Lo: $--,---
Pass/Fail Distance Threshold
A 15m target PASSES if the actual Kraken 15m close lands within this distance of the estimated target. Setting to 🎯 $5 Pinpoint activates L1 Zero-Bias Pinpoint training to pull estimates as close to a $5 gap as possible.
If next target is above the previous target (e.g. >$80,547), actual 15m close FAILS if it finishes below $80,547—even if inside the % corridor—and triggers a 1.5x Directional Flip ML penalty.
15m Close vs. ML Estimated Targets
Showing recent candles + live 15m target projection
Recent 15m Target Evaluations
Online Walk-Forward ML Architecture
At every 15-minute close on Kraken, the bot measures the exact residual error e_t = ActualClose - EstimatedTarget. If a target misses closely or fails outside your threshold, the model performs an immediate **Huber-loss gradient descent step** and updates a **Kalman residual bias tracker** before estimating the next 15m target.
Deploy this bot on a $3.50–$5.00/mo AWS Ubuntu instance. Enter your AWS Public IP below to link this exported app or Android APK directly to your 24/7 cloud server from anywhere.
Rolling Accuracy & Residual Convergence
Watch accuracy climb as ML adapts across the 200 targets
Live Learned Feature Weights (8 Features)
Updated t=200Last 200 Targets Pass/Fail Ledger
Evaluated at the exact 15m Kraken candle close