If you're interested in real-world ranking data, here is a list of sources:
FairVote American CVR repository. This repository contains cast vote records (CVRs) for most political American ranked-choice elections since 2004. The repo contains CVRs from statewide ranked-choice elections in Alaska and Maine, as well as local elections in cities such as San Francisco, Minneapolis, Oakland, etc.
Scottish Local Government Elections repository. Scottish council areas have used the method of single transferable vote, a multiwinner version of ranked-choice voting, since 2007. Elections occur every five years. This repository contains ballot data from every Scottish council election from which ballot data is available. Most of the data from 2007 is unavailable, but the repo has a complete record from the 2012, 2017, and 2022 election cycles, as well as data from off-cycle by-elections.
Australia Ranked-Choice Repository. Nick Stephanopoulos has collected CVRs from a few hundred local ranked-choice elections in Australia. The data was first analyzed in his paper Finding Condorcet. The data is mostly available by request.
Condorcet Internet Voting Service Repository. Andrew Myers has released ballot data from thousands of ranked-choice elections he has run on his Condorcet Internet Voting Service website. These elections are non-political, and range from elections held by professional organizations to elect officers to elections held by groups of friends to select a restaurant.
Preflib. A website created by Nicholas Mattei and Simon Rey, and maintained by Nicholas Mattei and Toby Walsh, preflib contains ranking data from political and non-political elections, as well as ranking data from non-electoral contexts (such as tennis world rankings data).
If you're interested in simulating election data or crunching real-world data, the following libraries have been developed recently.
VoteKit. A Python package developed by the MGGG redistricting group, VoteKit allows the user to generated simulated ballot data under models such as the Dirichlet or Plackett-Luce models. VoteKit can also calculate election winners under various winners when given an American-style CVR.
Preferential Voting Tools. A Python library developed by Wes Holliday and Eric Pacuit, Preferential Voting Tools can be used to run and study elections with different preferential methods, as well as some cardinal methods. Like VoteKit, this library has classical models of voter behavior (such as the impartial culture and anonymous impartial culture models) built in, and can handle real-world CVRs.
Single Transferable Vote R Package. Developed by Denis Mollison, this package allows for calculations of winner sets under various forms of STV, with corresponding visualizations. The package is particularly useful for analyzing Scottish council elections.
Pabuviz. Developed by Ulle Endriss and Markus Utke, Pabuviz is a computation and visualization tool for participatory budgeting.