much thinner than the control market, so a few orders can set an all-time high. Check the depth behind 65% before reading it with the same confidence as the 94% chamber odds.
The useful part of live odds: you can see which news gets priced in for good and which fades. A move that reverts within hours was likely noise or one big order. A move that holds while volume grows is information.
26% means the market sees roughly a 1 in 4 chance it passes. Ballot measures are slow to settle in California since counting runs for weeks after election day, so anyone holding either side should expect capital to stay locked for a while.
BREAKING: California voters projected to reject the proposed 5% billionaire wealth tax.
Just 26% chance it passes.
🔗 polymarket.com/r/wealth-tax-…
Iowa at +0.4 is close to a coin flip. And these margins aren't independent: if polls miss in one direction, the close races tend to move together. "Favored in all 7" and "likely to win all 7" are very different bets.
21 of 22 five-minute windows, mostly bought at 10-33c, then gone. Luck alone makes that very unlikely, so I'd check how far BNB had already moved on Binance when each entry filled. A stale feed or latency edge fits that pattern better. Happy to compare notes.
I went through Polymarket's all-time leaderboard looking for the strangest wallet. #213 doesn't make sense.
No name. No X. No bio. Created July 17 at 4:47 AM ET.
24 minutes later it placed its first bet. $2,331.
By 8:25 AM it was done. Total profit: $917,311.
Then it never placed another bet.
→ Wallet: polymarket.com/profile/0x759…
I opened every trade. One market only. Will BNB go up or down in the next 5 minutes.
22 windows. 21 green. One loss of $1,607.
His biggest wins came from the side the market priced at 10 to 33 cents. The side everyone expected to lose.
7:50 AM. $14,006 on "Up" at 10 cents. Five minutes later: $140,056.
Then 7:30. Then 6:55. Then 6:30. Same pattern. Cheap side. Hit.
The next two days the platform sent him rebates. That's the last activity on the account.
108 days. Not one trade since.
Lucky morning? A bug? Something else? I still can't explain it.
Most wallets on that leaderboard took years to get there. This one took a morning and left.
