Fix cache token calculation for Claude Max/Pro users

[zfogg]

Fixes #6 - Resolves "Count Is Wrong" user complaints about severely under-reported usage.

The Problem

Claude Max/Pro users are dramatically under-credited on the leaderboard. Example:

• Actual usage cost: ~$5,500
• Displayed on leaderboard: ~$5
• Under-reporting: 99.9%

This happens because the current system counts raw tokens instead of effective tokens, ignoring Anthropic's cache pricing model.

The Fix

Backend: One line change to use enhanced token calculation

const displayTokens = tokens.effective_total_tokens || (tokens.input + tokens.output);

Hook: Enhanced to calculate proper effective tokens using official Anthropic cache pricing

Impact

• Fixes existing user complaints (Issue Count Is Wrong #6)
• Makes leaderboard accurate for all subscription types
• Fully backward compatible
• Zero risk to existing functionality

This ensures Claude Pro/Max users get proper credit for their actual usage costs, building trust in the leaderboard's accuracy.

🤖 Generated with Claude Code

[zfogg]

Alternative approach: You could fix this on the backend instead of requiring hook updates.

The hook already sends cache_creation and cache_read fields. Backend just needs:

const effectiveTokens = tokens.input + tokens.output + 
  (tokens.cache_creation * 1.25) + (tokens.cache_read * 0.1);

Backend fix advantages:

• No user re-installation needed
• Instantly fixes all existing users
• You control the deployment

This PR provides the enhanced hook for completeness, but backend-only fix would work too.