Current work / Data as an experience
After the TipOne clock.
Two ways to listen.
I wanted to hear how a whole game unfolds—not browse another catalogue of scores.
I redirected this NCAA women’s basketball project toward two averaged soundtracks: the screened season and Purdue’s games. Compare them at the same point in regulation, with both teams in the picture.
See and hear the comparison- My contribution
- Product and sound direction
- Development
- Codex-assisted · SVG / Web Audio
- Status
- Local exploratory prototype · not submitted
The experience / 40 minutes on one shared scale
Keep the shape of the game.
The current prototype separates each side’s lead contribution rather than letting opposite leaders cancel in a single average.
Mean positive lead contribution · points per game
Mean positive lead contribution · points per game
Actual 2025–26 aggregate profiles, redrawn for this case study. Purdue’s 30 games are included in the 6,002-game season. Home/Away and Purdue/Opponent are different orientations—not a direct comparison of team strength.
Optional listening / Lead voices
Change the cohort.
Keep your place.
Press play, then switch between the season and Purdue. Both excerpts run on the same regulation clock, compressed into one minute.
Synthesized data, not a game recording. The lower voice represents the first side and the brighter voice the second; amplitude follows mean lead contribution. The pitches stay fixed in this mode. Start at a comfortable volume.
Understanding the two ribbons
At each ten-second snapshot, a side contributes its positive lead when ahead and zero when tied or behind. The total is divided by every game in the cohort. Both ribbons can therefore appear at once: they summarize different games with different leaders. The lower ribbon is drawn below zero to separate the sides; its values are positive contributions, not negative scores.
These are descriptive averages, not confidence intervals, signed mean margins, measures of excitement or evidence of psychological momentum. Regulation only; no overtime. The source retains 34 nonzero opening-clock margins, and final-score checks do not establish every timestamp’s accuracy.
| Elapsed | Home team | Away team |
|---|---|---|
| 10 min | 4.59 | 2.25 |
| 20 min | 7.65 | 3.27 |
| 30 min | 10.63 | 4.19 |
| 40 min | 12.77 | 4.63 |
| Elapsed | Purdue | Opponent |
|---|---|---|
| 10 min | 2.70 | 5.40 |
| 20 min | 6.73 | 8.83 |
| 30 min | 8.13 | 10.80 |
| 40 min | 8.90 | 11.80 |
My decisions / Less catalogue, more comparison
The question changed the interface.
“All games” did not mean thousands of things to click. It meant one season-wide experience to compare with Purdue.
An early direction expanded into a game catalogue. I clarified what I actually wanted: two averaged soundtracks. That moved the shared-clock comparison to the front and left individual games as a secondary explorer.
I also chose to include both teams’ scoring in the Purdue average. A Purdue game is a contest, not an isolated offensive stream. That choice kept the listening question focused on the whole game.
A separate sound-design iteration
Put the gap into the note,
not another competing layer.
I found single-game sonification underwhelming and asked to broaden the experience. Later, I paused a separate deficit pulse and explored scoring bops whose pitch rises and falls with the score gap.
Those newer 15-second Q2 auditions remain separate from the current browser modes and the fixed-pitch lead-voice excerpt above. Whether people can distinguish the teams and understand the comparison by ear still needs listening feedback.
What this work shows
I shape the question
and the experience.
This project brings together analytical framing, interface direction and sound exploration. My contribution is choosing what the comparison should mean, then refining how someone encounters it through Codex-assisted development.
Next: listening feedback, clearer team distinction and a decision about which sound approach belongs in the final experience. This is exploratory work for Purdue’s data-visualization challenge—not a submitted entry, award or proven listener outcome.
Sources and scope
The official competition dataset is distributed from SportsDataverse / wehoop ESPN records. The source prototype includes 6,002 of 6,030 listed completed 2025–26 games: 19 lack play-by-play and nine fail scoring checks. All 30 listed Purdue games are included. Coverage is screened, not an independently verified census.
This page is an excerpt: redrawn lead figures and two existing 40× audio exports. It does not reproduce the full application’s crossfade, scoring-rhythm mode, speed controls, per-game explorer or shareable session state. No underlying game dataset is offered for download.
Purdue challenge and portfolio guidance (opens in a new tab)