Blog / Field notes
Work in
public
Articles on quantitative trading, crypto strategy backtesting, market structure, and the boundary between an observed result and a proven edge.
16 / Field notes
Crypto Liquidation Data: Where Is the Trading Edge?
What 24 BTC forecasting tests, liquidation data and market-microstructure research establish before a market map becomes a paid signal.
KuCoin Dual Investment: A New Karaptic Lab
Buy low. Sell high. We’re going to collect the offers and test whether the yield compensates for the risk.
BTC Gamma: Gas, Brake, or an Unconfirmed Hypothesis?
Gamma concentration, 36 structural transitions, and the gap between an interesting drawdown relationship and a proven trading filter.
The Coin Rose. What Happened Along the Way?
3.3 million observations across 515 coins: larger rises can arrive with deeper falls and more negative weekly closes.
Volume Predicts the Future. The Future Just Always Arrives
Volume leads large moves by hours, but three tests found no useful separation from ordinary activity.
Why Crypto Move Size Is Hard to Predict: A Heavy-Tail Test
A concentrated tail, a failed magnitude model, and limited threshold ranking: what the latest Klines tests actually establish.
One Market Move or Several? The Problem of Crypto Trend Segmentation
How epsilon-tau, directional change and scale sensitivity help define movements before evaluating a market radar.
We Tested Momentum on the Exchange You'd Actually Trade It On
An 832-symbol Binance perpetuals replication shows how costs and an arbitrary rebalance date can change the conclusion.
Why an AI Trading Signal Is Not a Trading Strategy
Win rate, maximum gain, and a contract-risk badge answer narrower questions than whether a strategy can actually be traded.
A Trading Edge That Wasn't Wrong, Just Old
Early crypto research found a strong trend-following effect. In the post-July-2020 regime, it turned negative and statistically insignificant.
A 100% Win Rate, Without a Proven Trading Edge
Klines closed 15 of 15 paper trades positively. The number is real; what it proves is narrower than it first appears.
Nearly 4,000 AI Skills Scanned, 76 Were Malware. Who's Responsible?
A verified badge is not a security review. Agent skills inherit the reach of the systems that run them.
Why Bitcoin's Volatility Is Predictable and Its Direction Isn't
Volatility clusters and options price a range. Neither gives you tomorrow's direction.
Why I Stopped Trying to Make AI Predict Bitcoin
Detection is not prediction, and a beautiful backtest is not evidence of an edge.
We Started Recording the Market
Karaptic has begun building a synchronized archive of BTC derivatives data across Binance and Deribit.
The Grid Bot Illusion Is Also a Compliance Lesson
A return can be numerically accurate and still materially mislead when the fee tier, capital base, or holding period behind it is missing.