The Crypto Kabum

@Cryptokabum

Alive in crypto since 2013. never get tired of making it and them looosing it all

Brazil Katılım Mart 2018

186 Takip Edilen90 Takipçiler

The Crypto Kabum retweetledi

Superteam Brasil@SuperteamBR·5d

329 submissions. 10 winners. 🧵 The Solana AI Kit Skills @SuperteamEarn bounty results are in and we have some heavy hitters! Congrats to everyone who shipped, and thank you for making the Solana agentic space even more powerful 🦾🤖 🏆 The Top 5 Skills 🥇@RZ1989sol solana-cpi-safety: detects and prevents four cross-program invocation vulnerability classes + coverage of CPI return-data spoofing 🥈@blchead solana-auditor: production-grade security audit for any codebase, 18 checklists, 1,209 verification items, 109 known attack vectors 🥉@soltyagi solana-indexer: expert skill at reliably indexing on-chain Solana data into a database and serving analytics; idempotent, cursor-checkpointed, and reorg-safe 🏅 @sou_foka community-moderation/ Foka AI: comprehensive evasion-resistant moderation skill backed by Foka's vast expertise; it moderates scams, manages member contacts, triages support, and routes questions to the right person/links 🏅 @ZiBibro crypto-legal: legal and compliance triage for Solana-native founders across the United States, the EU, and Brazil juristinctions + we couldn't leave out @Cryptokabum position-manager with the most complete skill around CLMM management across @MeteoraAG @Raydium & @orca_so, who submitted a killer PR but forgot to submit to the bounty in time 😂 Up next👇

English

4.2K

The Crypto Kabum@Cryptokabum·27 May

@Ak_sire47 @Gustafssonkotte all my parquets are huge! i think compressed 30gb alredy, i can upload it to a drive and send you, but i dont know if you can use it in this plataform you are using. i run all my engines direct in python. i will dm you

English

Sire the video star★@Ak_sire47·27 May

@Cryptokabum @Gustafssonkotte oh hey share! I also got something like that from, i ran a realistic backtest with my strategy and retained 86% at 5min and 92% at 15min market, still wanna try with different mmp ranges, please share yours polybacktest.com

English

Sire the video star★@Ak_sire47·26 May

I've been reading your write ups from day one, a bunch all you figure out i figured ealier with slight variations in approach, would you like to read my write-up? you might find it useful.

Kotte@Gustafssonkotte

UPDATING THE CHEAT CODE FOR POLYMARKET The strategy looked great on paper and worse in the wallet. Today I found out why, and it was not the strategy. - The gap that bothered me · On historical data the setup held 96 percent · Live, by the export not the bot log, it ran in the low sixties · Same logic, same thresholds, a 30 point gap · When the model and the wallet disagree that hard, the model is measuring something the wallet never gets - I went looking in the execution path, not the logic · The brain was firing correctly, the right calls at the right moments · The losses were not bad decisions, they were good decisions arriving late · A signal is worthless if the fill happens after the moment has passed - What I found was an order sitting on its hands · The bot placed a resting limit order and then waited up to twelve seconds for it to fill · In the final stretch of these markets, twelve seconds is an eternity · It would set a price, the market would move, the order would hang unfilled · Then it would chase the price, paying more and arriving later - The historical edge was measured on a perfect snapshot · On paper I read the exact moment and the exact price · Live, by the time the order lands, the price has already walked away · The 96 percent was real, it just assumed an instant fill I was not getting - The fix was to stop waiting · Switched from a resting order to a fill-or-kill style that executes now or cancels · Set the price with a small cushion so it clears the book on the first touch · No more twelve second hope, no more chasing · Pay a hair more, but actually be in the trade when the signal says so - What I keep relearning · An edge on backtested snapshots is not the same edge in live execution · The slowest part of the loop is where the profit leaks out · The brain was never the problem, the hand was too slow · Behind one env flag, off and it reverts, as always Same strategy, same thresholds. I just stopped letting the order dawdle. Now I watch whether closing that gap moves the live number toward the paper one. I will report.

English

124

The Crypto Kabum@Cryptokabum·27 May

i used this guys to get the first data: telonex.io . i dont used this polybacktest. i did all my infrastructure in python. I made a collector copying their parkets structure, and some parts of this repo : github.com/Polymarket/pol… . got a vps in ireland to run the infrasctucture for the bot and the collector. im running live for 20 days now but just with 4 directional strategies, the others are maker arbitrage or position abitrage, those are getting better results.

English

The Crypto Kabum@Cryptokabum·27 May

Im Reading the Gustafsson relates from a while too. And i Took the fredom to read yours, if you want, i have a Lot of collected data from polymarket 5 and 15 min markets(sol, eth and BTC). Im running a collector for more than 3 months now. Last 3 months 10h a day thousands of trials i have found 7 direcional good strategies. 😂.

