Every sports slate on ThoroughWiz can carry one parlay, built only from that slate's own filed picks. Parlays are the most popular format in the book and the worst-priced one, and this guide exists so that ours are built under a published construction instead of a habit. The line that rides wherever our combos display:
The old construction, and what it cost
The first construction (Combos v1, then v1.1) built a three-leg cross-game parlay from the slate's strongest picks — and when qualifying legs ran short, it filled the remaining slots with moneyline favorites. It also built same-game parlays: a game's spread pick and total pick riding together. Both habits felt reasonable. Neither was tested.
So we tested them. The construction backtest replayed the lab's filed-pick machinery over its own history, and the harness first had to prove itself: it reproduced the lab's published corrected baseline exactly, in every era block, before a single parlay number was read. On that record, the old construction returned −2.04% ROI at recorded prices across 5,599 parlays. The extreme version — a parlay of every filed pick on the slate — returned −24.81%. Forcing legs to complete a ticket was the most expensive habit in the file.
Combos v1.2 — the construction now
- Cross-game only. Same-game parlays are prohibited and the form is retired: two legs from one game are correlated by construction — a game's spread and moneyline name the same outcome, and its spread and total share the same script.
- Exactly three legs, one per game.
- A leg qualifies only on strength and price: its recorded strength — the differential test's margin over the filed line — must be at least 2.0 points, and its filed price must sit between −120 and +100.
- The three strongest qualifying legs file. Fewer than three qualify, no parlay files, and the reason is recorded on the slate.
- Moneyline fillers are abolished. A leg is never forced, padded, or borrowed from another market to complete a ticket.
- The Rule v1.3 spots are never legs. Combos read only the differential spread and the NFL wind under.
On the same backtest, this construction returned +2.36% ROI at recorded prices (+2.17% at flat prices) across 3,930 parlays — and it tested no worse than the old construction in any league, and better in all four:
| League | Combos v1.2 | Old construction |
|---|---|---|
| NFL | −23.92% ROI (434 parlays) | −36.09% (444) |
| NBA | −5.85% (1,594) | −7.49% (2,932) |
| College football | −2.87% (395) | −6.46% (526) |
| College basketball | +19.98% (1,507) | +17.67% (1,697) |
The flag on that positive number
Read the pooled +2.36% the way the rule file reads it: college basketball supplies more than 100% of the pooled profit — the other three leagues are negative — so the pooled figure is presumed a one-league concentration, not a demonstrated edge. The construction was adopted because it is the least-bad tested construction in every league, not because the pooled line is positive. The same legs bet straight returned −1.72% in the test: the parlay form itself still costs against its own legs in three of four leagues. All of that stays printed next to the picks.
How a combo settles
A combo is built once per slate, before any of its games start, and is never rebuilt or restated. It wins only if every leg wins. A leg that pushes drops out and the combo reprices on the remaining legs at their filed prices — the standard book rule; if every leg pushes, the combo is a push. Combos settle from the same official finals as the single picks, into their own ledger, and never mix into the single-pick records.
Keep reading
How the Sports Lab Files a Pick →
The Sports Book — today's boards, picks and parlays →
Sources: the sports lab's Combos v1.2 amendment and its construction backtest record (Combos v1.2, 2026-11-08). The honest line at the top of this guide is the lab's own, verbatim.