Kifisia held AEK Athens to a 1-1 draw in the shock result of Matchweek 2, while our model's calls on both teams to score took a heavy hit, landing in only two of the seven fixtures. The overall round saw the forecasts post a perfect record on win-or-draw markets but stumble badly when predicting that both sides would find the net.
How our forecasts did
The both-teams-to-score topic proved the round's biggest headache, with just two of the seven calls graded correct. The model had been leaning towards open, goal-filled encounters, but three clean sheets across the card blew those calls wide open. At the other end of the scale, the win-or-draw calls achieved a flawless seven from seven, and the match-result forecasts landed five of seven, underlining the model's solid read on who would take something from each game.
Total goals returned four correct calls from seven, while first team to score landed five. The high-variance exact scoreline and half-time/full-time projections both delivered three correct from seven, which, given the low confidence attached to each of those calls, represents a fair return.
Results
Volos NFC 1-1 Iraklis 1908
Volos were forecast to claim a narrow win, but the contest finished level after both sides scored. The model's call for fewer than 2.5 goals and a Volos win-or-draw both landed, yet the specific predictions of a home victory, a clean sheet for one team and a 1-0 scoreline all missed, with the hosts falling short in a match that the analysis had flagged could open up late — and indeed both goals came in the second half.
Panetolikos 2-0 Kalamata
A firm call for a Panetolikos win proved spot on, and the hosts duly opened the scoring. The model, however, had anticipated a more attacking affair, forecasting both teams to score and more than 2.5 goals, but Kalamata's attack never got going against a Panetolikos side that built on their opening-day 3-1 home win. The exact scoreline call of 2-1 was one goal out, and the half-time/full-time prediction missed because Panetolikos led at the break rather than being level.
Aris Thessalonikis 2-1 OFI
A flawless round for the forecasts here: all seven graded calls came in, including the exact scoreline. Aris edged a contest the model had expected to be tight but goal-filled, with both teams scoring and over 2.5 goals landing, and the home side led at half-time and at the finish.
Atromitos 1-2 PAOK
PAOK completed a comeback win after falling behind early, which scuppered the model's firm call that the visitors would score first. The broader match picture unfolded as projected, however: a winner either way, over 2.5 goals and both teams scoring all landed, and the exact scoreline prediction of 1-2 was correct. The half-time call missed because Atromitos led at the interval, not PAOK.
Asteras Tripolis 0-1 Olympiakos Piraeus
Olympiakos ground out a single-goal victory, matching the model's match-result, exact-scoreline and half-time/full-time calls. The forecast's desire for a livelier contest was misguided, though: the low-confidence calls for both teams to score and over 2.5 goals were undone by a disciplined Olympiakos performance that kept Asteras quiet.
Kifisia 1-1 AEK Athens FC
AEK failed to get the win the model expected, surrendering an early lead at a Kifisia side that continues to draw heavily. The forecast for fewer than 2.5 goals and the call that AEK would score first both landed, but the both-teams-to-score and match-result predictions were incorrect, and the 0-1 scoreline never materialised.
Levadiakos 0-2 Panathinaikos
Panathinaikos cruised to a comfortable away win as predicted, and the model's firm call for the visitors to score first came in. However, Levadiakos could not breach the Panathinaikos defence, so the both-teams-to-score and over-2.5-goals forecasts were wrong, with the contest finishing two goals short of a goal-heavy afternoon.
What we got wrong: the both-teams-to-score collapses
Five of the seven both-teams-to-score calls missed, a swing that points squarely at the model's overestimation of attacking contributions from struggling sides. The analysis for Volos-Iraklis noted a recent Cup meeting that exploded with three late goals, but that pattern did not repeat, leaving the call that at least one side would be kept off the scoresheet incorrect. At Panetolikos, the model leaned on Kalamata's porous opening-day numbers — 19 shots conceded — yet the visitors failed to register, and the hosts kept a clean sheet.
The Kifisia-AEK draw was the headline outcome that bucked the forecasts. The model had expected a low-scoring away win, leaning on Kifisia's tendency to draw and their failure to score in three of five home games. Instead, they found a goal and held AEK, leaving the both-teams-to-score and match-result calls in the red. Even the Olympiakos and Panathinaikos fixtures, where the model got the winner right, saw unexpected shutouts that killed the both-sides-to-score predictions.
What the round changed
Matchweek 2 cemented the look of a division where the top sides are finding their feet but leaving little margin for error. The win-or-draw model went unblemished, and match results were largely on the money, but the round revealed that the model's read on game rhythm and goal distribution still needs refinement, particularly when it leans towards open, end-to-end encounters. With AEK dropping points and Panathinaikos picking up their first win, the early-season picture has tightened, and the forecasts will carry a sharper note of caution on goal-related calls in the weeks ahead.
