Frequently Asked Questions

An AI sports analytics workspace. It turns every game across NFL, NBA, MLB, NHL, soccer, UFC and golf into a breakdown: a win probability derived from the market, an AI-written matchup analysis, team and player statistics, and a graded public record of every read. See the About page for what it is and what it is not.

No. Sportslyx does not accept, hold or pay out money on wagers. The company board is graded at a flat notional stake and your own tracked picks at the stake you enter, purely to measure them — no money changes hands. It is an analytics product for information and entertainment, for adults only.

No tout service: the company board is part of every subscription, published before the start and graded in public. It publishes one read per game, dated and settled, so the record can be checked rather than marketed. Nothing on the site is betting advice.

Adults who want the numbers before the narrative: analytics-first fans, people who keep an honest record of their own reads, and model builders who want to test their own stat weights on live games. It is not for anyone under 18.

Each read starts from the current market line, which is de-vigged into a fair win probability. A per-game package of statistics, form, head-to-head history, injuries and availability is assembled from public feeds, and a language model writes a matchup breakdown around that number. The Methodology page describes every step.

Accuracy varies by sport, market and month, and the honest answer is the public record: the Pick Tracker shows settled reads, wins, losses, pushes, hit rate and ROI at a flat stake for each calendar month, across all sports and per sport. Treat every prediction as a statistical estimate; past results do not predict future results.

It is the market-derived win probability — the moneyline pair with the bookmaker’s margin removed. It is not the language model’s own certainty and it is not a promise; what the model output carries instead is a data-quality tier — rich, partial or weak. The same number appears on the board, on the card and inside the breakdown.

From the official final score, including overtime and shootouts where the market settles the game that way. MLB player props — the only sport with props — are graded from the box score, and a player who did not play is void, not a loss. A soccer draw is a push. Cancelled fights and postponed games are voided and removed. Golf previews are published for information only and are never graded. Picks close 15 minutes after the scheduled start.

One calendar month in UTC, product-wide. Hit rate is wins divided by wins plus losses, with pushes excluded. ROI is the net result at a flat stake divided by total stake. The month a read belongs to is the month it settled in.

A structured read of the matchup: a summary, the case for each side, what the market is pricing, what could move the result, and a closing read tied to the market-derived probability. It is written by a large language model from the per-game package and the line, then validated and cached.

A large language model accessed through OpenRouter, which keeps the product provider-agnostic. The model receives structured inputs and a data-quality tier — rich, partial or weak — that tempers how assertive the read may be.

They are pre-generated for the upcoming slate by scheduled jobs, so a game usually has its read ready before you open it, and then cached. The market-derived percentage on the card follows the line with each board refresh, but the written text is not rewritten when the line moves — a late scratch may not be in the read. A breakdown that fails validation is held briefly and regenerated.

Yes, and it is. Feeds can be late, inputs can be incomplete and language models can assert things that are not in their data. Structured inputs, a fixed brief and output validation reduce the risk without removing it. The AI Sports Analysis page lists where the analysis is strong and where it fails.

NFL, NBA, MLB, NHL, soccer — major European leagues, MLS and the main cups and internationals on the public feed — UFC and golf (PGA Tour, DP World Tour and LIV events). Golf previews are published for information only and are not part of the graded record. An esports section exists behind a feature flag. Each sport has its own hub, predictions board, statistics pages and guides.

The ESPN public scoreboard is the keyless primary source for schedules, scores, base lines and box scores in all seven sports. Optional providers — The Odds API, API-Sports, football-data — add enrichment where configured, a free odds feed supplies lines for some markets, and MLB player props come from the ESPN core endpoint. The Data Sources page lists everything.

Schedules, lines and scores refresh several times a day, most often on game day. Team profiles — name, abbreviation and crest — are refreshed daily; form, head-to-head, lines and injuries are assembled per game, not stored on a profile. Breakdowns are pre-generated ahead of the slate and cached; the text is not rewritten when a line moves, only the percentage on the card follows it.

Sections degrade rather than invent. The per-game package is marked partial or weak, the breakdown is told to hedge, and empty sections show an explanation. Data can be delayed, incomplete or unavailable, and the product says so when it is.

