Trading Polymarket: entries, exits and the numbers behind them. Building @PredyX_bot, a terminal for Polymarket traders. @zscdao
Joined March 2022
- Tweets2.3K
- Following259
- Followers318
- Likes2.7K
Pinned Tweet
Polymarket was 90.2% accurate a month before its markets resolved
Four hours out it hits 98.5%. Real money pricing real outcomes.
So I start every trade from the price and read the news second.
My entries and exits go up here, with the size and the result.
I think the market is still the best signal most ppl ignore.
Price first, headlines second.
30c to 60c on a recession-if-cut contract looks clean on paper
Conditional markets split already thin liquidity even further. If the Fed holds, the condition never triggers and your money just sat there.
Price the exit before the thesis.
prediction markets are still focused on the least interesting part of trading uncertainty.
it would be more interesting to trade what it means for the economy, so imagine the fed is expected to cut rates:
say the market gives a 30% chance of a recession if the fed cuts, but i think it's closer to 60% because the economy is weakening.
i buy that recession event contract for 30¢, and if more traders start agreeing with me, its price could rise toward 60¢. i can sell before the outcome is even known.
that's the opportunity in trading how events affect each other.
the real money is in the consequences, not the headlines.
5
165
Nov 11 vs Nov 12 cost dromedaryy $87K on one market
Nvidia's filing says the board accepted the resignation on Nov 11. The news broke Nov 12. The rules decide which date counts.
Kress sits at 15c while already interim CEO. Read the resolution text first.
JENSEN HUANG QUIT NVIDIA.
ONE TRADER MADE $125,745.
ANOTHER LOST $100,743.
Same market. Opposite sides.
dromedaryy bought YES on "out by Nov 12" at 49.7¢
433,033 shares → +$217,066
but also held YES on "out by Nov 11"
Jensen Huang left one day later
→ −$87,771
one day early cost $87K. still walked away +$125K
OxPredict went the other way
NO on Nov 12, NO on Nov 30, NO on Dec 31
sure Jensen Huang stays
the board said otherwise
→ −$100,743
nothing in between. one headline flipped both books
the next fight is already open: who becomes Nvidia CEO
Ian Buck 37¢, Colette Kress 15¢
polymarket.com/event/next-ce…
2
3
92
Kalshi founders face a Utah criminal court on 12/4
Arizona got a federal rescue. This judge already ruled Utah can enforce its laws against Kalshi, which makes the CFTC ask a long shot.
Slow judge, two-week clock. Brutal timing.
Three weeks from today, the co-founders of @Kalshi are scheduled to appear in a Utah courtroom to face criminal charges of promoting illegal gambling.
Unless a federal judge shuts down the criminal case... and we've actually seen that happen before, in Arizona.
The @CFTC asked a federal judge on 10/26 to "immediately" halt Utah's criminal case. But the judge has taken his time. There's a hearing set for next Friday. (And, unlike in Arizona, this judge already ruled that Utah can enforce its gambling laws against Kalshi.)
Anyway, the initial appearance in the criminal case is slated for 12/4. That means the federal judge will have two weeks to decide if that goes forward as planned.
Kalshi and its co-founders deny wrongdoing.
74
Primus zkTLS proves where the data came from, and that's about it
A verified API response can still be wrong, late, or answer a different question than the market asked.
The hard calls come down to rule wording.
Read the rules before the source.
Settlement is the part of prediction markets I never fully trusted
Someone has to say what actually happened
And you just hope their source is real
@primus_labs is working on exactly this
Their zkTLS proves the web or API data really came from where it claims
So the source gets verified before the outcome is decided
They are even exploring the same idea for private credit
Would you trust a market more if it settled this way?
67
Cole retweeted
Everyone is chasing the next Hyperliquid...
Most people show up after the points leaderboard is already a second job.
@QFEX is one of the few places that still feels early.
They’re building 24/7 perps on stocks, indices, and commodities (actual RWA exposure, not just more crypto pairs). Backed by $9.5M led by General Catalyst, with Y Combinator and Paul Graham in the round, team background is the kind that has produced people who later built the platforms everyone farms now.
