Home › About
About BetterBet
Who writes this, what it measures, and what it refuses to claim.
Who writes it
Johnathan Cruzen-Myer — Analysis and modelling, BetterBet. I build the models, run the backtests and write the pages. There is no editorial team and no contributor network; one person is responsible for every figure on this site, which is the main reason the methodology is published in as much detail as it is.
I am not a licensed financial adviser and nothing here is financial or betting advice. I do not sell tips, I do not take affiliate commission from bookmakers, and I do not publish a record I have not measured.
What the site is
Historical results, closing lines, totals, period scores and player statistics for the NRL, AFL, NFL and NBA, with a model for each market and an honest account of whether it works. Roughly 150 pages, built on eleven seasons of data.
The distinguishing feature is not the predictions. It is that every claim is tested the same way and published whichever way it falls — including, repeatedly, the finding that something does not work.
The standard every claim has to meet
- Walk-forward only. Anything used to price a game comes from games played before it. No model is ever scored on data it was fitted on.
- A matched baseline, never 50%. A trend is compared against the same teams' or players' own rate over exactly the games the claim fired on. Comparing to a coin flip flags every good team.
- Every test counted. Test ninety-six things at a 5% threshold and about five will clear it on noise. The number of tests run is published next to the number that passed.
- Out-of-sample or it does not count. A finding is split in half. If the second half does not look like the first, there was nothing there.
- The break-even stated. 52.4% at standard pricing, not 50%. A 51% trend is a losing bet at a number that looks like a winning one.
Things this site has found that do not work
Published as prominently as anything else, because a site that only reports its wins has not tested anything:
- No NRL referee produces a bettable trend — 4 of 96 tests cleared significance against 4.8 expected by chance, and none survived out-of-sample.
- Situational trends across four codes beat their matched baseline by −2.1 to −3.0 points.
- Backing hot streaks does not work in the NRL, AFL or NFL.
- The Best Bets board returns between −0.8% and −7.5% on real closing prices.
- Heavy favourites win 88–96% in the biggest spread bands and cover only 46–57%.
Mistakes, and how they are handled
Several published figures have been wrong and were corrected once found. A lookahead bug in the prop model overstated one market by 40 points; quarter totals were graded against a benchmark that was wrong by a third; a confidence figure was printed as "100% chance" because of a numerical underflow. Each is documented in the code at the point it occurred, and the corrected figures replaced the originals rather than sitting alongside them.
Data sources
- NRL and AFL results, closing odds, lines and totals — aussportsbetting.com
- NFL and NBA results, lines, totals, period scores and officials — sportsdatabase.com
- Player box scores, rosters and injury reports — ESPN's public site API
- Live odds — The Odds API
- NRL match detail — rugbyleagueproject.org; AFL results — afltables.com
Contact
Corrections are welcome and acted on. If a figure here is wrong, it should be fixed.
Informational only, not betting advice · 18+ · Gamble responsibly. If gambling is causing harm, Gambling Help Online is available on 1800 858 858.