What Live Trading Taught Us
Nine months of filled orders, not simulation. How the log changed stops, sizing and trades we stopped taking.
Read the post ›Blog
Research, signals, and strategy from our AI trading desk
Real-money desk: 291 market signals every 4 minutes across 9 assets on Kraken, Binance and OKX. Trade log includes costs.
New VPIN: 90 seconds before the move →291signals read before every buy and every sell
4 minfull decision cycle, day and night
9assets traded, three exchanges
Executed on
Kraken Binance OKXSpot on all nine assets. Futures on BTC, ETH and SOL only.
VPIN: The Indicator That Detects Smart Money 90 Seconds Before the Move
Order-flow toxicity in volume buckets. What VPIN sees before price reacts, and why that lead time is shorter than it looks.
Read the post ›Three defenses, by name
Most of what we write about risk comes from these three, because they are the parts of the system that act while nobody is watching.
Stop-Hunt Detector
Reads BTC, ETH and SOL together and flags coordinated manipulation within 60 seconds, so the desk does not hand over its stops.
Proactive Wick Buys
Limit orders parked at −1.5, −2.5 and −3.5% below reference. When a wick reaches down, the fill is already there.
Grid Auto-Reposition
Recentres the range when price drifts more than 5% off it: the difference between a working grid and one left behind.
The index
Ten posts from the desk
Showing all 10 posts.
What Live Trading Taught Us
Nine months of filled orders, not a simulation. What the log changed about our stops, our sizing, and the trades we stopped taking.
Read ›80% of BTC Volume is Bots. Here is How We Detect When Their Cycle Ends
Machine flow leaves a signature in order size and cadence. The three prints marking the end of a bot-driven leg, with our thresholds.
Read ›Why Position Size Matters More Than Entry
The same entry at two different sizes is two different strategies. Where the variance actually comes from, written out with the arithmetic.
Read ›Dynamic Hedge Engines: Why They Are Not Insurance
A hedge costs money every day it stays open. What a dynamic hedge engine pays for, and when it is no longer worth the carry.
Read ›VPIN: The Indicator That Detects Smart Money 90 Seconds Before the Move
Order-flow toxicity in volume buckets. What VPIN sees before price reacts, and why the lead time is shorter than the headline.
Read ›Funding Rates: The Hidden Tax Most Traders Don’t See
Perpetual funding is charged whether right or wrong. How eight-hourly payments rewrite leveraged-trade break-even.
Read ›How A Year Of Backtesting Killed Our Best-Looking Indicators
The prettiest equity curves died first. What survived 12 months of walk-forward testing, and what overfitting looks like inside.
Read ›Why Single Signals Lose and Multi-Signal Consensus Wins
One indicator is a coin flip with extra steps. How agreement across 291 signals changes outcome distribution, not hit rate.
Read ›Grid vs DCA: The Math Behind Two Bot Strategies
Grid harvests range, DCA averages drawdown. The same market, two very different loss profiles, laid out with the numbers behind each one.
Read ›The System Doesn’t Predict: Where the Edge Really Lives
Prediction is the wrong frame. The edge sits in execution, cost control and position size — and here is how much each one contributes.
Read ›Ten posts, and the eleventh publishes when the trade log gives us something worth explaining.
One email per post1. No forecasts2. No affiliate links3.
New research lands about twice a month: order flow, funding, sizing and risk machinery. Same site text, sent on publication day.
Stored in this browser only. Nothing is sent anywhere.
1 One message per published post, plus nothing else. No digests, no partner mailings.
2 The desk does not forecast price and the blog does not publish targets, signals to copy or expected returns.
3 Exchanges are named because the desk trades on them, not under any referral arrangement.