🚀 No hype. Just a clean, working setup to finally see what the bot actually does. Trading at a new level.
📊 The New Analytics Dashboard
We built a local dashboard that visualises everything:
candidate_log.csv – every signal, why it was blocked or selected
scan_stats.csv – per‑scan BTC regime, ADX, candidate counts
trade_outcomes.csv – all simulated trade results with PnL and exit reason
signal_publish_queue.csv – signals ready for Binance Square
It runs on the my Backup PC
The dashboard includes:
Scan Funnel – how many coins pass ADX, MTF, selection
Blocked Reasons – top filters that kill signals
Strategy vs Regime – which strategies work in trend, range, reversal
Regime Timeline – BTC regime over time (trend_up, range_low_vol, etc.)
Coin Drilldown – performance per coin, click to filter
Threshold Pressure – RSI/ADX/Vol distribution for winners vs losers
All with filters for strategy, regime, symbol and time range.
🔥 Heatmaps & Visuals
The dashboard gives you instant heatmap‑style insights:
Blocked Reasons by Strategy – quickly see if ADX, RSI or Volume is the bottleneck
Regime Performance – colour‑coded PnL per regime per strategy
Symbol Ranking – green for positive EV, red for negative, sorted by PnL
Example heatmap view:
https://via.placeholder.com/800x400?text=Heatmap+%E2%80%93+Strategy+Performance+by+Regime
🧠 Key Improvements in V7.6
1. Data Contract – Clean Separation
We defined a strict target data contract:
candidate_log.csv – events only (written by signals.py)
trade_outcomes.csv – outcomes only (written by simulate_outcomes.py)
scan_stats.csv – per‑scan BTC regime and summary
No more cross‑writing, no more confusion.
2. Filter Tuning – More Signals, Still Safe
We loosened filters just enough to get real SIM_TRADEs again:
q2_balanced now catches intraday breakouts (ADX ≥15, vol ratio ≥1.05)
q1_trend accepts weaker trends (trend_strength_min = 0, ADX ≥18)
q4_statistical now sees squeeze setups (bb_squeeze_min = 0.01)
q5_compression_revert tuned for range markets
Result: first SIM_TRADE in days – LTCUSDT LONG via q2_balanced.
3. Binance Square Auto‑Posting – Finally Working
The new publish pipeline:
signal_publish_worker.py creates payloads
signal_publish_dispatcher.py marks them ready
signal_publish_sender.py posts to Binance Square using the OpenAPI
Real‑time auto‑posts for every new SIM_TRADE and TRADE.
Example format:
text
$LTC
LTCUSDT LONG
Strategy: q2_balanced | Score: 41.5 | RSI: 65.1 | 4H: NEUTRAL
#AutomatedSignal #LtcLong
No more manual copy‑paste. The bot publishes itself.
🎬 The Video Logo – Our “Mad Scientist” Character
We also created a consistent video identity: a slightly unhinged inventor in a chaotic workshop.
He appears in short video clips explaining updates, pressing buttons, and showing holographic data.
📌 What’s Next – Following the V7.6 Roadmap
We’re working through the phases defined in V7.6_ROADMAP_DATENVERTRAG.md. Here’s the plan for the next days:
Phase 1: Data Contract Stabilisation (Ongoing)
We already implemented the target schemas.
Now we monitor that writers stick to the rules and no old scripts accidentally patch candidate_log.csv.
Phase 2: Logging Repaired
Next step: ensure every BLOCKED entry has a meaningful reason – especially for q4, which had a known logging bug.
Also extend simulate_outcomes.py to join btc_regime from scan_stats.csv into trade_outcomes.csv.
These are small, isolated fixes (max 2 files per change).
Phase 3: Auswertung Repariert (Analysis Fixed)
We already have the dashboard. Now we’ll align the daily report (daily_report.py) and optimize_filters.py to use the new schema and ignore legacy rows.
Set a “V7.6-valid-from” timestamp to trust only new data.
Phase 4: Audit Fixes
q4 Reason Bug: explicit logging for each block condition.
BTC Regime Join: as mentioned, add regime and scan_id to outcomes.
Startup Reconciliation: verify that open positions are always tracked in the registry (already partially implemented).
Phase 5: More SIM Trades, But Controlled
Let the loosened filters run for 24–48 hours to collect at least 20–30 fresh SIM_TRADEs.
Use the dashboard to see which strategies fire and which blockers remain dominant.
Only then consider small further adjustments – one parameter at a time, with before/after comparison.
Phase 6: Extended Measurement
Once we have a clean data set, we can finally use Sharpe, max drawdown, and walk‑forward validation as planned.
But that’s only after the base is solid.
🛡️ Still No Live Trading
We are still in simulation mode only.
Live trading will only be considered when:
The data pipeline is stable for 7+ days
Auto‑posting works without manual intervention
At least one strategy shows a positive expected value over a statistically meaningful sample (30+ trades)
Bottom line:
Today we shipped a working dashboard, cleaned up the data mess, and enabled auto‑posting to Binance Square.
Now we collect data, close the last audit gaps, and only then think about optimising – and eventually, maybe, live trading.
Built in public. One small step at a time.