The volume is already real, just not loud yet:
- over $10.9B in the last 30 days
- $607M on Nov 13th alone (DefiLlama)
That is not nothing it is just not the size that makes CT lose its mind yet.
The interesting part is what hasn’t been fully formalized.
No public points system neither no big Season 0 is live announcement that turns farming into a full-time competition.
early trading, feedback, and just showing up while the product is still growing is still relatively cheap.
There is no guarantee this pays.
team hasn’t published an exact formula or allocation
That’s the trade-off of being early.
I do rather test it while the cost of participation is still low and the room isn’t packed.
Once they formalize rewards and the leaderboard becomes public, the same volume that feels quiet today will suddenly have a lot more people fighting for it.
Being late usually just means more competition for the same upside. Being early isn’t a free lunch, but it was one of the few edges left in this cycle.
Testing while it’s still quiet >
[qfex.com/trade/US100-USD]
joining once everyone’s already in.
Louisiana just took Kalshi, Polymarket and 20+ venues to court
A Louisiana judge issued a temporary restraining order to halt trading on sports, elections and awards.
On the list: Kalshi, Polymarket US, Robinhood, Coinbase, Gemini, Crypto.com, Novig, ProphetX.
Attorney General Liz Murrill: no gaming license, no legal operation.
Illegal games of chance is a crime, with jail time and significant penalties.
She says judges across the country have overwhelmingly called this games of chance, and the platforms are making up new words to evade state games of chance laws.
Kalshi and Polymarket US already sued Louisiana officials in federal court in October.
The hearing on the order is set for Nov. 30.
2
10
483
Front Office Sports, October 2025: "The Trumps Are All In on Prediction Markets"
That all-in story has mostly unwound since. Brutal turnaround for a headline that big.
Trump Media & Technology dropped its plans to launch a prediction market of its own.
Donald Trump Jr. walked away from his advisory role at Kalshi.
Only Don Jr.'s ties to Polymarket are still standing.
2
16
342
Polymarket now prices two very different AI outcomes at the same 25.5%
Nvidia finishing 2026 as the world's largest company. And the AI bubble bursting by June 30, 2027.
Last Thursday the Nasdaq had its worst day since April 2025 and Nvidia fell 9%. Polymarket moves since last Monday:
Nvidia as largest company on Dec 31: 57% to 25.5%. Apple in the same race: 41% to 73%.
Anthropic valued at $2T+ by Dec 31: 66% to 30.5%. AI bubble by mid-2027: 22.5% to 25.5%.
Nvidia and Anthropic each lost more than 30 points. The bubble moved 3. Brutal gap.
That bubble market needs three of six triggers. One is Nvidia closing 50% below its record, and as of Friday it sits more than 20% below.
@polyfactual, a weekly podcast on Polymarket news, tracked the shift.
Polyfactual's read: traders marked down AI's leaders but barely touched the bubble, while US recession odds for 2026 fell to 3.9%.
Nvidia reports earnings on Nov 17, the day both 25.5% numbers get a real test.
Pegula goes for revenge on Kostyuk at the WTA Finals
Pegula leads their head-to-head 4-3, yet Kostyuk beat her in straight sets in Brisbane and Madrid this season.
Pegula's group opener vs Gauff was brutal: she saved a match point, took the second set, still lost in three.
Kostyuk found out she was playing one day before the tournament, then lost to Noskova 2-6, 4-6.
@iamigorekk, an AI analyst and researcher at Polymarket, notes Kostyuk came in short on match practice, so Pegula can pressure her second serve and control points from the return.
On Polymarket, Kostyuk trades near 63% to Pegula's 37%, on $11.77K volume.
3
18
473
1789 Capital put another $300M into Polymarket in August
An advisor whose fund keeps writing checks that size to your main rival is a hard seat to keep.
The midterm theory is a story. The $300M is the receipt.
Kalshi cut ties with Donald Trump Jr. !
They just kicked him out of the board of advisors.
Officially. It's because his fund "1789 Capital" invested $300M in their competitor Polymarket.
Realistically. The Trump ties have become too toxic after midterms.
The timeline proves it:
- Sepetember 2024 : 1789 Capital first invests in Polymarket
- January 2025 : Kalshi places Don Jr. on their advisory board
- August 2025 : Polymarket too places Don Jr. on their advisory board
- August 2026 : 1789 capital invest another $300M.
