Jugend forscht computer science project mentoring

Source: Stiftung Jugend forscht, Alterssparten and Jugend forscht junior pages (jugend-forscht.de), read on 3 November 2026. In a group, the age of the oldest member decides the category. In Mecklenburg-Vorpommern and the Saarland the junior category has regional contests only.

This rule is a real difference from the Bundeswettbewerb Künstliche Intelligenz, which accepts projects in English, and from the Bundeswettbewerb Informatik, where the programming language is free. A pupil whose strongest language is English may find one of those a better first research contest, and can come to Jugend forscht later.

import random
from statistics import median

def days_until_all_know(size, rng):
    knows = {0}
    days = 0
    while len(knows) < size:
        days += 1
        told = {rng.randrange(size) for pupil in knows}
        knows |= told
    return days

rng = random.Random(2027)
for size in [10, 20, 40, 80]:
    runs = sorted(days_until_all_know(size, rng) for _ in range(1000))
    print(size, "pupils: median", median(runs), "days, middle 90%", runs[50], "to", runs[949])

Simulation: our own model and code, seeded with random.Random(2027), run with Python 3 on 3 November 2026. The model is deliberately simple; real classes do not choose names at random.

The hypothesis is wrong, and that is the most useful outcome a young researcher can have. Doubling the class from 10 to 80 pupils, three doublings, adds about two days each time instead of doubling the time. The number of people who know grows faster and faster at first, roughly doubling each day, so a class twice as large needs only about one or two more days. A Jugend forscht write-up that states the wrong hypothesis, shows the data that refuted it, explains why, and names the limits of the model is a far stronger piece of science than one that shows a single run that happened to look impressive.

1. Repeat it. Run the code with a different seed. Which numbers change, and which stay the same? What does that tell you about 1,000 runs?
2. Change one thing. Let each pupil tell two names a day. Predict the new medians before running it, and keep everything else fixed.
3. Find the limit. Real pupils talk mostly to friends. How would you change the model to test that, and what data would you need?
4. Write it up. In your own language first, then in the language of your contest, write the question, the hypothesis, the method and one honest sentence about what the model cannot show.

Germany runs on UTC+1 in winter and UTC+2 in summer, as the Einheitenzeitgesetz provides; winter time began on 25 October 2026 and summer time resumes on 28 March 2027. India keeps UTC+5:30 the whole year, which is four and a half hours ahead of German winter time and three and a half ahead of summer time.

These are illustrations, not fixed batch times. Before the January deadline many one-to-one pupils meet weekly; after it, the focus moves to presentation practice for February.