diff --git a/docs/HELIUS_RATE_LIMITS.md b/docs/HELIUS_RATE_LIMITS.md
new file mode 100644
index 0000000..e771150
--- /dev/null
+++ b/docs/HELIUS_RATE_LIMITS.md
@@ -0,0 +1,146 @@
+# Helius RPC Rate Limits
+
+## Overview
+Our bot uses Helius RPC as the Solana endpoint. Understanding rate limits is critical for reliable operation.
+
+## Helius Rate Limit Tiers
+
+### Free Tier (Current)
+- **Burst:** 100 requests/second
+- **Sustained:** 10 requests/second
+- **Monthly:** 100k requests
+- **Best for:** Testing, small-scale trading
+
+### Paid Tiers
+- **Developer ($49/month):** 50 RPS sustained, 1M requests/month
+- **Professional ($199/month):** 100 RPS sustained, 10M requests/month
+- **Business ($499/month):** 300 RPS sustained, 100M requests/month
+
+## Bot Configuration (Nov 14, 2025)
+
+### Retry Strategy
+```typescript
+// lib/drift/orders.ts
+retryWithBackoff(operation, maxRetries=3, baseDelay=5000)
+// Progression: 5s → 10s → 20s
+```
+
+**Why 5s base delay?**
+- Free tier: 10 req/s = 1 request every 100ms maximum
+- If rate limited, need >1s to clear backlog
+- 5s gives ample time for Helius to recover
+- Prevents cascade of failed retries
+
+### Operations that Hit RPC
+
+**High frequency (every price update):**
+- `getPosition()` - Check current position size (for TP1/TP2 detection)
+
+**Medium frequency (on trade events):**
+- `placeExitOrders()` - Place TP/SL orders (3-6 requests: cancel old + place new)
+- `cancelAllOrders()` - Cancel orders (2-3 requests)
+
+**Low frequency (trade open/close):**
+- `placePerpOrder()` + `confirmTransaction()` - Open position (2 requests)
+- `closePosition()` + `confirmTransaction()` - Close position (2 requests)
+
+### Estimated Daily Usage
+
+**1 trade/hour scenario:**
+- Open: 2 RPC calls
+- Monitoring: ~1 call/10s = 360 calls/hour
+- TP1 hit: 6 calls (cancel + place breakeven orders)
+- Close: 2 calls
+- **Per trade:** ~370 requests
+- **Per day (24 trades):** ~8,880 requests
+
+**Free tier capacity:** 100,000/month = ~3,300/day
+**Headroom:** 3,300 - 8,880 = **DEFICIT of -5,580/day**
+
+## Monitoring
+
+### Check current rate limit health:
+```bash
+curl http://localhost:3001/api/analytics/rate-limits | jq .
+```
+
+### SQL query for rate limit events:
+```sql
+SELECT 
+  "eventType",
+  COUNT(*) as count,
+  MAX("createdAt") as last_occurrence
+FROM "SystemEvent"
+WHERE "eventType" IN ('rate_limit_hit', 'rate_limit_recovered', 'rate_limit_exhausted')
+  AND "createdAt" > NOW() - INTERVAL '24 hours'
+GROUP BY "eventType";
+```
+
+## Warning Signs
+
+### 🔴 Critical (Upgrade Required)
+- `rate_limit_exhausted` events > 0/day
+- `rate_limit_hit` events > 20/day
+- Positions closed late due to rate limits
+- Missing trades due to failed opens
+
+### ⚠️ Warning (Monitor Closely)
+- `rate_limit_hit` events 10-20/day
+- Recovery times > 10 seconds
+- Frequent retries during high volatility
+
+### ✅ Healthy
+- `rate_limit_hit` events < 5/day
+- All recoveries within 2 retries
+- No exhausted events
+
+## Optimization Strategies
+
+### Short-term (Current)
+1. ✅ Increased retry delays (5s base)
+2. ✅ Rate limit-aware error handling
+3. ✅ Database logging for analytics
+
+### Medium-term (If limits become issue)
+1. **Reduce monitoring frequency**:
+   - Current: Every Pyth price update (~1s)
+   - Alternative: Sample every 5s instead
+   
+2. **Batch order operations**:
+   - Combine cancel + place into fewer RPC calls
+   - Use Drift SDK's batch order methods
+
+3. **Cache position data**:
+   - Only call `getPosition()` when price crosses key levels
+   - Trust Position Manager's tracked size between checks
+
+### Long-term (Growth)
+1. **Upgrade to Developer tier ($49/month)**:
+   - 5x sustained capacity (50 RPS vs 10 RPS)
+   - 10x monthly quota (1M vs 100k)
+   - Cost: ~$2/day for reliable trading
+
+2. **Implement request queueing**:
+   - Rate limiter middleware
+   - Priority queue (closes > opens > monitoring)
+   - Automatic backoff when approaching limits
+
+## Cost-Benefit Analysis
+
+**Current situation:**
+- Free tier: $0/month
+- Risk: Missed trades, late closes, orphaned positions
+- Manual intervention required
+
+**Developer tier ($49/month):**
+- Eliminates rate limit issues
+- Reliable trade execution
+- Cost per trade: ~$0.10 (assuming 500 trades/month)
+- ROI: One avoided $5 loss pays for 50 trades worth of RPC cost
+
+**Recommendation:** Upgrade to Developer tier once reaching 200+ trades/month or experiencing frequent rate limit events.
+
+## References
+- [Helius Documentation](https://docs.helius.dev/)
+- [Helius Pricing](https://www.helius.dev/pricing)
+- [Solana RPC Methods](https://docs.solana.com/api/http)