Built for punters who want the reasoning

The racing tipping industry runs on confidence, not evidence. Services sell a name and a staking plan, rarely a track record, and almost never the reasoning. If the tip loses, you never find out whether the logic was sound or the process was noise.

PuntPulse takes the opposite approach. We built a database of Australian greyhound racing — years of form — then added thoroughbred racing and UK greyhounds, and trained machine-learning models that rate each runner from those numbers alone. Crucially, the AI tells you why: which parts of a runner's form pushed its rating up or down, how confident the read is, and where it disagrees with the market.

Principles

• Explainable over clever. If we can't show you why the model likes a runner, we don't ship it.
• Probabilities, not tips. You get the shape of the race and decide what to do with it.
• Honest accounting. Every model is tested on races it has never seen, and we say plainly what it hasn't shown — including a profit.
• Responsible by default. Nothing here is advice, and we'd rather you bet less and think more.

What we haven't proven

Beating the market. The models' win chances are useful, but in our own testing they haven't shown a profit after bookmaker margins that we'd stand behind, and we won't claim one until they do. Most runners the model rates highly still lose.

Where it's going

AI ratings for more regions (New Zealand, UK and Irish gallops) and a public accuracy dashboard. The card stays the same — transparent, quick to read, and upfront about uncertainty.