Bug: Board endpoint player rating at start of match changes during game

Documentation at https://www.chess.com/news/view/published-data-api#pubapi-endpoint-match-board
is quite specific: "rating": 1492, // the player's rating at the start of the game ...

I would like to use the player ELO rating at the start of a match (with other factors) to try to forecast potential outcomes (e.g. multiple regression). Currently ratings associated with match boards are changing through the game. It is the starting player ELO at the time the match starts, but then the API returns the rating that is current at the time of the request. When the game finishes, the rating that is kept is the rating of the players at the end of the game.

After exchanging ideas with @skelos (he has reported this as documentation mismatch at https://www.chess.com/clubs/forum/view/in-progress-match-board-has-diverged-from-the-documentation), but this seems to be a bug as:
1) documentation says it is ELO rating at the start of the match
2) that seems to be the preferable option (not sure there is a use case to support changing player ratings throughout the game.
3) while it might be interesting to have the player rating at the end of the game (as it happens now when the game finishes), that could be another piece of information.

In conclusion, I would expect the rating to reflect the player rating at the start of the game, and stay that way whether the game is in progress or finished.

Proof:
Match started on Juy 15th, when I had 1860 ELO daily, on July 22nd I lost a game and got to 1832, board API has current (as of July 22nd) or last ELO, and that will continue to change until the game finishes, then the ELO would not longer change and reflect my ELO at the end of the game (not the start of the game). https://api.chess.com/pub/match/913368/71

I understand rating history would be a future endpoint to enable point in time rating recovery or historical analysis, but still I would prefer the board endpoint rating to stay as of the start of the match. I understand last ELO might be more useful for some purposes, but I can recover that from the player stats.

It would also be nice to add the rating at the end of the game.

PD: Thans to @skelos for his valuable feedback!

A comment: chess.com uses the Glicko rating system, not Elo. There are papers and 'Net posts available. I only found one with a worked example for predicting results (as opposed to adjusting ratings post-events) but as you've mentioned multiple regression I'll be very very quiet now. I doubt you need maths help.

It's confusing that the ratings are labelled "Elo" in the PGN files, but I think that's per the PGN specification and the best compromise available.

You are right. References are inconsistent, but it should be Glicko. That's why in the long run having access to historical ratings would be quite interesting, as we could retrieve also the RD at the start of the match...

Ah, yes. If I am using the starting rating for any calculations, I want the RD that went with it. Otherwise I won't get consistent results.

(Whether the consistent results would be any good is a worthy question, but I have a few predictions for matches saved here and there and will be interested to see how close they come when the matches finish.)