Strategy Board
Plenty of sites show you a backtested ROI as if it were live. A backtest can be re-fitted until it looks good — that is why we killed ~50 of our own models. The experiments below run the hard way: rules frozen first, then measured forward on live prices, wins and losses alike. The one replay on this page, value signals at detection, is labelled as what it is: a backtest.
The Italian Draw (TIPS-EXP-001)
Starting · 2026-27Thesis: the draw is systematically underpriced in Serie A and Serie B. Rules frozen and hashed before the 2026-27 season: back the X at Pinnacle opening odds ≤ 3.50, flat stakes, CLV measured same-book. Success and failure criteria declared in advance — the result gets published either way.
Pre-registered · SHA-256 + OpenTimestamps · starts with the 2026-27 season
Full details →AutoBet at Betfair prices (TIPS-EXP-005) — earlier series closed
Running · 2026-07 →Thesis: value signals can be executed at real Betfair Exchange prices once the artefacts are removed by rule (fresh fair price, within 3 h of kick-off, liquidity ≥500). Current series TIPS-EXP-005. v1 (TIPS-EXP-002) ran 262 bets: +0.8% flat-stake, CLV +2.5% vs the devigged close (corrected from a published +6.5%), with a public autopsy. Gate: 150+ bets with flat ROI above zero.
111 settled simulated bets · 46 won · ROI -11.6% on 1054 units staked
Avg execution edge vs live sharp benchmark: +5% (112 bets)
v1 concluded: 262 bets, flat-stake ROI +0.8%, CLV +2.5% vs devigged close. Read the full autopsy →
Full details →Value flags at soft-bookmaker prices (TIPS-EXP-006)
Decided · 2026-10Thesis: the site's public value flags can be taken at the soft bookmaker that showed the price, at a price that beats the devigged Pinnacle close. Paper test of every flag from registration, entry at the first price of the same bookmaker we observe within 60 minutes after the flag. Decision at 300 takeable flags or 31 December 2026.
Decision (17 Oct 2026, 08:25 UTC, first 300 takeable flags): inconclusive; average CLV +0.57% (95% interval -0.50% to +1.64%).
After the decision point (descriptive, not part of the test): 913 more takeable flags settled, average CLV at the bookmaker's price +1.08% (95% interval +0.43% to +1.73%).
Full details →TIPS-EXP-007 · Value on goal markets (paper)
Running · 2026-10 →Soft-bookmaker prices on over/under and both teams to score at least 5% above Pinnacle's fair price on the same line, measured against the devigged close. Recorded on paper, listed after kick-off; decision at 300 flags with a close or 31 March 2027.
Full details →Value signals at detection (backtest)
Backtest · 2026-06 →Thesis: the value flags shown on this site are worth something. This is a backtest, not a live record: every settled match since June 2026 is replayed on the prices we had stored, using only prices live at that moment, the first time the rule (best soft price above the devigged sharp benchmark, edge ≥ 2.5%) would have fired, flat stakes. Re-run weekly. It is a stricter subset than the track-record backtest (all flags from 0.5%), so its numbers differ.
5445 settled signals · 1906 won · flat-stake ROI +0.4% (± 4.4 pp at 95%) · avg CLV +2% vs the devigged close
Correction, 24 September 2026: until then this panel scored each signal at the highest soft price ever seen for that match, not the price available when the flag fired. That overstated the result (it showed +7.3% ROI and +5.0% CLV). The replayed figures above replace it.
Full details →Small samples mean wide error bars — a positive ROI inside its confidence interval is a hypothesis, not a proof; we say so on every card. Full methodology, raw data and external timestamps are public. Where a strategy fails, the card stays up: the graveyard is part of the record.
