Trader SimBot

Simple crypto trading bot simulator, running against Bitcoin market historical data. You can create change configuration variables, or create your own bot with deeper strategies.

requirements

node, typescript, mongoDB

also ts-node for running scripts at the cmd-line:

npm i -g ts-node

config

cd server
cp src/config/example.env src/config/local.env

install

cd server
npm i

Gettting data to work with

Get some historial BTC-USD price data

note: there's already some data in the repo, but this will load latest data. you'll want to run this script to load it into mongoDB though.

src/scripts/get-binance-data.sh

this will download a few years of hourly CSV data, and load into a local mongoDB.

Running a test bot

Using ts-node to run typescript at the command line from 'server' directory. If you use nodemon it will rerun on each code edit. The runners folder contains some bots run with different configs.

cd server
ts-node src/runners/SwingRun.ts

you'll get something like this:

$ NODE_ENV=local ts-node src/runners/SwingRun.ts
[      DbConn]  |  got dbHandle
[    SwingRun]  |  priceList.length 767
[    SwingRun]  |  market.start  4345.95
[    SwingRun]  |  market.end    6687.99
[    SwingRun]  |  profit.total:     -131.3299999999981
[    SwingRun]  |  log: swinger.log.csv

Logs

So, not a good trading strategy. It lost $131 even though the market went up by nearly $2342 in the same period.

priceList.length = list of price points in the data (eg duration/interval) 767 prices ~= 31 days (binance historical data is at hourly intervals) market start/end = how much market changed in the period

Well, its just a basic bot without much strategy or signals yet. But this simple framework allows other ideas to be tested.

Note: at this time the bot just buys 1BTC each tx, no matter what the price.

This creates a log file in logs/swinger.log.csv (see below)

Basic how it works:

The MoodyBot (mood swings geddit?) just trades on basic swings, with no awareness of pressure/resistance/volume or any other technical indicators (coming soon!)

If you look into SwingRun you'll see:

The DateRange:

  let finder = {
    date: {
      $gte: new Date("2017-10-01T00:00:00.000Z"),
      $lte: new Date("2017-11-01T23:00:00.000Z"),
    }
  }

the Bot Config:

const config: IBotConfig = {
  logfile: 'swinger.log',
  calcConfig: {
    stepDown: 10,
    stepUp: 5
  }
}

This is passed into the very simple bot MoodyBot. stepDown/Up is a setting for how many (USD) the bot will consider a change of 'direction'. Down can be bigger than up as often drops are sharper than rises. If a pattern of directional change is recognized it will trigger a 'swing'.

A 'swing' is defined in Kalk.ts eg:

public calcSwing(miniChart: string): string {
  let sw = '-'
  if (/DDUU$/.test(miniChart)) sw = 'S-U' // swing up
  if (/UU$/.test(miniChart)) sw = 'R-U' // run up
  if (/UDD$/.test(miniChart)) sw = 'S-D'
  if (/DD$/.test(miniChart)) sw = 'R-D'

Then later the U or D triggers a BUY/SELL action. Of course these rules can overlap and apply based on order. So a UUDD SwingDown would also get caught on a DD run-down.

So if the bot sees DDUU two down ticks followed by two up-ticks it will treat this as a swing up S-U. We also check for some other DDD meaning a run going down R-U

The samples run at the same rate as the data being fed in.

So based on these swings the bot will call an action

Viewing results

You can see more details of the trading log in a csv file eg logs/swinger.log.csv

As the trader sim runs it writes logs here.

Logs

(viewing the CSV file with a VScode plugin) The date formats are compatible with google docs if you want to import and see graphs.

Columns:

diff1 - difference last two prices
dir - up/down related to diff1
total - current running profit

Seeing the trades graphed

A full log of the trades is saved logs You can view in a graph form by running the client app.

cd client
yarn start

you can then see the strata graph and zoom in to check what happened. Red arrow down = sell event (trade start) Green arrow up = buy event (trade end) Dotted line = held position

Trade Run

So the profit/loss per trade is the delta between the green/red pairs (not shown on the graph currently)

You can click to zoom in on the trade sequence:

Trade Run

so here you can see close up of the graph. At tick 914

Bot bought in at 4557
hodl a bit
sold too late after drop
bought in again
sold and made small profit (all paper trades)

add binance keys to use realtime data

edit local.env to add these keys if you want to use realtime Binance market data. The bot does NOT currently do any trades, but in any case you can create an API key that only has 'read' capability

BinanceApiKey=YOUR_API_KEY_HERE
BinanceApiSecret=YOUR_API_SECRET_HERE

Otherwise just leave the keys out, we can also run against historical downloaded data.

TODO

more modular strategies
Add more technical analysis types (tulip)
Client to visualize trades in graph format
Profitable bot strategy :D

Please get in touch if you'd like to help add features! dc AT rikai.co

dcsan / swingbot

readme