A ticket built from legs in one game — moneyline or spread, total and team totals — priced from a single model of how the score is likely to unfold; player props are covered for MLB only and are not a combo leg. Correlated legs are priced together rather than multiplied as if independent, so the price reflects what the legs are really worth as a set.

A tool for building your own stat-weighted model: choose the statistics you trust, set the weights, run a live game and save the result. Each run is stored next to the market number and the AI read and is graded after the final result, so you can see whether your weights hold up.

Both are available on the Pro plan and above. They share one pool of runs — 20 a month on Pro, unlimited on Unlimited. The 30-day free trial every new account starts with includes 1 Pick Builder run; after it, subscribe to Pro or Unlimited.

Yes. Saved combos and Studio runs are settled by the same jobs that grade the company board and your tracked picks, from official final scores and box scores, and they appear in your monthly record.

Three plans, the same prices for everyone, billed in USD with tax charged on top: Starter at $9 a month or $86.40 a year, Pro at $19 a month or $182.40 a year, Unlimited at $59 a month or $566.40 a year — yearly billing saves 20%. Starter includes 50 AI tokens a month — 5 open every Our Picks call for a day, 1 opens an AI match analysis. Pro includes unlimited AI match analyses and AI chat, up to 10 Our Picks a day, 20 Pick Builder and Parlay Builder runs a month and VIP Picks; Unlimited removes those limits. Full details are on the pricing page.

Create an account — free, no card — and a 30-day trial starts with 10 AI match analyses and 1 Pick Builder run for the whole trial; they do not refill. VIP Picks and same-day Our Picks are not part of the trial — they open on a paid plan. When it ends nothing is charged and nothing renews; to keep going, pick a plan — monthly, or yearly at 20% off, plus tax, cancel any time. One trial per person.

Partly. Without an account you can see the predicted winner and win chance for every upcoming match on the free preview. The public pages — statistics, game pages and the graded track record — are open too. Breakdowns, Our Picks, VIP Picks, Pick Builder, Parlays, Battle and the assistant need a free account.

On Pro and Unlimited, AI match analyses and AI chat are unlimited. Pro meters Our Picks (up to 10 a day) and Builder runs (20 a month for Pick Builder and Parlays together, refilled monthly); Unlimited meters nothing. The free trial has a fixed allowance: 10 AI match analyses + 1 Pick Builder run. Allowances are a usage quota, not currency, and have no cash value.

Upgrades on the same billing period apply immediately: you pay only the difference between the two plans today and your renewal date stays the same. Downgrades take effect at the start of the next period, so the period you have paid for continues on the current plan. Switching from monthly to yearly starts a new yearly period right away with credit for the unused part of the month; switching from yearly to monthly takes effect at the end of the paid year. Plan changes are made from Manage subscription in the avatar menu.

Cancel any time from Manage subscription in the avatar menu or through the Stripe Billing Portal linked there. Cancellation takes effect at the end of the period you have paid for; the free trial has nothing to cancel — it simply ends. Refund rules, including consumer withdrawal rights and charges made in error, are set out in the Terms of Service.

Through the Stripe Billing Portal, opened from Manage subscription in the avatar menu. Card details go directly to Stripe and never reach Sportslyx servers. Invoices and payment history live in the same portal.

Tap “Start free 30-day trial” at the top of any page (“Free trial” on a phone), then register with an email address and a password — or continue with Google — confirm you are 18 or older and accept the Terms of Service and Privacy Policy. Your 30-day free trial starts immediately, with no card — 10 AI match analyses + 1 Pick Builder run — and the plans screen shows Starter, Pro and Unlimited if you want more.

Use the “Forgot password” link on the sign-in page. A one-time reset link is emailed to the address on the account; it stops working as soon as the password is changed.

You can delete your account yourself at any time from Account → Danger zone; deletion signs you out everywhere, closes any active subscription at the end of the paid period and removes your personal data. To request a copy of your data, or to have the account deleted for you, write to support@sportslyx.com.

Email support@sportslyx.com from the address you registered with. For a settlement or data question, include the sport, the date and the game. The Contact page explains what to include for each kind of request.