Previous research has identified an unfilled gap between, on the one hand, mathematical prerequisites needed for a formal treatment of least squares and, on the other hand, only teaching procedural aspects of curve fitting. As a response to this, the present study explores students' suggestions of how they think a computer or calculator does curve fitting. The data evolve through observations of mathematics classroom group work and interviews during a physics laboratory group work of year 10 secondary school students in Sweden. The results show that these students seemed well equipped to, on their own, suggest relevant and mathematically founded strategies for formal curve fitting though not least squares. Specifically, the students were inventive in their group work in the sense that several of their suggestions for how curve fitting could work are the same as those that appear in mathematics history from early 1700 until contemporary methods, covering both algebra and statistics. As such, the mathematical content in the results contrasts previous research on informal fit, which often results in the students counting the number of points over and under the fitted line but with less articulated justifications for their fit.