Altora Analytics
About us
Where data meets the edge, and why we publish everything, including the failures.
What Altora is
Altora Analytics is an independent, self-funded analytics project, founded in 2026. We build statistical models for eight sports (football, tennis, darts, snooker, rugby, esports, basketball and handball) and run systematic trading models across FX, equity indices and commodities. There's no tipster network behind this, no marketing agency, and no team of "senior traders". There are models, a public ledger, and the discipline connecting them.
The exact counts live on the front page. They are generated from the actual data files on every deploy, so they can't drift from the truth.
Who's behind this
I'm Sam Carter. Altora is my project, built and run by me, with a lot of automation doing the daily legwork.
It started with frustration. I signed up to tipsters like everyone else, and saw how the game works from the paying side: winners paraded, losers deleted, "97% win rate" screenshots, edited history everywhere. So I built my own football model instead, partly to see if I could do better, mostly because I wanted numbers I could actually trust. I enjoyed building it more than I expected. Once it was running honestly, with every pick logged and every loss kept, the obvious question was whether the same discipline could carry across a whole suite of sports, and into systematic trading. That's Altora.
I'm here to try and prove one, out in the open.
Every prediction is logged at the odds available when it was published, settled against the real result, and kept forever. That includes the losses, and the bets my own filters said not to back. And no model keeps its place because it used to work: when one stops performing, it has to earn its way back through fresh validation, exactly like a new idea would. If the models are good, the record will show it. If they're not, you'll see that too. That's the point.
How we work
- Models earn their place. Nothing goes live until it beats the incumbent on held-out data: both halves of the sample, under the exact configuration we serve. Ideas that fail are published in the research log alongside the ones that shipped.
- The record is frozen. Once a pick is logged it cannot be quietly edited. If we ever correct something before a match starts, the old row is voided visibly and the correction is logged as a new entry.
- The market is respected. We publish our probability next to the market's price. When our model disagrees wildly with a sharp market, we withhold the pick rather than pretend we've found magic. That gap is usually a data problem, not an edge.
- We measure ourselves against the sharpest price in the business: the close. Every pick is logged at the earlier, softer opening price and then compared against where the market finished. The aim is to be early enough to beat the open; the closing line is the referee that tells us whether we did. Both prices go in the record, so you don't have to take our word for it.
- The models are never finished. They retrain on schedule as new results arrive, and new ideas are constantly tested against the incumbents. An upgrade only ships when it proves itself on data it hasn't seen, and the ones that fail are logged too. What you see today should be measurably better in six months, and the record will show whether it is.
- Automation handles the routine. The pipelines run every day, fetching, predicting, settling and publishing, with health checks that alert us when anything silently breaks. That includes one that verifies no settled row has ever been altered.
The methodology page carries the full detail: data sources, model families, validation gates and known weaknesses, including the parts that don't flatter us.
What we're not
We're not tipsters, and nothing here is betting advice or financial advice. We publish calibrated model output with a verifiable track record. What you do with it is your decision and your responsibility. Past performance proves the process is honest; it doesn't promise future profit. If someone tells you their model guarantees profits, they're selling something. We'd rather show you the ledger.
Get in touch
Questions, corrections, or something on the site that doesn't add up? Use the contact form. Corrections are genuinely welcome. The fastest way to make this project better is to catch us being wrong.