Frequently Asked

Yes. It is the live MeowKuda AI output for the upcoming meeting — the same numbers a subscriber sees, just limited to Race 1. It is generated by the production model from the captured race card, not written by hand.

With Google. There is no password to remember — tap Continue with Google and pick the account your subscription is on. The same works in the app: Google opens in your browser and the signed-in session is handed back to the app.

The whole meeting, not a sample. Every race on the card with the full runner table — rank order, field-calibrated win probability, fair odds, live tote win/place odds, the model-vs-market edge and the last-win-by-condition columns. Plus:

• The MeowKuda AI assistant — ask about a race, a horse or a past meeting.
• The members’ race-day room — talk to other subscribers during a meeting.
• The full form book — every horse, jockey, trainer, owner, sire and dam, each with career records, splits, conditions and the rail configuration they ran under.
• Instant alerts — results, new or revised cards, and room messages.
• The fixture calendar and the race simulator.
• No ads anywhere.

Five tiers, and every one includes exactly the same thing — only the price and the commitment differ: a 7-day free trial, Monthly $50, Quarterly $135, 6 months $240, Annual $420. Billing is handled by Whop and no card details are taken on this site. See Pricing.

Yes. Before the checkout opens you tick a box accepting our Terms & Conditions and Privacy Policy. That acceptance is recorded against your account with the date and the version — and if those documents ever change version, you will be asked again.

A chat assistant that answers questions about a race, a horse or a past meeting, using the same records this site publishes. It is included in every plan. It does not give gambling advice and it will not tell you a horse is a sure thing.

Yes — a race-day room for subscribers, where members talk to each other during a meeting. It is not advice from MeowKuda AI, and its alerts can be switched off in Settings.

Every horse, jockey, trainer, owner, sire and dam on the card is tappable and opens a profile: career record, splits by venue, distance and going, record by class, cup and group races, and the rail configuration each run was under. The core history runs from 2015 and the results archive reaches back to 2000. Where we hold nothing, the profile says so rather than showing a zero.

Yes — /sim replays a meeting race by race.

Yes. Result alerts fire the moment a result is settled, plus alerts when a new meeting’s card is published or revised, and when someone posts in the room. Each has its own switch in Settings, and the app can deliver them with the app closed.

Yes — every published race day, Sha Tin and Happy Valley, with each day’s venue, race count and first-race time. It sits under the meeting card in the app, and in Settings.

In an earlier card-path backtest the ranker's top pick won 41.1% of 56 races. That has not held up live: over the 34 races since, the card's top pick won 6 while the market favourite won 12 — on that stretch we are behind the market. The honest, leak-free AUC is 0.8288 for Happy Valley and 0.8788 for Sha Tin. Those are real model numbers, and they still mean the top pick loses most races.

Because a single “accuracy” number is usually meaningless or invented. Hit rate depends on how many picks you count, at what odds, over which races. We publish the metric we can defend — top-pick winner rate over a fixed set of races — and the baseline next to it, so you can judge for yourself. On our most recent 34 served races the market favourite beat our card, and we say so rather than quoting the older backtest alone.

Edge is the model's win probability minus the market's implied win probability, in points. A positive edge means the model rates the horse higher than the market does; a negative edge means the market is shorter than the model. Both appear on the card; we do not hide the negative ones. Important: Edge is a model output, not a proven edge. Measured on settled races, the model's probabilities are over-confident at the top of the field, so a positive edge is a prompt to look closer, not a signal to bet.

Because the market is not open yet. Cards can be generated before win/place prices are published, and we would rather show a dash than a made-up number. Odds fill in once the market opens, and the card is re-rendered.

Because a placing should read at a glance. The first four finishers get a coloured coin — gold, silver, bronze, then slate — with the placing written inside. The on-screen results deliberately show no medal emoji; the alert that arrives on your phone uses medals instead, because a system notification cannot render the badge.

Not on this site. The model behind this site is a course ranker (Sha Tin and Happy Valley). Other data work exists behind the scenes but is not sold here.

No. MeowKuda AI is a statistical model that publishes its output for entertainment and analysis. It is not financial advice, it does not know your circumstances, and no tip is a guaranteed winner. Never stake money you cannot afford to lose.