English

Sire the video star★@Ak_sire47·26 May

@Gustafssonkotte will update the write up when i'm done with tonight's update. One thing to note, i went from entering in the last 10 secs to 300s for 15min and 15sec for 5min if all my conditions are met, also reduced MIN_MOVE_PCT= 0.35 as well. Generally we can talk better in dm

English

The Crypto Kabum@Cryptokabum·25 May

@AlterEgo_eth do you have nother valid code ? this one expired :/

English

Alter Ego@AlterEgo_eth·24 May

You can also find and analyze traders with similar narrow-bin + scaling strategies using Parity Terminal: predictparity.com/?code=3vxuh7cw

English

913

Alter Ego@AlterEgo_eth·24 May

How ColdMath Dominates Polymarket Weather Markets (Narrow 1°C Bins + Aggressive Scaling + Multi-Model Bias Correction) I went through more than 1,200 of his recent trades using Parity to understand the system behind the results What looked like random activity at first turned out to be an extremely disciplined and sophisticated strategy 1. Ultra-Narrow Bins - The Core of His Strategy He actively trades very precise values and tight ranges, often just 1 degree. Recent examples include: "Will the highest temperature in Singapore be 34°C?" "Will the highest temperature in Kuala Lumpur be 29°C?" "Will the highest temperature in Houston be between 70-71°F?" This is not random. Consistently profiting from such narrow outcomes is only possible with very high forecast accuracy. 2. Aggressive Position Scaling (Scaling In) One of his most distinctive traits is entering the same outcome dozens of times with small amounts ($0.01 to $10–50, sometimes up to $100–200) Example from today: he made 8–10 separate buys in the same market at 7, 10, 16, 46, 115, and 290 shares. This is classic technique: minimizing slippage, averaging entry price, and the ability to quickly stop adding to a position if the forecast changes. 3. Extremely High Trade Volume He trades a lot. Some days he makes dozens, sometimes hundreds of trades. This clearly indicates a fully automated bot 4. Aggressive Betting on 1¢ Yes/No Outcomes He often builds massive positions in very cheap outcomes (1¢), which shows high confidence in his probability calculation after all adjustments.What we can infer about his strategy: • He most likely uses a multi-model blend • He applies bias correction at least at the city + horizon level • He has a dynamic component that adjusts model weights based on recent performance • Strong emphasis on real-time data (METAR, etc.) ColdMath is not just “a bot that pulls ECMWF” It’s a well-thought-out system that combines multiple forecast sources, bias correction, and extremely disciplined position management through micro-scaling This is exactly why he delivers outstanding results even with a relatively average win rate (~63-65%)

Alter Ego@AlterEgo_eth

Here are the 10 Best GitHub Weather Bots for Polymarket Right Now (Beginner to Advanced) I went through dozens of repositories and filtered out the ones that are actually usable in 2026 From simple ready-to-run setups to serious multi-model ensembles with Kelly sizing and live execution - here’s what’s worth your time > Beginner-friendly / quick start 1. MoonsatProtocol/Polymarket-Weather-Bot: TypeScript, Kelly Criterion sizing, scans NWS forecasts, finds mispriced temperature markets, simulation mode + JSON logs. Has a paid advanced version Repo: github.com/MoonsatProtoco… 2. MusicBoiyzzz/Polymarket-Weather-Bot: same Kelly + NWS stack, focus on easy setup, ships with a ready-to-run .zip Repo: github.com/MusicBoiyzzz/P… 3. hcharper/polyBot-Weather: designed for regular hardware (Mac Mini, home PC). Gaussian probability model on NOAA data for weather markets. Also covers crypto via Black-Scholes and binary arbitrage. Copy trading built in. Streamlit dashboard Repo: github.com/hcharper/polyB… > Multi-model ensemble (serious edge) 1. alteregoeth-ai/weatherbot: pure Python, three forecast sources: ECMWF, HRRR for US cities, METAR observations. Scans temperature markets, enters when EV is positive. Kelly sizing + simulation mode Repo: github.com/alteregoeth-ai… 2. AadiXD200/polymarket-weather-bot: autonomous quant bot, multi-model ensemble forecasting, Kelly sizing, live CLOB execution. One of the more complete implementations. Repo: github.com/AadiXD200/poly… 3. yangyuan-zhen/PolyWeather: monitors 52 cities, uses DEB (Dynamic Error Balancing): blends multiple model forecasts weighted by recent accuracy per city. Generates calibrated probability buckets, scans for mispricing. Web dashboard + Telegram bot Repo: github.com/yangyuan-zhen/… > GFS ensemble specialists 1. suislanchez/polymarket-kalshi-weather-bot: trades Polymarket and Kalshi simultaneously. Uses 31-member GFS ensemble from Open-Meteo, converts ensemble agreement into probability, enters when edge vs market price exceeds 8%. React dashboard, FastAPI backend, SQLite tracking. Repo: github.com/suislanchez/po… 2. kylecwalden/polymarket-sniper-bot: sniper bot with a dedicated weather strategy using GFS 31-member ensemble. Also includes crypto strategies. Repo: github.com/kylecwalden/po… > Kalshi-focused (transferable logic) 1. AruneshDev/Automated-Trading-System-Kalshi-Weather-Model - ML-based system (XGBoost) built specifically for weather event contracts on Kalshi. The model architecture is directly transferable to Polymarket weather markets. Repo: github.com/AruneshDev/Aut… 2. cpratim/Kalshi-Weather-Trading: simple weather trading bot for Kalshi, clean codebase, good reference for understanding the base execution layer Repo: github.com/cpratim/Kalshi… You don’t need to reinvent the wheel. These 10 repositories already solved 80% of the hard problems Pick one that matches your skill level, study how they structure the code, and build on top of it The traders making consistent money on weather markets right now aren’t the ones writing everything from scratch - they’re the ones who start with good foundations

English

102

25.3K

The Crypto Kabum@Cryptokabum·27 Nis

@Gustafssonkotte my bot has a similar structure, if you have a good gpu, try to traine this model github.com/shiyu-coder/Kr… . I trained it with btc 1min OHLC and im getting good results here.

English

178