Every edge is logged as a virtual bet at the line shown and notionally sized, then settled against the real result and measured for closing-line value. Each line runs on its own bankroll so it can be validated separately. Tap a card to dig in.
A DIRECTIONAL line, not a value scan. The other strategies search every market for the biggest edge; this one takes the single result the model's own predicted scoreline implies — 2-1 backs the home side, 1-1 backs the draw — and backs it on the 1X2 line. Because the selection is fixed, the thing being tested is the SIZING: quarter-Kelly on that pick's edge against the offered price, so a strong, well-priced call is staked many times a weak one. A pick the market prices better than the model still gets recorded, at a token carry stake, so the record shows every call the rule made and not just its flattering half. Unvalidated — never staked or recommended.
How bad the worst losing run got, and how long it took to climb out — the betting analogue of a trading drawdown report.
| Market | Bets | Won | Staked | P&L | ROI | Avg CLV |
|---|---|---|---|---|---|---|
| 1X2 | 0 | 0 | $0.00 | +$0.00 | — | -7.4% |
| Total | 0 | 0 | $0.00 | +$0.00 | — | -7.4% |
| Status | Selection | Odds | CLV | Stake | P&L |
|---|---|---|---|---|---|
| void | Bal Khalsa 1X2 | 2.50 | -7.4% | $50.00 | +$0.00 |
