# Strategy C — Home picks via highest B365 CLV

**Discovered:** 2026-05-16
**Status:** Backtested only, not yet live

## Rule
For each PRE snapshot, compute B365 CLV for all 3 outcomes (H/D/A).
If `clv_b365_home` is the highest among the three, bet HOME at the Kalshi ask.

## Backtest results (flat $10 stake, 386 games in live_data corpus)
| Snapshot window | N bets | WR | P&L | ROI |
|---|---|---|---|---|
| -200 min | 134 | 53.0% | +$69.38 | +5.8% |
| -480 min | 153 | 56.2% | +$137.10 | +10.0% |
| -720 min | 135 | 54.1% | +$69.35 | +6.1% |

WR is consistent across windows (~54%). Earlier snapshots slightly stronger.

## How this compares to current strategy (Kalshi CLV > 5% only)
- Current strategy at -480 min: 93 bets, 22.6% WR, +24.7% ROI (small bets, big % return)
- Strategy C: ~150 bets, 54% WR, +10% ROI (larger flat bets, smaller % return)
- These are *complementary*: current strategy = small away/underdog bets, Strategy C = home favorites/medium-priced bets

## Caveats
- Layering Strategy C on top of Kalshi CLV > 5% (Strategy B) makes it WORSE — combining filters reduces sample too aggressively
- Flat $10 staking used; Kelly sizing not yet tested
- Draw picks via B365 CLV give ~26% WR (no edge), away picks give 30-37% WR (marginal)
- The HOME signal is the only strong one — don't generalize "bet highest B365 CLV" across all outcomes

## Open questions to revisit
1. Kelly sizing on Strategy C — does it beat flat $10?
2. Does ROI hold across leagues, or is it concentrated in specific ones?
3. What CLV magnitude threshold (if any) further improves WR?
4. Does -480 min hold up with more data, or is it sample-size noise?
5. Can we combine A + C as two separate concurrent strategies (not as a filter)?