September 2026 cricket review: upsets, models, market

How the month went

September 2026 was a month of overlapping T20 storylines: the Caribbean Premier League ran alongside a chaos-heavy Women’s CPL, while South Africa’s CSA T20 Challenge ticked along with fewer shocks. The standout moments weren’t subtle — they were innings and spells that blew games open in a handful of overs, the sort of swings that make pre-match prices look silly after the fact.

The loudest statement came from Dasun Shanaka, who detonated 121 off 43 balls (SR 281.4, 12 sixes) for St Kitts and Nevis Patriots against Saint Lucia Kings. With Andre Fletcher adding 64 off 38 and a 93-run stand involving Shanaka, the Patriots didn’t just win — they won by 137 runs, and Jeremiah Louis finished the job with 5 for 10 in 2.5 overs. In the same competition, Shimron Hetmyer’s 102 off 45 (SR 226.7) for Jamaica Kingsmen against Guyana Amazon Warriors underlined how quickly totals could be chased down or made irrelevant, even when the finish was tight (Kingsmen home by 5 wickets with 2 balls remaining).

If the CPL supplied the fireworks, the WCPL supplied the volatility. Guyana Amazon Warriors W repeatedly ended up on the right side of close finishes and low totals: a 1-wicket win with 0 ball remaining (D/L method) went Jamaica Empress W’s way after a 95-run partnership between Realeanna Grimmond (45 off 46) and Shemaine Campbelle (57 off 33), and later Guyana edged Trinbago Knight Riders W by 2 wickets with 3 balls remaining. The CSA T20 Challenge had fewer headline shocks, but when it produced a batting burst it was emphatic: Sinethemba Qeshile’s 77 off 28 (SR 275.0, 8 sixes) powered Titans to a 5-wicket win with 31 balls remaining against Western Province.

The upsets

The month’s biggest market miss by margin was also the most brutal on the field. Saint Lucia Kings went off at 1.7491 against St Kitts and Nevis Patriots (2.0436), and still lost by 137 runs. Shanaka’s 121 off 43 set the tone, and Louis’ 5 for 10 ensured there was no late correction — it was a one-way game built on an extreme powerplay-to-death squeeze that the pre-match line simply didn’t price in.

Several of the most instructive upsets came from teams winning as clear outsiders and doing it fast. Antigua & Barbuda Falcons were 2.35 away to Guyana Amazon Warriors (1.57) and won by 9 wickets with 88 balls remaining after Guyana were held to 76; the bowling pair of Sufiyan Muqeem (4 for 13) and Shadab Khan (4 for 16) made it a collapse rather than a normal loss. Falcons also beat Guyana again earlier at 2.5289 (Guyana 1.52), this time by 65 runs — two results in the same matchup that read less like coin-flip variance and more like a tactical or matchup edge that the market didn’t fully respect in that window.

Jamaica Kingsmen’s 9-wicket win with 35 balls remaining over Barbados Tridents was another clean example. The Tridents were priced 1.6773, Kingsmen 2.1636, but Maaz Sadaqat’s 112 off 49 (SR 228.6) turned it into a chase with daylight, backed by a 138-run stand involving Sadaqat. In the WCPL, the most pronounced price upset was Guyana Amazon Warriors W at 2.75 beating Trinbago Knight Riders W (1.37) by 25 runs — a game defined by pressure and control rather than one huge score, with Guyana logging 55% dot balls and Shabnim Ismail returning 3 for 21.

South Africa’s lone flagged upset in the period came when North West Dragons (2.05) beat Warriors (1.65) by 7 wickets with 27 balls remaining; Wihan Lubbe’s 3 for 9 was the kind of early damage that makes even modest chases comfortable.

Where the methods landed

Across the 34 settled matches, the bookmaker benchmark still set the pace on pure hit-rate: 58% (19/33). None of the four forecasting methods matched that accuracy, with AI 2 the best of them at 50% (17/34), followed by AI 1 at 44% (15/34), Maths at 41% (14/34), and ML at 38% (13/34).

Profitability was harsh across the board — and it’s a useful reminder that accuracy and ROI are not the same thing, especially when a method (or the market) is frequently aligned with short prices. Even with the best accuracy, the bookmakers finished at -8.0% ROI (-265). All four methods were deeper in the red: AI 2 at -19.1% (-650), AI 1 at -27.4% (-930), Maths at -30.1% (-1024), and ML at -40.0% (-1360).

The most revealing pattern in the featured fixtures was how often everyone agreed with the market — and still got punished by the same surprises. In multiple major upsets (Patriots over Kings; Kingsmen over Tridents; Falcons over Guyana at 1.57/2.35; Dragons over Warriors), every method shown sided with the favourite, mirroring the market’s view. When disagreement did appear, it was notable: in Trinbago Knight Riders vs Antigua & Barbuda Falcons (Falcons won by 8 runs at 2.16), three methods (AI 1, AI 2, Maths) took the outsider correctly while ML and the bookmakers stayed with Trinbago at 1.6778. In the WCPL, the models were also more willing than the market to lean into Jamaica Empress W in the 1-wicket D/L finish — all four methods picked the Empress while the bookmakers favoured Guyana Amazon Warriors W at 1.8.

What it means

September’s data reads like a stress test for any favourite-leaning approach: the CPL produced 7 upsets in 17 matches, and the WCPL went even further with 6 in 8. That kind of distribution doesn’t just lower accuracy; it also punishes portfolios that concentrate stake on short prices, because the occasional miss is expensive and the wins don’t pay much back.

At the same time, the month also showed how narrow the edge can be even for the benchmark: the bookmakers led on accuracy, yet still finished negative on ROI. The practical takeaway is simple and non-glamorous: in volatile T20 environments, being “more right” is not automatically the same as being “paid more”, and the most important differences often show up in the small set of matches where you’re willing to oppose the market — and whether those calls are repeatable beyond a single month’s noise.