So Kalshi knew Don Jr. was involved with Polymarket since 2024 before even naming him advisor.
And now a few days after the Trump family suffers a huge electoral defeat they suddenly realize it's a problem ?
We all know what is happening.
2
5
233
31 forecasts from one GFS model is still one opinion
28 of 31 above 70F prints 90%, but all 31 share the same physics and blind spots. That 8% gap can be model error.
A $10K paper bankroll never had to get filled at 70c.
A free bot on GitHub scans Polymarket weather markets using 31 forecasts at once
I opened the code
I expected some heavy ML model
Instead it's simple counting, repeated on every market, every 5 minutes
Its game plan:
1. Every 5 minutes it pulls the open temperature markets on Polymarket: NYC, Chicago, Miami, LA, Denver
2. It downloads the GFS forecast from Open-Meteo. Not one forecast, 31 versions of it, each started from slightly different data
3. It counts how many of the 31 land above the market's line. 28 of 31 above 70F = 90%
4. It compares that 90% with the Polymarket price. Gap of 8% or more = signal
5. It never pays more than 70¢, sizes with a cautious Kelly (15%) and caps every trade at $100
6. Down $300 in a day and it stops itself
Where does the edge come from?
Your weather app shows one number. The bot sees all 31. Say the app says 72F but 9 of 31 forecasts say 69F. The 72F bucket suddenly looks a lot less certain than its price
The catch:
> It doesn't place real orders. It paper trades with a $10K virtual bankroll and shows you the signals
Clicking buy is on you
2
117
Cole retweeted
Copytrading on Polymarket can be profitable, but knowing WHO to copy is just as important as knowing HOW to copy
Take juice-fruit for example: gravia.trade/copy-trading?ca…
Personally since I'm mostly into sports, I'd rather find traders who specialize in sports markets and have a proven track record
NFL, Serie A, Europa League, tennis and more. Just look at some of his recent wins, a $400 copytrading backtest over the last 30 days would have turned into $812 (+103%)
With @graviahq you can find and backtest traders across your favourite niches to see who's actually worth copying
Nate Silver's fight with the markets got scored by a Kalshi study
A Brier gap that thin, graded by the venue itself, proves very little. Rerun the same 488 races on Polymarket prices.
Markets keep growing either way, and Polymarket belongs on that scoreboard.
Nate Silver vs. Kalshi’s CEO is the best fight of the midterms.
Tarek says the markets called it. Nate says “give me a break.”
Then a quant pulled 488 races and ran the numbers. Kalshi wins. 👇
tl;dr:
– Tarek: the people called the midterms on Kalshi
– Nate: polls and models led, markets just followed
– 488 US races since 2024: Kalshi had ~12% less forecasting error than the models
– head-to-head with Silver Bulletin on the 2024 presidential race: Kalshi wins, 0.0312 to 0.0348
– Kalshi’s own 2026 study: lowest Brier score of any forecast on Election Day, Silver Bulletin included
– Nate has been a Polymarket advisor since 2024
Lesson learned: markets don’t beat the models. They eat them.
Full details:
–- The fight timline
The night after the election, Tarek posted a victory lap: “the power to decide what America sees changed hands.”
Nate quote-tweeted it. His point: the polls and models gave the direction, and the better models were more confident and got there faster than the markets.
Then he went further on Silver Bulletin, saying markets are “low-key stealing valor from the models.”
–- The scoreboard
Quant Chad pulled 488 US races from 2024 to now and scored Kalshi’s pre-election odds against Silver Bulletin, FiveThirtyEight and JHK.
Average Brier score (lower is better):
Kalshi: 0.0282
Models: 0.0320
That’s roughly 12% less forecasting error for Kalshi.
Against Nate directly, on the 2024 presidential race across 50 states: Kalshi 0.0312, Silver Bulletin 0.0348.
Kalshi’s own 2026 study points the same way. At 9 a.m. on Election Day, Kalshi had the lowest Brier score of every forecast: Kalshi 0.0139, Silver Bulletin Deluxe 0.0141.
Kalshi also says House control passed 99% at 8:10 p.m. ET and Senate control at 9 p.m., hours ahead of the networks.
