STEEM Price Feed: STEEM Currency Price Feed for Witnesses (https://github.com/steemed/steem-price-feed/)

Introduction

STEEM Price Feed is yet another
STEEM price feed for witnesses.
It has several advantages over simpler feed scripts including:

• Extensive configurability through a YAML config file.
• The ability to add exchanges through the config file.
• Volume-weighted prices that proportionally reduce the influence of low volume markets.
• Stochastic update based on variance analysis of price history.

Stochastic Updating

One innovation of STEEM Price Feed is that the decision to update at
any cycle is probabilistic. If the current estimated price satisfies the
"hard" publication criteria (exceeds a minimum price fluctation
and minimum waiting period), then the variance of the feed history
is calculated and the current price is converted to a
z-score. This score is
then converted to a cumulative probability using the
standard error function
and if a random number in the interval [0,1) is less than the cumulative
probability, then the current price estimate is published. If not,
then STEEM Price Feed tries again after a waiting period.

Rationale for Stochastic Updating

The stochastic updating utilized by STEEM Price Feed is based on decision
theory, and attempts to answer the question "is the price observed now
significant enough to warrant an update?" The answer to this question
is probabilistic in nature because any significance in a price change has
a component of uncertainty. Updating therefore reflects this uncertainty
in that updating is more likely when there is a higher probability that a price
change is significant (i.e. not due to the random variation observed in
the price history). Because markets are extremely complex
phenomena with far too many variables to model, the most reasonable estimate
for this uncertainty is a
normal distribution.

Dependencies

STEEM Price Feed has only two dependencies: PyYaml and
Requests.
On Ubuntu, these dependencies
can be installed with the following command:

sudo apt-get install python-yaml python-requests

Configuration

An example configuration file is provided in the examples directory as feed-example.yaml.
The configuration file is specified when calling the steem-price-feed.py script:

python steem-price-feed.py /home/ima/steem/feed.yaml

Although the instructions below may appear intimidating,
only a couple of settings at the top of the configuration file actually need to be changed.
These settings are

• witness_name
• wallet_password
• rpc_user
• rpc_password

It is perfectly reasonable to take the rest of the settings as the defaults.

The configuration file has two main sections, settings and market_data.

Section: settings

The settings are

• witness_name: the witness name (e.g. steemed)
• wallet_password: the witness password
• min_publish_interval : do not publish a new feed if it has not been this many hours since the last
• max_publish_interval : pulish a new feed if this many hours have passed since the last publication
• rpc_ip: the ip that the wallet server (cli_wallet) runs on (e.g. 127.0.0.1)
• rpc_port: the port that the wallet server runs on (e.g. 8091)
• rpc_user: the user set for cli_wallet using the --server-rpc-user flag
• rpc_password: the password set for cli_wallet using the --server-rpc-password flag
• default_base: a base price for 1 STEEM, denominated in SBD; used if all else fails
• log_file: optional; a file to print errors and debugging information;
  if not provided, stdout will be used (default: null)
• debug: optional; emit debugging information if set to true (default: false)

Section: market_data

Market data subsections are market pairs, expressed as query_base, where query is the asset
that is priced in terms of the base asset. For example, the btc_usd subsection means that
BTC pricing information will be reported in terms of USD. Volume will be reported in terms of BTC (query).

Each market_data subsection has sub-subsections for each exchange. For example, current btc_usd exchanges
are bitfinex, coinbase, okcoin, bitstamp. Each exchange has two subsections, price and volume
that specify how to retrieve these amounts. Key-value pairs for price and volume are:

• url: the base API url for the query (e.g. https://www.okcoin.com/api/v1/ticker.do)
• query_params: form query parameters used for http GET, express as a JSON object (e.g. {"symbol": "btc_usd"})
• accessor: specifies how to traverse the JSON object to extract the needed information,
  expressed as a list of object members (keys) and/or array indices, intermingled

Example of market_data

An example of a market_data for the OKCoin subsection is:

market_data :
  btc_usd :
    ...
    okcoin :
      price :
        url : "https://www.okcoin.com/api/v1/ticker.do"
        query_params :
           {"symbol": "btc_usd"}
        accessor : ["ticker", "last"]
      volume :
        url : "https://www.okcoin.com/api/v1/ticker.do"
        query_params : {"symbol": "btc_usd"}
        accessor : ["ticker", "vol"]

