ootrstats
is a tool for generating various statistics on the Ocarina of Time randomizer by generating a large number of seeds and analyzing them.
-
Install Rust:
- On Windows, download and run rustup-init.exe and follow its instructions. If asked to install Visual C++ prerequisites, use the “Quick install via the Visual Studio Community installer” option. You can uncheck the option to launch Visual Studio when done.
- On other platforms, please see the Rust website for instructions.
-
Open a command line:
- On Windows, right-click the start button, then click “Terminal”, “Windows PowerShell”, or “Command Prompt”.
- On other platforms, look for an app named “Terminal” or similar.
-
In the command line, run the following command. Depending on your computer, this may take a while; you can configure ootrstats while it's running.
cargo install --git=https://github.com/fenhl/ootrstats --branch=main ootrstats-supervisor
- Install the updater:
cargo install cargo-update
- Update
ootrstats
(and everything else installed viacargo install
):cargo install-update --all --git
ootrstats
requires a configuration file, which should be a JSON object located at $XDG_CONFIG_DIRS/ootrstats.json
on Unix, or %APPDATA%\Fenhl\ootrstats\config\config.json
on Windows. It takes the following required entry:
workers
: An array of worker configurations. You should specify at least one worker so seeds can be rolled. While ootrstats tries to make full use of all workers in the list, the order of the list serves as a priority (workers defined earlier in the list are preferred) in tiebreaker situations, such as when rolling fewer seeds than there are workers.
And the following optional entries:
log
: Iftrue
, the supervisor will create a text file namedootrstats.log
in the working directory with debug info. The default isfalse
.statsDir
: A path to a directory where the statistics will be stored. Defaults to$XDG_DATA_DIRS/ootrstats
on Unix, or%APPDATA%\Fenhl\ootrstats\data
on Windows.
A worker is a computer that rolls seeds. Each worker configuration is a JSON object with the following required entries:
name
: A string which will be displayed on the progress display on the command line, as well as in error messages.kind
: One of the section headers listed below.
And the following optional entry:
bench
: Iffalse
, this worker is skipped when thebench
subcommand is used. The default istrue
.
Depending on the kind
, there are additional entries:
A worker that runs on the same computer as the supervisor program. It takes the following additional required entry:
baseRomPath
: An absolute path to the vanilla OoT rom. See the randomizer's documentation for details.
And the following optional entries:
wslBaseRomPath
: An absolute path to the vanilla OoT rom that will be used if the randomizer is run inside WSL, i.e. for thebench
subcommand if this worker is running on Windows. Defaults tobaseRomPath
if not specified.cores
: The maximum number of instances of the randomizer to run in parallel. If this number is 0 or negative, it will be added to the number of available CPU cores, e.g. if 6 cores are detected and a number of-2
is given, 4 cores will be used. Defaults to-1
.
A worker that listens to WebSocket connections from the supervisor. To set up, do the following on the worker computer:
- Install Rust
- Run
cargo install --git=https://github.com/fenhl/ootrstats --branch=main ootrstats-worker-daemon
- Create a JSON file at
$XDG_CONFIG_DIRS/ootrstats-worker-daemon.json
on Unix or%APPDATA%\Fenhl\ootrstats\config\worker-daemon.json
on Windows, containing a JSON object with the following entries:password
(required): A password string that the supervisor will use to connect to this worker.address
(optional): The IP address on which the worker daemon will listen. Defaults to127.0.0.1
, meaning only local connections will be accepted and you will need a reverse proxy like nginx. Change to0.0.0.0
to accept connections from anywhere.
- Start the worker daemon, e.g. by editing
assets/ootrstats-worker.service
inside a clone of this repository to adjust the username, then runningsudo systemctl enable --now assets/ootrstats-worker.service
if the worker is on a Linux distro that uses systemd. - Make the worker daemon reachable from the network, e.g. by enabling (an edited copy of)
assets/ootrstats.fenhl.net.nginx
if nginx is installed on the worker or by configuring theaddress
.
The worker configuration on the supervisor takes the following additional required entries:
hostname
: The hostname or IP address of the worker, optionally with the port after a:
separator (port defaults to 443 iftls
istrue
, or to 80 otherwise).password
: The password from step 3 of the worker setup described above.baseRomPath
: An absolute path to the vanilla OoT rom on the worker computer. See the randomizer's documentation for details.
And the following optional entries:
tls
: Whether to use a secure WebSocket connection. If enabled, the worker needs a TLS certificate. The default istrue
.wslBaseRomPath
: An absolute path to the vanilla OoT rom that will be used if the randomizer is run inside WSL, i.e. for thebench
subcommand if this worker is running on Windows. Defaults tobaseRomPath
if not specified.priorityUsers
: A list of usernames. The worker will not start rolling any new seeds while any of these users are signed in. Only supported by workers running on Windows.
