distributed-stockfish v0.1.0

DICEY CLI

Before you get started, make sure that you have access to a backend deployment URL (see Backend README). TL;DR only requires running a simple Bash script if you have your own AWS account. You can either store that URL in an environment variable as described on that page or pass it as a URL via the --url option on the CLI.

Now, ensure that you have Node JS 14+ installed. Next, you can install this package from NPM with npm i -g distributed-stockfish. Then you can run it like so (may require starting a new terminal session); note the many optional parameters:

dicey eval [--fen "fenStr"] [--depth 54] [--chunkDepth 13] [--parallelism 5]
[--lookahead 6] [--branch 3] [--url backendAPIUrl] [--quiet]

A quick overview of the optional optional:

• fen: A FEN string for the position to evaluate. Defaults to the starting position.
• depth: The depth that you want to evaluate at. Dicey supports really big depths -- like 60+!
• chunkDepth: The depth of an individual chunk. This should be a reasonably large fraction of depth
• to reduce the total chunks you are processing. Also, the sum chunkDepth + lookahead should hover around 15 - 20 for best accuracy without causing individual nodes to take a long time to process.
• parallelism: This lever goes allllll the way up! Technically depends on the Lambda concurrency settings on your account, but probably can support several hundred parallel processes without a problem. (But: watch that AWS bill)
• branch: How many principal variations to show, also the branch factor for each partial variation. As this goes up you may see better accuracy but also an exponential increase in nodes, so be careful!
• lookahead: Ideally around 5 + 6, helps with accuracy. Again, recommend keeping the sum chunkDepth + lookahead at around 15 - 20.
• quiet: See less output

Required:

• url: Your personal Dicey Backend URL, see Backend README. You can also supply this as an environment variable via the instructions there.