And Kalshi’s data is public. Nate’s 2026 race forecasts are paywalled, so the 488-race comparison had to use JHK for 2026.
–- Where Nate has a point
The models won 3 of the 5 categories in that data: 2024 Senate, 2024 House and 2026 Senate.
And a NYT analysis found markets did no better than the forecasts for most of the race.
– Worth knowing
Nate is a Polymarket advisor. Kalshi is now offering $1M to researchers to study its election data.
My read:
Both are right, and that’s the point.
The models gave the direction. The markets took the models, the polls, the insiders and the money, and priced all of it at once. That’s why the market comes out ahead across 488 races.
Nate isn’t wrong -- He’s an input.
EP11 * No Silver Bullet
7
335
RN1 is a Polymarket bot sitting on $15.6M in all-time profit
@PolyScalping, a Polymarket trader stats site, pulled its stats: 85,896 predictions, ranked #4 out of 3.3M traders on Polymarket.
The last 30 days alone, Oct 13 to Nov 10, added +$818K. That's an absurd month for any account.
Quick math: $15.6M over 85,896 predictions is about $182 per prediction on average.
PolyScalping's read: RN1 is the most successful active bot on Polymarket rn.
Jumper at 82c YES pays 18% for trusting a team's Q4 promise
Teams miss quarters all the time. NO at 19c is what you pay if you think this one slips, and I don't think that's cheap.
A guessed checker date is not free money.
Jumper token launch by 31 2026
On polymarket
18% free money right ?
TGE in Q4 confirmed by the team
At this point polymarket odds looks free money .
Personally I set a limit order 50 € doesn't matter its filled or not .
I always put minimum 1:1 or more
Anything can happen .
But i guess checker at November 20
13 november snapshot date 🌹
Hope for best... 💭
1
9
302
Polymarket wipes the order book when a market's wording gets clarified
Every resting order is cancelled. That's how a clarification works, per Polymarket's own docs.
The point from Ronak (@apsiri13), a self-described Polymarket lover on X: events move in ways nobody planned, and wording that looked obvious at launch can turn ambiguous after trading has started.
Clearing the book means nobody keeps trading against an outdated reading of the contract. Brutal if you had orders sitting there.
Ronak's read: change nothing and ppl get confused, change carelessly and trust breaks.
For Ronak, handling that line well is an underrated piece of market infrastructure.
2
16
705
Cole retweeted
Every perp DEX says its tech is the moat. I checked who actually built 5 of them, and only 2 clear my bar on both team and product 👀
My scores, team | product:
> @variational_io: quants with their own hedge fund since 2017, then engineering and quant trading at a big OTC desk. Zero fees, 550+ markets, $1.5B+ OI. 10 | 9
> @risextrade: HFT founder, own L2, bought BSX (ex-Coinbase, Kraken, Jump). Fully onchain orderbook, ~1ms. 9 | 9
> @arcus_xyz: incubated at dYdX Labs, built with Robinhood Crypto. 24/7 stock tokens, RWA perps. 8 | 7
> @QFEX: ex-Citadel, Tower, Flow Traders, YC. Stock, FX and commodity perps. 8 | 7
> @TxFlow_L1: own L1, 250K TPS claim. But the team runs the validators for now and most of its track record isn't public. 6 | 8
Early on the team matters more. A perp is a matching engine, a risk engine and liquidity. The first liquidation cascade shows who really built it.
Later it's the product. Nobody stays for a resume.
Variational built that OI over 20 months of private beta. RISEx's CEO held back points until the engine was stable, against his own growth team. That's a team decision you can see as uptime.
My rule: the team decides if I start farming, the product decides if I keep going. I'm farming both.
Both are still young tho. My opinion, not advice.
Team or product, what do you check first?
Polymarket is cutting staff to get ready for a possible 2027 IPO
Front Office Sports reports layoffs of some full-time employees last week, plus sharp cuts to contractors, mostly in marketing, as part of a broad reorganization.
Those viral posts claiming 30% of staff got cut? Not accurate, per the outlet's sources.
The company is also hiring new execs to "professionalize" ahead of going public.
IPO timeline set, but the POLY token question is still open.
4
19
1,298
