Quasar
Quasar is built on a simple idea: a trading system should not rely on trust or human intervention. It should read the market, understand conditions, and act with controlled risk. At its core is ARC, Adaptive Regime Control. The market is always shifting between conditions like trends, volatility, and noise. Quasar continuously identifies these regimes and adjusts its behavior. It does not use one fixed strategy. It first understands the environment, then responds accordingly. In unstable markets it becomes defensive, in clear trends it allows continuation, and in unclear conditions it reduces activity. The system is driven by deeper market data, not basic indicators. It looks at order flow, volume behavior, CVD, liquidations, open interest, funding, and trade aggression. These reflect how participants are positioned and behaving, not just price movement. This allows Quasar to read market intent, not just outcomes. Risk is built into every decision. Position sizing and exposure depend on both market conditions and system performance. When conditions are poor or performance drops, risk reduces. When alignment improves, exposure can scale. This creates a self-regulating system. Quasar is trustless because every action follows defined logic. There is no discretion or hidden decision-making. It operates consistently based on data, adapting to changing conditions while maintaining strict risk control.
Category tags:
"1. Application of Technology: 4.5 / 5 Justification: The data engineering here is extremely impressive for a hackathon. Most inputs to typical trading bots are just Price and Volume. Quasar mathematically calculates cvd_accel (Cumulative Volume Delta acceleration), oi_change (Open Interest change), and long_liq_ratio (liquidation ratios) in numpy to classify market regimes. Bringing order-flow dynamics into the AI evaluation is top-tier. 2. Presentation: 4 / 5 Justification: Very clean presentation. Next.js dashboard, FastAPI, Postgres, and Docker Compose scripts make it easy to deploy. The narrative of "Adaptive Regime Control" over "Predictive Chatbots" is highly professional and appeals to quantitative judges. 3. Business Value: 4.5 / 5 Justification: Very High. Order flow and liquidity dynamics (like funding rates and liquidations) drive modern crypto markets much more than standard moving averages. A bot that uses AI specifically to read liquidity structure and position aggregation to scale risk exposure is exactly what proprietary trading desks build internally. 4. Originality: 4.25 / 5 Justification: While "Ask OpenAI for a second opinion" is standard, using statistical numpy slopes on complex market microstructure variables (like Order Flow Delta and Trade Aggression) as the primary gatekeeper is a highly original angle for this hackathon compared to the dozens of "give the LLM price data and ask whether to buy or sell" projects. ⚖️ Pros & Cons Pros: Microstructure Features: Deep order-book features (CVD, liquidations, OI, funding rates, trade aggression) are objectively superior data inputs for short-term futures trading compared to lagging technical indicators on OHLCV candles. Veto-Only LLM: Limiting OpenAI to a veto/confidence-adjust mechanism on top of a deterministic mathematical engine reduces hallucination risk and saves massive amounts of API compute cost (it only pings the LLM if the math already likes the trade). Full Automation: Monitors open positions and automatically closes them out on Take Profit, Stop Loss, or a strict 1-hour timeout, preventing trades from "drifting" into losing regimes. Cons: Kraken Futures Constraint: The bot targets Kraken Futures. While very profitable, many hackathon judges or retail users might only be comfortable (or legally allowed) to trade Spot crypto, somewhat limiting the immediate testability for some audiences. Opaque Math: The regime classification is done through heavily customized numpy polynomial fits (np.polyfit) rather than standard libraries, which might make it difficult for an open-source contributor or judge to quickly tell if the math is actually sound or if it is curve-fit garbage."
Sanem Avcil
For Innovators & Creators