2026

Picks whoever's higher on the ladder. Home team wins ties. It's the dumbest predictor here on purpose — if the fancy ones can't beat this, something's wrong.

Looks at each team's last 10 games and counts the wins, weighting recent ones a bit more. A win's a win, whether you scraped through by 1 or smashed them by 40. Tiny home bias to break the close ones.

Same idea as Recent Form but cares about how you won. A 40-point flogging counts way more than a 2-point thriller, and beating the Storm matters more than beating the Tigers.

Looks at how many points a team's scored vs conceded across the season and runs it through the Pythagorean formula (yes, the one from baseball). Teams that score plenty but keep losing close games get a bit of love here.

Same Pythagorean idea but only the last 8 games. So a team that started 0-4 and has now won 5 straight finally looks good. The exponent was tuned on 5 years of NRL data instead of borrowed from the NFL.

Standard Elo, like the chess one. Every team has a rating, you win and you go up, lose and you go down. Bigger wins move you more, and beating a stronger team moves you more too. Home teams get a +70 bonus, and ratings drift back toward the middle each new season to handle roster turnover.

Knows who's actually running out. Takes the team lists off nrl.com and judges each player on what they've actually done lately: run metres, line breaks, tackle breaks, tackles, errors and so on from the nrl.com match stats, weighted toward their last few games. A small ML model, retrained every round on past games only, works out which of those stats really win matches.

Our own player Elo, built from nrl.com match stats back to 2021. Every player has a rating and a side is the average of its 17. When a team wins, everyone who played goes up (more for more minutes), but players who outscored the usual fantasy points for their position take a bigger share of the credit than teammates who went missing. Then it averages the named side.

A simple ML model. Feed it three numbers — form difference, scoring difference, ladder gap — and it figures out which combinations have historically meant which team wins. Gets retrained for every round so it can't cheat by peeking at future games.

What the market reckons, converted from decimal odds (with the vig stripped out). Current week's odds come from nrl.com, historical ones from aussportsbetting.com. Bookies do this for a living and it shows — over a full season this is usually the single best predictor on its own.

The big one. Blends three predictors — Bookies, Player Form and Player Elo — weighted by how useful each has been lately, into a single pick. Back-testing showed adding the rest only added noise. The percentage is calibrated against history, so 70% really means it wins about 7 times in 10 — picks at 70%+ get the ★ Banker badge, 60–70% is a Lean, and anything under 60% is honestly labelled a Coin flip. If you've got no strong gut feeling on a match, just tip whatever this says.