where did the ball actually land?
Some urban ranges only see a shot up to the net. Your job: predict where it actually lands - then prove it on shots you've never seen.
submissions close 18 september - stand a chance to win an apple ipad
the brief
what you'll build
01 the data you get
Measured parameters of real data recorded at a range near Stellenbosch, the training set includes the input parameters as well as the landing location and apex we want you to predict.
02 what you predict
On a short urban range the landing and apex would not be measured, for the test set on input parameters you need to predict these, your predictions will be scored on the held-back portion of the data.
03 pick your approach
Physics-based modelling, machine learning, or a hybrid of both - entirely your call. AI-assisted approaches are welcome, just don't burn the budget on tokens or model runs.
04 make it believable
Show how the ball would have flown from launch to landing if no net was present. Show believable bounce and roll for a bit of extra credit.
05 Prove It Generalises
Your predictions are scored against the test set with inputs only - shots your model has never seen the answers to.
06 Write It Up
A sample Writeup submission is provided on Kaggle, add your prediction and code to the writeup and a short summary on what you did as well as any files and links you think is needed.
enter on kaggle
The data, the full spec, and your submission all live on Kaggle. Head over, download what you need, and submit your Writeup when you're ready.
go to kaggle
scoring & rules
what you need to know
scoring
100 points total, judged offline after close - build to generalise, not to memorise.
prediction challenge
full trajectory display
approach
data analysis & viz
novelty
bounce & roll
50 pts
20 pts
10 pts
10 pts
5 pts
5 pts
*Pass/fail check: Writeups must stay under 3,000 words and 25 graphics to be eligible.
rules
  • Open to Stellenbosch University students
  • Individual entries only - no teams
  • External data & tools are fine if free, public, and clearly cited
  • AI is welcome - just disclose how you used it
  • One Kaggle Writeup per team, submitted before the deadline
  • Judged offline after close, winners announced at prize giving
goals
  • Full trajectory - a realistic, animated flight from net to landing
  • Bounce-and-roll physics after impact
  • Generalisation - a model that holds up on shots it's never seen
  • Clear analysis of what the data actually shows
  • Novelty in how you use the data
how to submit
submitting your entry
01
Create a free Kaggle account and join the competition.
02
Download the training data, the test set, and the starter script from the Data tab.
03
Build your model, generate predictions for the test set, and create your submission.
04
Create a Kaggle Writeup - attach your notebook or repo link, your submission, and an optional demo link (video ≤3 min, hosted on YouTube).
05
Select a Track and submit before the deadline. Final submissions only - no drafts.
join the competition & win an apple ipad
Submissions close 18 September 2026.