Arbitrage Enhancements:
1. Updated cost calculation:
Real bid price = Southern bid – export tariffs – transport fees – storage fee
(Storage fee accounts for overnight holding before conversion)
2. Added one-sided market making:
⦁ Example: Local bid/ask at 1008/1014 vs. Southern real bid/ask at 1001/1009
⦁ Place local ask at 1011 → if filled, convert next day at 1009 → +2 profit
⦁ Current arb bot v2 yield: ~3.8k (parameter-adjusted, guarding against overfitting)
Climate Modeling Approach:
Sunlight Prediction:
1. Calculate accumulated sunlight:
sum(rate(t) × timestep)
(timestep = 12/10000 since 12 daylight hours ÷ 10k timestamps)
2. Project remaining sunlight:
remaining_hours × current_rate(t)
3. Total estimate:
If (accumulated + projected) < 17,500 (7h × 2500 baseline) → SHORT
Humidity Strategy:
⦁ 60-80 range: No action
⦁ >80:
If humidity ↑ over last 200 steps → LONG (yield drop expected)
If humidity ↓ → SHORT (recovery anticipated)
⦁ <60:
If humidity ↓ → LONG (further scarcity)
If humidity ↑ → SHORT (replenishment coming)
Current Limitations:
⦁ Backtest data covers only 1/10 of a day
⦁ Shorting glitches cause PnL spikes
⦁ Untestable in current system
Call for Expertise:
How would you recommend modeling sunlight/humidity effects more effectively?
