Dataset v0.24.0 · 2026-10-18 · doi:10.5281/zenodo.23217259
sim-highline
Put your simulated galaxy relation next to every published simulation and observation, and see which of those comparisons are fair. Two curves that look different may measure different things: a different stellar mass aperture, IMF or population cut. Every curve here carries the definitions that decide it.
Since your last visit the site has changed (now v0.24.0). What is new
Start here
- Turn the physics dials
- Two simulated universes, IllustrisTNG and SIMBA, each with four feedback dials. Turn them and watch galaxies reshape; ask whether one universe can be made to look like the other, and see which properties stay apart.
- 2,054 simulations · in the browser
- Look up a relation
- Pick a relation and an epoch: the published simulations and observations on one plot, each labelled with its source (official table, paper table, fit or digitized figure) and whether its definitions allow comparison with the others.
- 2,041 curves · in the browser
- Check your own run
- Paste your curve, say what it measures, and get offsets against the comparable published curves, as a table and a CSV. Nothing leaves your browser; there is an example to try first.
- stays in the browser
- Take the data
- Tidy CSV and JSON of every curve with its definitions, BibTeX for what you plot, and a Python loader (
pip install .). Cite the dataset and the original papers. - CC BY 4.0
Try this. What do simulations predict for galaxies at z ≈ 8? · How bright are galaxies in the UV at z ≈ 8, the quantity JWST measures? · How heavy are the black holes in different simulations? · What if supernova winds were twice as strong in both universes? · Is my own curve consistent with the published ones?
- If you run simulations
- Start with Check your own run: your curve against everyone else's, and which comparisons are fair.
- If you need predictions to compare with data
- Start with Look up a relation: simulations and observations on one plot, each with its source and definitions.
- If you are learning how feedback shapes galaxies
- Start with the dials, then read the glossary.
How far to trust it. Not reviewed by any outside expert. 55% of simulation curves are read from published figures by script and are never used for ranking. Where an author or collaboration releases a table, use that. This is not a leaderboard. Methods, limits has the details; to report an error or ask for a takedown, open an issue on the repository.
What is in it
| Relation | Records | Simulations | Observations | Rankable | z-range |
|---|---|---|---|---|---|
| Stellar mass function gsmf | 365 | 17 | 12 | 217 | 0 – 15 |
| Mass–size relation size | 214 | 10 | 5 | 67 | 0 – 10 |
| Gas-phase mass–metallicity relation mzr | 163 | 12 | 14 | 142 | 0 – 13 |
| Stellar–halo mass relation shmr | 148 | 15 | 4 | 60 | 0 – 17 |
| Star-forming main sequence sfms | 132 | 11 | 10 | 90 | 0 – 17 |
| Quenched fraction quenched | 124 | 6 | 4 | 32 | 0 – 7 |
| Black hole–stellar mass relation bh | 115 | 12 | 3 | 24 | 0 – 11 |
| UV luminosity function uvlf | 114 | 7 | 5 | 51 | 2.1 – 16 |
| Stellar mass–stellar metallicity relation zstar | 83 | 8 | 2 | 40 | 0 – 10 |
| Specific SFR ssfr | 65 | 10 | 3 | 25 | 0 – 13 |
| 34 further relations (all) | 518 | 21 | 61 | 192 | 0 – 15 |
| All 44 relation codes (the Explorer groups them into 44 relations) | 2,041 | 22 | 117 | 940 | 0 – 17 |
Simulations and observations count distinct sources; records count curves. Rankable: definitions and sampling allow the curve to be used for ranking. The full table, evidence tiers, files and licences are on the Data page.
What it has shown
The dataset and the CAMELS ensemble (2,054 simulations of two feedback models) were used to ask six questions. Where a study is marked pre-registered, its decision rule was written down before the outcome was computed. Most results are null or qualified, and the pages say so. All studies and how they are registered.
| Study | Question | Status | Headline |
|---|---|---|---|
| How much is just definitions? | How much of the disagreement between simulations is just definitions? | Result | 4% of 2,567 cross-simulation pairs fully matched |
| Can matching definitions close the gap? | Can documented definitions close the gap between simulated z ≈ 0 mass functions? | Pre-registered · result | R = 0.14 (95% CI −0.89 to 0.51), p = 0.08: not detected |
| What can retuning change in cosmic star formation? | Do simulations disagree more in the early Universe, and can retuning remove the difference between two codes? | Pre-registered · result | Funnel not detected, factor 2.2; retuning: 0.22 to 0.08 dex (removable) |
| Can two universes agree on everything at once? | Can retuning make two feedback models agree on every galaxy observable at once? | Pre-registered · result | Joint D = 4.8 (structural); 2 of 11 families structural alone |
| Does the shape of star formation tell the codes apart? | Does the share of galaxies in the green valley, and the width of the star-forming sequence, separate the two codes beyond the quenched fraction and the main sequence? | Pre-registered · result | shape alone removable; added to the usual two: adds nothing detectable; added to everything: adds nothing detectable |
| How much does a model's output change with resolution? | How far do published curves of one model move when only the resolution changes, measured in CAMELS noise units and against the distance between IllustrisTNG and SIMBA? | Pre-registered · result | 28 of 30 same-box steps change more than the CAMELS noise; 16 of 19 are at least the TNG–SIMBA distance |