Query URL and query_params

To query the BTC:USD price for OKCoin, this configuration specifies the URL

https://www.okcoin.com/api/v1/ticker.do?symbol:btc_usd

Notice how the query_paramswere appended as form submission data to the URL:

query_params : {"symbol": "btc_usd"}

Accessing the JSON: acessor

The formatted JSON returned by this query as of the time of writing these instructions is:

{
  "date":"1462307028",
  "ticker": {
              "buy":"449.25", "high":"449.62", "last":"449.25",
              "low":"440.75", "sell":"449.4", "vol":"10025.1961"
            }
}

Assuming the JSON object representing this result is called OK, accessing the last price
would look like:

OK["ticker"]["last"]

This corresponds to the acessor provided for the query in the YAML configuration file:

accessor: ["ticker", "last"]

Because the syntax of accessing array elements is exactly the same as accessing object members
except that member names are strings and array indices are integers, the accessor allows
simply mixing member names and array indices. For example, as of the time of writing these
instructions, the formatted Bittrex market summary for STEEM is:

{
  "success":true,
  "message":"",
  "result":
  [
    {
      "MarketName":"BTC-STEEM", "High":0.00120025, "Low":0.00090000,
      "Volume":35316.37017575, "Last":0.00107443, "BaseVolume":34.61316041,
      "TimeStamp":"2016-05-03T20:35:41.713", "Bid":0.00097042, "Ask":0.00105730,
      "OpenBuyOrders":105, "OpenSellOrders":341, "PrevDay":0.00109784,
      "Created":"2016-04-17T01:22:13.35"
    }
  ]
}

Notice that this API call has a single element array that contains all of the BTC:STEEM market
information. If this JSON object were called TREX, then the volume information would be accessed
like this:

TREX["result"][0]["Volume"]

The corresponding accessor line would then be

accessor : ["result", 0, "Volume"]

Usage

The STEEM Price Feed service is meant to be run as a daemon, although it can easily
be run manually like a normal python script for debugging.

The services steem-cli and steem-price-feed will start on boot if properly configured.
Their setup is described below.

If stopped, these services can be started with the following commands after they are set up:

sudo service steem-cli start
sudo service steem-price-feed start

These commands assume that a suitable steemd instance is running and listening at
the default RPC port.

Upstart Service steem-cli

Because publishing a price feed requires an open wallet, an instance of cli_wallet must
be run as a daemon process, listening on an RPC port. On Ubuntu,
this is best achieved using Upstart services.

To create an Upstart service for cli_wallet, add this simple initialization file at /etc/init/steem-cli.conf
(edited for your system, especially changing the password to something very strong):

# steem-cli service - steem cli_wallet service for witness

description "Steem CLI"
author "Ima Witness <ima@example.com>"

# Stanzas
#
# Stanzas control when and how a process is started and stopped
# See a list of stanzas here: //upstart.ubuntu.com/wiki/Stanzas

# When to start the service
start on runlevel [2345]

# When to stop the service
stop on runlevel [016]

# Automatically restart process if crashed
respawn

# Essentially lets upstart know the process will detach itself to the background
# This option does not seem to be of great importance, so it does not need to be set.
# expect fork

# Specify working directory
chdir /home/ima/steem

# Specify the process/command to start, e.g.
exec ./cli_wallet -u user -p password \
                --rpc-endpoint=127.0.0.1:8091 -d 2>cli-debug.log 1>cli-error.log

Upstart Service steem-price-feed

Similarly, it is desirable for STEEM Price Feed (steem-price-feed.py) to
run as an upstart service so that it starts on reboot and respawns in case
it is terminated undesirably.

Put the following file at /etc/init/steem-price-feed.conf (edited for your
own system).

# steem-price-feed service - steem-price-feed service for witness

description "STEEM Price Feed"
author "Ima Witness <ima@example.com>"

# Stanzas
#
# Stanzas control when and how a process is started and stopped
# See a list of stanzas here: //upstart.ubuntu.com/wiki/Stanzas

# When to start the service
start on runlevel [2345]

# When to stop the service
stop on runlevel [016]

# Automatically restart process if crashed
respawn

# Essentially lets upstart know the process will detach itself to the background
# This option does not seem to be of great importance, so it does not need to be set.
# expect fork

# Specify working directory
chdir /home/ima/steem

# Specify the process/command to start, e.g.
exec /usr/bin/python steem-price-feed.py feed.yaml 2>feed-debug.log 1>feed-error.log