Completed results
Scores appear only when the feed supplies identifiable NBA teams, valid status, a matching game date, and plausible scores.
The play before the play. The matchups behind the numbers. A clearer way to understand the NBA — and the evidence behind every prediction.
Start wherever curiosity takes you. Move from the question to the evidence without getting lost in another dashboard.
Explore records, scoring margins, and form when sourced statistics are available. Every figure has a meaning — and every gap in the data stays visible.
The idea behind Aureum / Our Mission
Basketball analysis should invite you in, not leave you guessing what the numbers mean. We turn complicated matchups into understandable research, while making every uncertainty visible.
Explore our principles
A percentage matters only when you understand what it describes. We translate technical measures into useful basketball context, without pretending every number tells the whole story.
Put two franchises on the same page. Compare what their results actually say, with side-by-side context instead of numbers without explanations.
Explore archived team-season research observations from the earlier research project—shooting, efficiency, records, and how the numbers moved between 2024–25 and 2025–26. Historical, not live.

Behind the research
Divit Walia · Founder
I'm interested in what happens when basketball intuition meets evidence. Aureum Tipoff grew from that curiosity: I wanted to explore the game more deeply and make statistical reasoning approachable to anyone who follows the NBA.
My goal isn't to make every forecast sound certain. It's to show what we know, what we're assuming, and what still needs to be tested. The better the evidence, the more meaningful the conversation.
We distinguish completed-game evidence, archived pregame predictions, and adjustable hypothetical scenarios. No invented statistics. No disguised simulations. No false certainty.
Choose two NBA teams and explore how different assumed team strengths affect a hypothetical outcome. These are educational scenarios, not forecasts of real games.
Move the sliders to represent your own hypothetical assessment of each team. The site does not assign ratings based on real-world results.
The demonstration adds a modest fixed home adjustment. The value is an illustrative design choice, not an estimated NBA effect.
Real lineups, validated individual player effects, opponent-adjusted efficiency, schedule, and live injury verification. The Research Data Desk separately tracks an Elo-based score model.
Place two teams side by side. Current-season results and Elo appear only when sourced data is available; the comparison explains each metric and never silently fills gaps.
The results feed hasn't been verified yet. You can still browse teams and experiment with clearly labeled hypothetical scenarios.
Three different kinds of information — always labeled separately.
Good analytics is not just a number. It is a reproducible process, a clear explanation, and a record of what worked—and what didn't.
Interactive, manually supplied team strengths. This teaches how probability responds to assumptions but does not predict actual NBA games.
Real team efficiency, injury, and schedule values are intentionally withheld until they can be sourced, timestamped, and checked.
There is no published game-level win rate, calibration score, or backtest result for this new platform yet.
Let A and B be the manually selected strength scores from 0 to 100. The illustrative away-team probability is calculated as:
The −3 term favors the home team in the demonstration. The divisor 16 controls how sharply the percentage changes. Neither parameter has been fitted to actual NBA games.
Record the provider, season, collection time, and data completeness for team and player information.
Use historical games and features that were genuinely known before each tipoff. Keep later games out of the training window.
Measure Brier score, log loss, calibration, and accuracy against simple baselines. Report the number and dates of evaluated games.
Timestamp forecasts before tipoff, preserve them, and show results alongside the original probabilities.
Our Mission
We believe basketball analytics should be clear enough to question, responsible enough to admit uncertainty, and engaging enough to make you want to learn more.
Meet the founder on the homepage ↗Aureum Tipoff is an independent basketball intelligence project. Its goal is to make advanced basketball concepts approachable while holding predictions to a higher standard of transparency.
The name combines a gold-inspired identity with one of basketball's defining moments: the opening tipoff. The ambition is simple—help fans understand not just what happened, but why.
Explain every metric and why it matters.
Separate verified facts, assumptions, and uncertainty.
Never invent game outcomes, live information, or model performance.
The Gold Standard in Basketball Intelligence
Independent project. Not affiliated with the NBA or its teams. NBA team names are used for identification.
Build a transparent basketball scenario. These controls describe hypothetical conditions, not reported injuries or real team performance.
Saved only in this browser. Nothing is uploaded to a server.
No statistics background required. Search the language of modern basketball, see how each metric is calculated, and learn when it can mislead.
Use team form, scheduled rest, and confirmed player status available at the forecast timestamp.
Archive the exact pregame percentage. Never quietly replace a prediction after the result.
If games assigned 70% happen about 70% of the time, those forecasts are better calibrated.
Thirty stories, one league. Real team performance, explained without jargon.
Points per game is a scoring average, not a measure of possessions. Point differential summarizes scoring margin; it does not by itself explain opponents, player availability, or tactical matchup fit.
What does the evidence actually support? Explore timestamped NBA forecasts when available, then test transparent what-if assumptions without mistaking them for measured player effects.
Each available forecast shows its original recorded timestamp, matchup and modeled probability. Past predictions are never rewritten after results are known.
Start with a team's sourced Elo rating when a validated research snapshot exists. Change which team is home, or model a hypothetical availability setback. This is a sensitivity study, not a verified player-injury prediction.
We don't treat speculative player-impact settings as an injury-aware trained model. Confirmed lineup effects, tactical fit and quantified player impacts require additional data and prospective validation.
Two complete NBA regular seasons. Thirty franchises. See how team performance changed, what the numbers mean, and what they cannot prove.
2024–25 / 2025–26Compare efficiency, shooting, possession control, and results across two completed NBA seasons. Choose a metric, search for a franchise, and explore the year-over-year change. Built for basketball fans—not just statisticians.
These 60 team-season records are an earlier independent research dataset, originally attributed to RealGM advanced team statistics. We've checked internal consistency, not independently verified each number against a fresh third-party download.
Historical trends here are entirely separate from Aureum Tipoff's pending live NBA feed and game-level prediction model.
Loading historical research...
Select any team to explore a detailed comparison.
Season-level measures can describe a franchise's outcomes and style, but they cannot determine the winner of an individual upcoming game. Correlation isn't causation. Changes between seasons may reflect different players, coaches, opponents, injuries, or many other factors.
Historical archive: earlier independent research files, originally attributed to RealGM. This data is not an automatically updated current-season NBA feed, and no injury adjustments or game-level model accuracy are inferred from it.
Explore dated 2026–27 NBA game results and schedules from an attributed third-party scoreboard. Completed scores are distinguished from games still in progress and from future fixtures. We show a feed error rather than an imagined result.
Checking source availability…
Source: ESPN NBA scoreboard. Additional official cross-check: NBA.com Games. Automated per-game independent cross-verification is not yet deployed.
Scores appear only when the feed supplies identifiable NBA teams, valid status, a matching game date, and plausible scores.
No fabricated injury statuses or changes in win probability are applied without a sourced, timestamped player report and a validated model.
Measured accuracy requires predictions recorded before tipoff, a fixed evaluation set, and published Brier score and calibration evidence.
A professional NBA model must distinguish what it could measure from what it has actually measured. This atlas explains categories analysts use without presenting unverified features as current statistics.
Loading factor explanations…