Bizzi

Polymarket beat CNN, Fox News and AP Last week poly dropped a breakdown of how it performed during the US midterms. Want to show you too and once again highlight the difference between voting with your opinion vs voting with your money. Numbers there are such that traditional media should just open an account on Polymarket and start reposting. Polymarket took 10 key races and compared when the market hit 99% and posted the call on X vs when CNN, Fox News and AP announced the result. Result: Poly was first 10 out of 10 times. Median lead approximately 31 minutes. > Michigan Senate – Poly called it at 9:46 PM, the first media outlet was Fox at 1:50 AM. +4 hours 4 minutes. CNN and AP came in closer to 2:30 AM.
> Texas and Kansas Senate – +1 hour 33 minutes ahead of CNN each.
> Florida Governor – 8:38 PM vs. 9:09 PM for AP, +31 minutes. Newsrooms wait for the votes to be counted and for lawyers to sign off on posting. Prediction markets don't need lawyers for that because the moment someone gets information, it instantly turns into price. The stats on favorites and resolution are interesting too. > The favorite won 419 of 426 resolved races. That's 98.4%.
> 7 upsets, with 78 more races still being counted.
> Senate 97%, governors 97.1%, House 98.6%.
> Control of Congress: the market gave both chambers to the Democrats a month out, a week out and a day out. Senate the day before the election – 73.9%, House – 94.7%. Both calls hit. Yes, 98.4% sounds very impressive, but let's be realistic about it. Most of the hundreds of House races are districts where the outcome was clear before the campaign even started. Anyone could call those and they inflate the overall chance. What really matters is calibration. If the market prices something at 70% then over the long run those events should happen roughly 7 times out of 10 (not 10 out of 10). Moreover the 60–70% bucket, favorites won only 60% of races – they lost 2 out of 5. In the 50–60% bucket – 75%. I don't want to draw some overcomplicated conclusion, so I'll keep it simple – trading on news is -EV (exceptions Bloomberg terminals, the Truth Social API and being related to Trump).

When Fable 5.5?

All I need is an affair with Chinese crypto mafia

Last few days the Anthropic IPO market tried to shake many traders out but I bought it instead. Quick recap if you missed the first post: I was long YES on “Anthropic IPO by Dec 31" from 83-85¢. Then Thursday happened. Nov 5: Nasdaq 100 −6.4% (worst day since 2020), S&P 500 −4.4%, semis −12%, NVDA −9.3%. Democratic wins in key data center states sparked fears of tougher AI rules and a slower buildout and the whole AI trade got dumped in one session. The drop on this market started right at the US open (91¢ before the open, 79¢ an hour later, 66¢ low Friday morning, 87.6¢ rn). And nothing about the IPO itself had changed. Nothing. No withdrawal on EDGAR, no postponed headline from anyone credible. Just fear that a $1T+ listing can't price into a market that just had its worst day in six years. It was the momentum where you had best entries. So I went shopping. I split bids across a few accounts and got filled at 72¢ on some and 70¢ on others. So then Anthropic proceeds with roadshow next week, target listing week of Nov 16 (before Thanksgiving). Latest reported range is $1.25T to $1.75T, raising $80B to $100B. But bro a cheaper IPO is still an IPO and this market doesn't care about the price tag.

Anthropic IPO by ?

I took this screenshot on purpose because I knew this situation will be. On the one hand, your knowledge and theory about new GPT image model release. On the other, thin market which won’t allow you to earn money on that edge. It’s just sad.

Kimi K3.1 Pro shipped open weights and Polymarket still gives it a ~4% shot at debuting top 5 on Text Arena. That's a bond guys. K3.1 Pro is a post trained K3 with the same architecture, just a bigger post training budget. Moonshot sold it on coding, agents, Blender, CAD, finance modeling. Not on being a better chat model. And Text Arena overall is mostly general prompts where coding is doesn’t matter much. So to crack top 5 it has to land 32 points above its own predecessor and jump 20 spots on day one. I've been sitting in No for a while. Yeah spread is wide rn so don't market buy at 99.5. Park a limit at 95, 96 or 97 and let someone fill you. That's roughly 3 to 5% in a few days.

Polymarket perps comp is live (Nov 2 to Nov 16) and I think most people are gonna farm it wrong. $50K pool, Top 50 get paid (1st takes $10K, and even 50th gets $250). But read the scoring before you click anything. 4x the volume only doubles your multiplier. Fees, though, hit every single fill and come straight out of PnL. So the classic degen move (spam market orders, max lev) works against you. Big volume on a red account just makes your score more negative. How I'd play it: 1) Farm volume only on boosted markets. A 3x boost gives ~1.73x score on the same notional and your PnL stays untouched. 2) Use limit orders. Maker rebates add to PnL, taker fees eat it. 3) Check funding before entry. Funding you receive counts as PnL (funding you pay gets subtracted). 4) Close old positions first. 5) Boosts don't count toward the $100K minimum. Closes do, though, so $50K in and $50K out gets you there. 6) Open positions get marked at oracle price at the end (don’t let a late wick decide your rank). Small green PnL × clean boosted volume beats a huge volume number on a red account. Also, no wash trading or alt wallets. That's a DQ for every linked account. Under 2 weeks left rn. Take part of the tournament: polymarket.com/perps?c=01ccl…

October AI markets on Polymarket are closed. Let’s look where I made money last month. 1) Best Chinese model in October Early in the month Xiaomi's MiMo V2.6 Pro jumped into the Text Arena top 12 with 1491 and passed every Chinese model. It held that lead the whole month but the market never fully believed it would stay there (everyone was waiting on new Chinese releases), so the price kept floating around 50 to 60%. I started building position right after Xiaomi landed in the top. Moreover I decided to build multi position with 2 legs. Xiaomi leg was the main size (avg entry ~60% or a bit lower). Alibaba leg at ~20% as a hedge (Qwen 4 rumors ran for half the month and I didn't want it flipping Xiaomi). Whole position cost ~82 cents. 2) Second best Chinese company Came in late here, about a week before the market close. Alibaba was trading ~85% and the risk was Alibaba dropping a new model in the last days and reshuffling everything. But the clock was tight (a model like that needs time to debut on Text Arena), the weekend was coming and the chance of a release landing in time was close to zero for me. Couple days later Alibaba officially confirmed new models drop this week. Doubts gone and the market turned into a full bond you could keep adding to at 90%. Only thing in the way was thin liquidity. I was also watching Kimi K3.1 Pro but that model is built for agentic work and wouldn't have performed like that on Text Arena. 3) ByteDance and text to video Found this one by accident while researching other markets. NO on ByteDance having the best text to video model was 89c. So on the leaderboard ByteDance's best model, Seedance 2.0, sits 5th, behind both Gemini Omni models, FLUX 3 Video and Grok Imagine. And the leaderboard hadn't even updated since September 21 (idk why). Mapped out which video models were shipping soon and realized this was a bond too. November is already loading and I think we gonna have the sensational month because of Fable, Astra and Carbon. So we will have to catch those opportunities. Moreover Qwen 4 should ship this week and has real odds of breaking out on Text Arena, but no aggressive sizing until I see what they're optimizing for. Just have a look at the Chinese models market.