The attached files calculate Elo using the USCF formula.
It won't be identical to the value calculated by bayeselo or elostat, but it will answer fairly nearly.
I guess that you can encode the math in a spreadsheet without too much trouble.
I guess that you already know it, since you are a pretty sharp fellow, but Elo is a relative number, so that Elo of one pool is not equivalent to the Elo that another pool of contestants have.
I explain what I wanted to do. Compare progress ELO because I saw a big jump from +37 to +39 in the last moments of the test. And so I wanted to identify the responsible lot of this jump.
If there's a way diverted, for example, the way to generate PGN files for each batch with only the results. I could then directly use the Rémi's tool