Ivan Varvar

A profitable Polymarket wallet doesn't tell you what you'd earn by arriving late. The trader's entry price isn't automatically yours. Polywhaler's Nov 7 changelog describes batch simulations with two versions of the same copying scenario: at the trader's prices, or estimated after a delay and fees, including partial fills. The useful question shifts from "who made money?" to "could I repeat those trades on my terms?" Both versions are models, not live returns or a promise that copying pays.

Two traders can buy the same outcome at the same price and still pay different fees. Polymarket US's published fee plan, scheduled for Nov. 6, generally gives higher-volume accounts lower fees for taking existing offers. For a hypothetical 100-contract purchase from existing offers at 50 cents, the lowest-volume tier would mean $2.38 in fees versus $1.74 at the highest, before rebates. New accounts start at the best standard rate for seven calendar days, counting the day of their first trade. In November, existing accounts get the better of their October or rolling 30-day tier.

Research in chat. You still place the trades. Kalshi's Nov. 6 MCP update describes AI access to odds, order books and portfolio risk. Its docs rule out placing or canceling orders, but allow price-alert creation and deletion with separate permission.

Kalshi announced single-stock perpetual futures on Nov 5, saying they cover 50+ companies with long/short exposure and no scheduled expiry. The distinction: stock-price exposure, not shares or binary event bets.

The Brewers’ World Series YES book became much thinner near its midpoint. Combined bid-and-ask depth within ±2 percentage points fell from about $19.2k to $7.3k and was still around $7.1k at the latest snapshot, even though the midpoint was unchanged at 63.85% between the later readings. Available size at the best bid dropped from 196.65 to 39.24 YES—not enough for a 100-YES sale entirely at that price. Yet a static 500-YES sale estimate fell by only about $0.19, from $315.91 to $315.72. The book was substantially thinner, but the cash effect for this trade size was modest.

🚨 EU tightens the rules on stablecoins! ESMA says MiCA-authorised crypto firms should stop offering services involving non-compliant stablecoins — including trading, transfers and custody. Existing exposures should be addressed within 3 months. ⚖️ Investor protection or regulatory overreach? What do you think? 👇 #MiCA #Stablecoins #Crypto

👀 Ever heard of the “Shy Neighbor Effect”? In 2024, a French trader used private polls asking: “Who will your neighbors vote for?” He bet big on Trump via Polymarket and reportedly made $85M in profit. 💰 🧠 Brilliant strategy or just luck? What do you think? 👇 #Polymarket

Heard about the Polymarket / Strategy BTC fiasco? Strategy reportedly sold 32 BTC before the market deadline. But because the confirmation came later, the market resolved NO. So traders who were right about the real-world event still lost. That’s resolution risk in one of its clearest forms. What do you think — fair resolution or complete failure? Write your thoughts in the comments. #Polymarket

Displayed YES order-book depth in the market “Will the Republicans win the Georgia governor race in 2026?” fell from $13,388 to $6,525, then recovered to $7,954, only partway back to its starting level. These selected snapshots measure USD resting orders on both sides within ±2 percentage points of the midpoint, not trades or a probability change. The latest sample was 59.4% of initial depth, and the observed sequence confirmed neither the low nor a full recovery. If depth reaches $13,388 in a later snapshot, it would be back to its starting level. A return to $6,525 would erase the rebound.

How to lose $700k on Polymarket when you’re 100% right Simple math: 9 wallets control >53% of active UMA oracle voting power. They can officially declare any lie the truth. 🧵 Breakdown of a clean “oracle attack”: 1. Cheap accumulation Whales quietly buy the clearly wrong outcome (“NO”) at ~$0.03. Retail, 97% sure of “YES”, dumps hundreds of thousands into the pool with zero hesitation. 2. Token deposit The real-world event happens. Whales post the tiny $750 bond and open a dispute on UMA. Retail laughs: “They just threw away $750.” 3. The Vote of Nine The same 9 wallets that hold majority active UMA voting power vote for the obvious falsehood in unison. Majority rules. The lie becomes the official on-chain truth. 4. Pool wipe Polymarket’s smart contract automatically pays the entire pool ($700k+) to the “NO” holders. Honest trader’s balance: $0. Code is law. The real game This is Profit from Inaccuracy. Whales don’t care if UMA’s price takes a temporary hit from the reputational hit — a $50k token loss is easily covered by a $700k+ market extraction. Even worse: if a small UMA holder votes for the actual truth, they get slashed. Their tokens are taken and given to the winning (lying) side. Pure penalty for honesty. Bottom line On Polymarket you are not trading the probability of real-world events. You are betting on whether it is currently profitable for nine anonymous wallets to lie to the smart contract. #Polymarket #UMA #Crypto #DeFi #Web3

Price up. Observed flow leaning NO. On Sept 8, Khachanov’s US Open YES probability rose 1.8 percentage points while observed flow leaned the other way. What would you check before trusting that move? A look inside VARVAR ↓ #Polymarket

News moves prediction markets. Now the context moves with them. We’ve added a news engine that connects relevant stories to live markets: • Latest market news on the homepage • Relevant news inside each market • Sources, timestamps, and relationship context See the story behind the probability. #PredictionMarkets #Polymarket

Interesting ≠ Executable We fixed the freezes, rebuilt the runtime, and added a single execution layer across the terminal. While testing 150 markets, we found a hidden issue: some markets looked tradable even though their order books were stale and execution was effectively blocked. We fixed the classification without changing rankings. The terminal now separates two questions: Is this market worth watching? Is it actually safe to consider an entry? That distinction matters.

Lesson for anyone running heavy backends on live trading platforms: One careless window function or forgotten full ORM object and your service starts choking the entire system at the worst possible time. Visibility + obsession with boundaries + controlled experimentation = winning. VARVAR public data agent memory hardening: complete. Now it runs clean and fast even under load. Who else is fighting memory leaks and temp file explosions in prod? Drop your war stories below.

The numbers that hit different: Before: 48.6 GB peak, 50+ GB temp in 6h After: ~685 MB on the largest worker, temp rate down 85% Ran controlled profiles (100 → 500 → 1000 markets) + full 60-minute soak. Zero API failures, swap = 0, workers rock-solid at 20/20. This wasn’t a quick fix. This was infrastructure surgery. The real flex: how we validated it. Not “it works on my machine.” We built: PID-bound monitor Stage-level JSONL profiling Page accounting (unaccounted time < 0.5%) Strict idle-transaction tracking Every step is reproducible. This is engineering with proof, not hope.

What we did: Broke everything into clean pages (100 markets each), built lightweight OrderbookSnapshotSummary, switched details to full sessionless mode: read → close session → heavy Python compute → write. All prefetch families now chunked to max 25 markets. Added hard safe-boundaries, controlled pause/resume, and runtime deadlines. Now the agent finally behaves like a disciplined market maker — no more memory fights and no holding transactions hostage. Stay tuned. The best part is coming next 👀