Worried abouut rising costs of Stockfish used on Lichess

Thanks for your concerns, but our costs are handled right now.

Indeed automated game analysis is CPU intensive but it's mostly run by volunteer computers (lichess.org/help/contribute)

We're all anti-capitalists, anyway!

Oh, so it's a POLITICAL site, too, not just a chess site?

One of those places where one must smoke Gitanes and have long, curly hair in order to hang out? Did everyone arrive on skateboards, surfboards, or in VW beetles... or is there room for a hard-working capitalist here too?!

My ten cents; if it would help the fine ppl who run Lichess to use a slightly weaker - but cheaper - engine, that's fine by me. It wd still be a helluva lot better than me

Stockfish seems perfect in every way. There's no reason to use a different engine.

When I find free time I'll investigate methods for accelerating analysis (esp. for openings and endgames).

There are many ways to improve the speed of analysis and even make it better. For one consider caching the analyzed positions in the database. You can save position, evaluation, stockfish version. And reevaluate only when you change stockfish. Saving position straightforwardly will on average require 65 bytes, so you will end up with less than 80bytes per record. So on a single Redis instance you will be able to have 1bln positions in <80Gb or RAM.

This will save a lot of recomputations especially in the beginning of the game. 1bln positions is really a lot. I highly doubt that in all lichess's life it calculated more than 10bln positions. You do not need to save all of them. Just first 30 moves of every game.

For the totally lost positions there might be no need to calculate position to a high depth. It might be enough to say that you are lost.