On the command line, run the ootrstats-supervisor
command, followed by any options you would like to change from their defaults, followed by an optional subcommand. If no subcommand is given, the supervisor will simply generate the spoiler/error logs, place them in the statsDir
, and display the path where they were placed.
On Windows, it is recommended to use PowerShell to run ootrstats-supervisor
, as the legacy Windows command prompt (cmd
) may interpret options incorrectly.
The supervisor can be interrupted cleanly using C or D. If this is used, the supervisor will wait for seeds currently being rolled to finish before exiting, but will no longer start rolling any new seeds.
--suite
: Runs the benchmarking suite. Should usually be combined with thebench
subcommand. Cannot be combined with--preset
,--settings
, or--rsl
. The benchmarking suite consists of:- the Default/Beginner preset
- the current main tournament settings
- the current multiworld tournament settings
- Hell Mode
- a version of the random settings script adjusted for compatibility with main Dev
--rsl
: Roll seeds using the random settings script.-u
,--github-user
: Specifies the GitHub user or organization name from which to clone the randomizer (or the random settings script if combined with--rsl
). Defaults toOoTRandomizer
(ormatthewkirby
if combined with--rsl
).--repo
: Specifies the repository name on GitHub from which to clone the randomizer (or the random settings script if combined with--rsl
). Defaults toOoT-Randomizer
(orplando-random-settings
if combined with--rsl
).-b
,--branch
: Specifies the git branch of the randomizer (or of the random settings script if combined with--rsl
) to clone. Defaults to the repository's default branch.--rev
: Specifies the git revision of the randomizer (or of the random settings script if combined with--rsl
) to clone. Must be given as an unabbreviated git commit hash. Cannot be combined with--branch
.-p
,--preset
: The name or an alias of the settings preset to use. Defaults to the Default/Beginner preset. Cannot be combined with--rsl
,--settings
, or--suite
.--settings
: The settings string to use for the randomizer. Cannot be combined with--preset
,--rsl
, or--suite
.--draft
: Simulates a settings draft from the given file. Seeassets/draft
for examples. Cannot be combined with--preset
,--settings
, or--rsl
.--json-settings
: Specifies a JSON object of settings on the command line that will override the given preset, settings string, or draft picks. Cannot be combined with--rsl
.--world-counts
: Each seed will override the value of theworld_count
setting to be equal to its seed ID (plus 1 because seed IDs start at 0). Restricts the--num-seeds
option to a maximum of 255. Cannot be combined with--rsl
.--patch
: Generate.zpf
/.zpfz
patch files and include them in thestatsDir
. Cannot be combined with--rsl
or with thebench
subcommand.
-n
,--num-seeds
: Specifies the sample size, i.e. how many seeds to roll. Any existing seeds will be reused. Defaults to 16384.--race
: If there are more available cores than remaining seeds, roll the same seed multiple times, racing the instances of the randomizer against each other to keep the one that finishes first. This option should not be used for statistics since it will skew results, but it can be useful when generating seeds for other purposes.--retry-failures
: If the randomizer errors, retry instead of recording as a failure. Care should be taken when using this option for statistics since it may skew results, but it can be useful when generating seeds for other purposes. Cannot be combined with thefailures
subcommand.--clean
: Delete any existing stats instead of reusing them.-w
,--worker
: Use only the specified worker(s). May be specified multiple times. Cannot be combined with--exclude-worker
.-x
,--exclude-worker
: Don't use the specified worker(s). May be specified multiple times. Cannot be combined with--worker
.--json-messages
: Produce status updates on stderr and command results on stdout in JSON Lines format instead of the normal human-readable status display and command output.
Benchmarks the randomizer by measuring the average number of CPU instructions required to successfully generate a seed, taking into account the failure rate of the randomizer.
This subcommand requires workers to have access to perf
, which is only available for Linux. Workers running on Windows will attempt to use WSL. To install perf
on an Ubuntu or Debian distro running inside WSL, run apt-get install linux-tools-generic
and copy/symlink /usr/lib/linux-tools/*-generic/perf
into your PATH
.
Results will be displayed on stdout.
If this subcommand is run with the --raw-data
option, it will output the following data instead of displaying a summary: Each seed's data is printed on a separate line, starting with the character s
for success or f
for failure, followed by a space, followed by the number of instructions taken.
Runs the given JQ filter (a required positional argument) on every spoiler log, and displays how many times each distinct value occurs in the outputs. Failed seeds are ignored. Results will be displayed on stdout.
Displays the 10 most common exceptions returned by the randomizer, grouped by the location in the code where they were raised. Results will be displayed on stdout.
Collects statistics about the chest appearances in Mido's house, and saves them as a JSON file to the given path (a required positional argument). Used for generating the midos.house logo.