# Net-of-Fees Edge — StillOS Calibration Evidence

**Claim under test:** the calibration gap is still positive after exchange fees and the $0.01/call cost.

## Formula (per $1 contract, exploiting an overpriced longshot via a NO position)
```
gross_edge  = (1 - P_true) - (1 - P_implied)         # = P_implied - P_true
fee_per_side = 0.07 * P * (1 - P)
net_edge    = gross_edge - 2*fee_per_side - (0.01 / N_contracts)
```
- `P_implied` = market-implied probability (candle mid).
- `P_true`    = empirically-resolved rate from brier_table.csv.
- `N_contracts` = position size in contracts; the $0.01 call cost amortizes across it.

## Worked example — 1-10% band, 24h horizon (n=369)
- P_implied = 0.038, P_true = 0.019
- gross_edge = 0.038 - 0.019 = **$0.0190** per contract
- fee_per_side = 0.07*0.038*0.962 = $0.00256; round-trip = $0.00512
- net edge @ N=1 contract:  $0.0039
- net edge @ N=10 contracts: $0.0129
- net edge @ N=100 contracts: $0.0138

The $0.01 call cost is the binding cost only at tiny size; above ~$10 notional the exchange fee dominates and the edge stays positive.

## Sensitivity (1 line)
Raising the fee coefficient 0.07 -> 0.10 (+43%) cuts net edge per contract from $0.0129 to $0.0107 (still positive); the edge survives realistic fee schedules.

_Generated 2026-06-27T04:55:00Z by agent stillos (f4d20644-971d-4f20-9322-6219452c2016). Reproduce from brier_table.csv._