TrollFit.
Fit better data.
Data analysis and curve fitting for experimental work — estimate uncertainties, compare models and create publication-ready plots without writing code.
- The account stores no email or name
- Runs in your browser
- Your data stays on your computer
Better fits. Clearer results.
No trolling. Just fitting.
Fit experimental data, estimate uncertainties,
compare models and create publication-ready plots —
without writing a single line of code.
No bloated software. No unnecessary complexity.
Just powerful tools for scientific data.
Free during the beta · Access by invitation · Nothing to install
Everything you need for experimental data
Powerful tools. Simple interface. Reliable results.
Curve fitting
Linear, polynomial of any degree, exponential, power, logarithmic, rational, 1–3 Gaussian peaks, Lorentz, sigmoid, Michaelis–Menten, Fourier series — or your own equation.
Compare models
Fit every model at once and see them ranked by AICc with Akaike weights — no more guessing which curve the data really support.
Uncertainties done right
Weighted fits with σy, χ²/dof and parameter correlations. Errors in x too — orthogonal distance regression, verified against SciPy.
Automatic fit check
Flags outliers, systematic residuals and non-normal errors, and warns about strongly correlated parameters. Exclude an outlier with one click.
Confidence & prediction bands
95 % bands for the curve and for new measurements. Evaluate the fit x → y or back y → x, always with its uncertainty.
Limits & global fits
Fix parameters or bound them to a range. Fit several datasets together with shared parameters and test whether sharing holds.
Digitize graphs
Read data straight from a picture — even an old black-and-white scan. Snap to markers, find all similar ones, or trace a curve.
Integrals & derivatives
Area under a fit between any limits, tangent and slope at any point, and fits combined with +, − or ×.
Your data, your way
Excel, ODS, CSV or a paste from any spreadsheet. Transform columns (ln, 1/x, units) with uncertainties carried along. Up to five datasets.
Publication-ready export
SVG, PDF, PNG or TIFF at the exact journal size — residual panel, fit results, annotations, themes and TeX labels, straight into your paper or thesis.
Private by design
Everything runs in your browser. Your measurements never leave your computer — nothing to install, nothing uploaded.
Learn it in minutes
A guided tour walks through a whole analysis, and every button has an animated hint. Sample datasets to play with.
Looking for something specific? See the complete feature list →
Figures, straight from the app
Every image below was fitted and exported by TrollFit itself. No touch-ups in other software.
Overlapping peaks on a sloped background
A double Gaussian separates two Raman bands. The tilted baseline is fitted too and drawn dashed, and the residual panel shows nothing is left over.
Radioactive decay, weighted fit
Counts with Poisson uncertainties σ = √N. A true weighted exponential fit on a log axis, with the half-life marked and the fit box showing χ²/ν.
Calibration curve and an unknown sample
A linear calibration with confidence and prediction bands. The unknown concentration is read back from its absorbance, including its uncertainty.
Enzyme kinetics, global fit
Two Michaelis–Menten datasets fitted together with a shared Vmax. Colour-blind safe colours and different marker shapes keep it readable in print.
Errors in x and y
Ohm’s law with uncertainties on both axes, fitted by orthogonal distance regression. Normalized residuals sit below the plot, typeset like a LaTeX thesis.
Periodic signal for a slide
Three days of temperature fitted with a Fourier series. Its period and harmonics are found automatically, and the nights are shaded for a dark presentation slide.
Every figure saves as vector SVG or PDF, or as PNG and TIFF up to 1200 dpi, at the exact size your journal asks for. All export options →
Nothing is a black box.
Every model, estimator and test in TrollFit is written out in a complete mathematical reference — so every number the app shows can be checked by hand.
Reference manual
The maths behind every tool, with the formulas the code implements, the constants it uses and how it is validated against SciPy and statsmodels.
- Nonlinear least squares, weights & uncertainties
- Errors in x: orthogonal distance regression
- Model comparison (AICc) & fit diagnostics
- Parameter limits & global fits
- Peaks, Fourier series & custom equations
- Graph digitizer & numerical validation
Free during the beta
The whole of TrollFit is free while it is in beta. For now the beta is by invitation — ask for an account number by email. After the beta TrollFit will become a paid subscription, and we want to be upfront about it.
- Every feature free, no limits — now
- New features and fixes with every release
- Paid subscription after the beta, announced well in advance
- Never charged unless you choose a plan yourself
- Your data and results always stay yours
Used TrollFit in your work?
Please cite it with the version you used — it helps others reproduce your analysis. Appreciated, never required.
Svoboda, Š. (2026). TrollFit: data analysis and curve fitting (Version 0.5.0) [Computer software]. Troll Softworks. https://trollsoftworks.com
@software{trollfit,
author = {Svoboda, {\v S}imon},
title = {TrollFit: Data Analysis and Curve Fitting},
version = {0.5.0},
year = {2026},
publisher = {Troll Softworks},
url = {https://trollsoftworks.com}
}
For specific methods you can also cite the reference manual — its appendix A has both entries and an example methods paragraph.
Good software should get out of your way.
Troll Softworks is an independent software studio — small for now, but built to grow — making thoughtful tools for people who care about doing science properly.
Focused
We build software around real problems.
Thoughtful
Powerful features without unnecessary complexity.
Built to last
No hype. No shortcuts. Just good software.
Hi, I’m Šimon — founder of Troll Softworks.
Troll Softworks started as my personal project, and TrollFit is its first product — so far designed, written and tested by me. I started it because fitting data was always more painful than it needed to be: tools that are either too simple to trust or too complicated to enjoy. So I built the one I wanted to use — careful with the maths, honest about uncertainties, and pleasant to work with.
There are no investors and no ad money behind it — just a lot of test cases and the people who send feedback. If something is wrong, missing or could be better, write to me — every message is read by the person who will fix it.
simon.svoboda@trollsoftworks.com →Want to try TrollFit?
TrollFit is in a closed beta. Send us a short email — who you are and what you would like to use it for — and we will send you an account number. Students and researchers are especially welcome. It is free during the beta, and your data never leaves your computer.
No mail app? Write to info@trollsoftworks.com. Bug reports and ideas are welcome there too — every message gets a real reply.