Model live Fixtures68 Avg λ2.69 O2.5 ≥55%23 Value prices22 In play0 30-day hit66% 08:41 WAT18+
Wednesday 18 November 2026

Yesterday's predictions, graded

129 fixtures with 1X2 probabilities, expected goals and the model's selection. Tap a match for its score grid. Graded so far: 81 right · 46 wrong.

Tips by · xGaura analystModel refreshed 19 Nov 2026, 08:41 WAT · Settled publicly on the track record

Final scores

NigeriaNPFLxG1X2O2.5BTTSPick · EV

BrazilSerie AxG1X2O2.5BTTSPick · EV

SpainCopa Del ReyxG1X2O2.5BTTSPick · EV

WorldFriendliesxG1X2O2.5BTTSPick · EV

ArgentinaTorneo Federal AxG1X2O2.5BTTSPick · EV

ColombiaCopa ColombiaxG1X2O2.5BTTSPick · EV

North & Central AmericaCONCACAF Nations LeaguexG1X2O2.5BTTSPick · EV

PanamaLiga Panameña De FútbolxG1X2O2.5BTTSPick · EV

Turkey2. LigxG1X2O2.5BTTSPick · EV

UkraineU19 LeaguexG1X2O2.5BTTSPick · EV

South-KoreaWK LeaguexG1X2O2.5BTTSPick · EV

WorldASEAN Club ChampionshipxG1X2O2.5BTTSPick · EV

PolandCentral Youth LeaguexG1X2O2.5BTTSPick · EV

AlbaniaCupxG1X2O2.5BTTSPick · EV

SerbiaCupxG1X2O2.5BTTSPick · EV

BulgariaThird League NortheastxG1X2O2.5BTTSPick · EV

BulgariaThird League SouthwestxG1X2O2.5BTTSPick · EV

SloveniaCupxG1X2O2.5BTTSPick · EV

Yesterday in figures

Wednesday 18 November 2026: 129 fixtures in 49 competitions. Busiest: Turkey 2. Lig (15), Spain Tercera División RFEF Group 13 (7), Ukraine U19 League (6). Selections by market: 4 1X2, 4 double chance, 85 goal lines, 36 BTTS. Thirty-day accuracy 66% from 8,047 graded picks (results).

How to read each row

  • Kick-off shows in your local time; live games show the minute in red.
  • xG: expected goals for the home side (top) and away side (bottom).
  • 1 · X · 2: home, draw and away probabilities in per cent – the darker the cell, the higher the number; the outlined cell is the likeliest result.
  • O2.5 · BTTS: probability of three or more goals, and of both teams scoring.
  • Pick · EV: the model's selection, its probability, the best odds and the expected value (probability × odds − 1).
  • After the final whistle the pick is marked right or wrong on the 90-minute score.

Looking for price edges instead? Open the value bet index. League scoring rates are on league statistics, ratings in the xG database.

xGaura research desk2 min read

Yesterday's predictions and results: learning from the scoreboard

This page keeps Wednesday 18 November's full board exactly as it stood at kick-off, with each pick now graded on the 90-minute score. It is the least glamorous page on the site and one of the most useful, because reviewing finished forecasts is how you learn whether a model – or your own judgement – actually works.

Why we never edit a graded pick

Every probability you see here was published before the match started. Once the game kicks off, the row is frozen: no quiet deletions of losing picks, no retrospective changes to the selection. That rule is what makes the track record meaningful. A tipster who can tidy up the past can make any record look good.

Reading a day of results properly

A single day is a tiny sample. On a typical weekday the model might grade sixty picks; even if every probability were perfectly accurate, the hit rate could land anywhere between the low fifties and the high sixties purely through chance. So do not ask "did it win yesterday?" Ask better questions:

  • Did picks rated 70% or more land roughly seven times in ten?
  • Were the losses concentrated in one league or one market, which might point to a weakness, or spread evenly, which points to normal variance?
  • Did the picks with positive expected value return a profit, even if the overall hit rate was ordinary?

Expected goals versus actual goals

Each finished match still shows its pre-match expected goals. Compare them with the final score. A 2.4 v 0.6 projection that ended 0–1 is not proof the model was wrong – the favoured side may have hit the woodwork three times. But if the actual goals regularly fall short of the projections in one competition, the league's scoring average may have shifted, and the ratings will absorb that within a few rounds.

The difference between a bad pick and a bad result

A bad pick is one where the probability was overstated or the price offered no value. A bad result is a sound pick that happened to lose. Over a handful of games they look identical; over a few hundred they separate clearly. The calibration page is designed to show which of the two you are looking at, because it compares stated probabilities with observed frequencies across thousands of forecasts.

Using yesterday to improve tomorrow

Keep a simple log. Note the matches you would have backed, the odds you could have taken, and whether the closing price moved towards or away from your selection. Closing-line movement is a powerful signal: if the market consistently moves your way after you bet, your reading is probably sharper than average, regardless of short-term wins and losses. The bankroll tracker and ROI calculator help you keep that log honest.

Where to go next

The full archive of graded picks sits on the results page. If you want to see how individual markets have performed over thirty or ninety days, the leaderboard ranks them by profit at level stakes.

Frequently asked questions

How many fixtures are modelled yesterday?

129 – every match with bookmaker odds, from the top five leagues and UEFA competitions down to second tiers worldwide. Each has its own match model page.

What do the 1, X and 2 cells mean?

Home, draw and away probabilities in per cent. They add up to 100; the outlined cell is the most likely result.

What is EV?

Expected value: probability × odds − 1. Positive EV means the odds pay more than the risk, on average.

What is the xG line?

Expected goals for each side – the inputs to the score grid. Higher totals favour overs and BTTS; a big gap between the two favours the stronger team.

How is the model pick chosen?

From every market, the one with the best mix of probability and price. Its percentage is the model's probability and the @ figure is the best bookmaker odds.