Frequently asked questions.

This is the right question to ask — the industry is full of deleted losses and backdated winners. We built the system so we can't quietly delete a loss or backdate a win. Every live prediction is timestamped the moment it's generated and written to an append-only ledger; database rules reject any edit or deletion. Each day's ledger is then hashed and published to a public record that's timestamped onto the Bitcoin blockchain — an external clock we don't control. The win rates and results you see are computed from that same ledger, losses included.

Three layers. An append-only database stores every prediction with a SHA-256 fingerprint — edits and deletes are rejected at the engine level. Each day the ledger is hashed and published to a public ledger and anchored onto the Bitcoin blockchain via OpenTimestamps — an independent timestamp that doesn't rely on us or GitHub. The alpha record is sealed at github.com/SplitWinner/audit_trail_alpha. The methodology and the verifier are open and public.

The timestamp is external and can't be moved. A live prediction is recorded the moment the morning model run produces it — before that day's games — and that record is anchored publicly. On each pick you'll see an audit receipt with the recorded time. A pick recorded before kickoff is shown as a live call; anything evaluated after the fact is labeled a backtest, plainly. We never present a backtest as a live call.

We publish rolling win rate and ROI in the app, computed from the same immutable ledger that anchors every pick — wins and losses included. Independent research on ML sports models reports accuracy roughly in the 55–68% range depending on sport and market; even a few points above the ~52.4% breakeven at -110 odds can be meaningful over a large sample. We will never claim 80%+ accuracy or guaranteed returns.

No. No honest prediction service can. Sports outcomes carry real variance. SplitWinner identifies a statistical edge — spots where our models' probabilities differ from the market — and any value shows up in the aggregate over many predictions, not in any single pick. Past performance does not guarantee future results.

Free picks are usually one person's opinion, with no verifiable record and every incentive to show only winners. Some tout services even send opposite picks to different groups so half always appear to win. SplitWinner logs every prediction — wins and losses alike — in an append-only ledger that's publicly anchored, rather than relying on screenshots. The track record is computed from that ledger, not curated by hand.

Daily ML predictions with confidence scores across NFL, NBA, MLB, NHL, NCAAF, and NCAAB. An audit receipt on every pick showing when it was recorded and anchored. Insights — win rate, ROI, and performance analytics by model and by sport. SplitWinnerGPT — an AI assistant grounded in your predictions and the model's historical performance. Optional Kalshi integration to trade the model's markets from your own Kalshi account. SplitWinner is an analytics platform, not a picks-in-a-DM service.

Open the app's Today tab. Each prediction shows the model's pick, a confidence score, supporting context, and an audit receipt. Treat it as one input in your decision — not the only one. If you connect Kalshi, you can trade the model's markets from your own Kalshi account, right next to the pick. We don't place trades for you or hold your funds — every order is yours, sent directly to Kalshi, at your own risk.

An AI assistant grounded in your predictions and the model's historical performance. Ask it why a pick looks the way it does, how a sport has been trending, or what the track record shows. It's part of an active Daily Prediction pass.

No. SplitWinner is an analytics and prediction platform — we never accept wagers or hold your funds. The optional Kalshi integration lets you trade event contracts on Kalshi's own platform, a separate CFTC-regulated exchange. We provide the analysis; Kalshi handles execution and custody.

A pick is a person's opinion — "I like the Lakers tonight." A prediction is a model-generated probability with a confidence score, logged and scored against the outcome. No cherry-picking, no story — just results measured over a real sample.

A data pipeline ingests game data from verified sources and feeds a proprietary ML training-and-inference system. Models learn from historical data; predictions are conformal-calibrated against held-out outcomes so the confidence numbers carry a coverage guarantee, not a guess. The fuller breakdown is in the in-app methodology screen, and every production model is fingerprinted in the public ledger.

Historical game data, team and player statistics, schedule context, and other sport-specific features, updated on a recurring schedule. We're transparent about the approach but don't publish the exact features or how they're engineered — that's the part worth protecting.

Not the future — edges. The model looks for spots where its probability diverges from the market's closing line. Independent research has reported ML models reaching roughly 55–68% accuracy across major leagues, above the ~52.4% breakeven at standard -110 odds. The edge is real but modest, and there's no magic. The value is disciplined, positive-expected-value decisions over many predictions — not one lucky weekend.

It happens regularly — no model wins them all. We publish losses next to wins because that's the only way trust is earned, and because the ledger makes hiding them impossible anyway. The value is in the aggregate: small, consistent edges over many predictions, if they hold up. Judge it by the published record.

As little as possible. You sign in with Apple or Google, so we never see or store a password. We don't collect your SSN, banking details, or government ID. If you connect Kalshi, your Kalshi API key never leaves your device — it's stored in the device's secure keychain and used to sign requests locally. It is never sent to, logged by, or stored on SplitWinner's servers.

Several layers, on by default. The connection to our API is certificate-pinned, so a forged or intercepted certificate is rejected. Requests are cryptographically signed, so they can't be forged or tampered with in transit. The app checks for device-tampering signals (such as jailbreaking) and locks paid content on a compromised device. Crash and error reports never attach screenshots or your data in the released app. And the integrity tables behind your predictions are append-only at the database level — edits and deletes are structurally rejected.

No. We don't sell your personal data, and we don't share it for anyone else's advertising. We rely on a small set of standard service providers — for sign-in, payments, analytics, and crash reporting — purely to operate the app.

No. Your predictions, activity, and account details are private. If you connect Kalshi, your positions, balances, and trade history are visible only to you.

One drop. A pass unlocks that day's prediction in full — plus SplitWinnerGPT and the rest of the paid surface — through the next 5 AM ET rollover. It's sold as one prediction, not 24 hours. The window always ends at 5 AM ET, after even late West Coast games have settled, so you can see your slate's results. A late-night buyer pays for the same single drop.

That depends on your volume. A few points of edge only turn into meaningful returns at scale, so SplitWinner is built for serious bettors, not casual $5 parlays. You're paying for the whole surface — the prediction, the verifiable record, the analytics, and the AI assistant — not just a number. Review the track record before deciding.

Passes are purchased through the App Store or Google Play, so refunds are handled by Apple or Google under their policies — not by SplitWinner directly. We'd encourage you to review the track record and the methodology screen before buying.

SplitWinner is an information and analytics platform, not a sportsbook — we don't accept wagers or hold funds. Providing sports analysis is an information service. SplitWinner itself isn't gambling; it's data-driven analysis, and how you use it is your decision. If you choose to trade on Kalshi, that happens on Kalshi — a CFTC-regulated exchange where those products are event contracts. This isn't legal advice; rules vary by location.

Whether you can trade depends on Kalshi, which operates under federal CFTC oversight and is available in most US states, though some restrict certain contract types. Using SplitWinner's analysis isn't placing a bet. Check Kalshi's website for current state-by-state availability.

SplitWinner doesn't give tax advice. In the US, prediction-market profits and betting winnings are generally taxable income, and Kalshi provides its users with tax forms (such as a 1099). Consult a qualified tax professional for your situation.

SplitWinner isn't a sportsbook and doesn't place trades or hold your money — if you trade, it's through your own Kalshi account, where you set your own limits, and every order is yours at your own risk. We show analysis, not pressure: no manufactured urgency, no guaranteed winners, no upsells. Bet only what you can afford to lose, and if it stops being fun, step away.

If you or someone you know is struggling, contact the National Council on Problem Gambling at 1-800-522-4700 (call or text, 24/7) or visit ncpgambling.org. Most states also offer self-exclusion programs through their gaming commissions.