Regarding the normalization range of RMSD

[Asecretboy]

I noticed that in the PR(#223 (comment)), it mentioned the RMSD normalization range as [0, 0.03], but in the code, it’s actually normalized to [0, 0.3] (

matbench-discovery/site/src/lib/metrics.ts

Line 141 in a48c373

const baseline = 0.3 // in Å, a reasonable baseline for poor performance given worst performing model at time of writing is AlphaNet-MPTrj at 0.0227 Å

). I’m not sure if this is a bug or if the correct normalization range is indeed [0, 0.3].

[janosh]

that's a typo, thanks for catching that! 🙏