Sneaky

Can we make money by buying sneakers low on one platform and selling high on another?

Sneaky is a prototype with multi-venue feed, price-margin-based strategy and history / volume / volatility analytics tooling for trading sneakers on consignment stores and exchanges.

Sneaky系统从多个平台 (毒app，StockX) 抓取球鞋价格数据、存储，并决定哪些款式有合适的价差、volatility、volume，提供倒卖建议。

For research / study purposes only.

What to expect

• Feed

  • Scrapes venues for current price listing and historical transactions.
  • Three main modes of operation:
  • Query: given search terms, scrape models on a venue, record static information (e.g. style ID, title, release date, product ID / URL query key on the venue, feed/static_info_serializer.py).
  • Update: given static info produced by query, dump current price readings and transactions since the last update to local file system (feed/time_series_serializer.py).
  • Study: given style ID and size, query a venue for current listing prices and historical transactions data.
  • Venues:
  • Du (__feed/du_feed.py__, fully implemented): scrapes Du WeChat app for product detail and transaction history.
  • StockX (__feed/stockx_feed.js__, fully implemented): scrapes StockX (web service) for current best bid and offer.
  • Flightclub (flightclub/, on v1): scrapy-based crawler to parse sell prices from FlightClub webpage.
  • Goat (not implemented).
  • Ebay (not pursued, low match rate and lack of authenticity guarantee would greatly hurt automation).

• Strategy

  • Given time series data feed produced and a configurable set of options, print the most lucrative models for purchase (strategy/strategy.py).

• Analytics

  • Given time series data and a model, print stats or plot historical transactions (__feed/du_analyzer.py__).

• Model of operation

  • Run query every month / year to account for new releases and potential changes in (StyleID, platform-specific ID) map.
  • After scraping static info from multiple venues, run a join to produce items for which we have data on multiple venues (__feed/csv_merge.py__)
  • Cron update everyday to produce a sample of today's prices and record transactions since yesterday.
  • Run strategy whenever you want to see its recommendations based off of currently recorded readings.
  • Before making a purchase, run analytics to double check its volume, volatility and historical transaction prices.

How to run

• Dependencies
• Create credentials
• Feed

  
  # Du product details => du.mapping.{now}.csv
  ./du_feed.py --mode query --kw ../stockx/query_kw.txt --pages 20

StockX product details => stockx.mapping.{now}.csv

This is recommended to circumvent an antibot mechanism enforced by StockX

./stockx_query.sh

Merge

./csv_merge.py --output merged.20191225.csv stockx.20191225.csv du.mapping.20191221-150959.csv

Du current listing and historical transactions => data/{model}/{size}.json

./du_feed.py --mode update --start_from merged.20191225.csv --transaction_history_date 20190801 --transaction_history_maxpage 20 --min_interval_seconds 3600

StockX current listing and historical transactions => data/{model}/{size}.json

This is recommended to circumvent an anti-bot mechanism enforced by StockX

./stockx_update.sh merged.20191225.csv

* Strategy
```sh
# Time series price/transaction data data/{model}/{size}.json => recommendations
./strategy.py --start_from ../feed/merged.20191225.csv

• Analytics

  
  # plot Du historical transaction prices
  ./du_analyzer.py --style_id 881426-009 --size 7.0 --mode plot

produce Du historical transaction statistics

./du_analyzer.py --style_id 881426-009 --size 7.0 --mode stats

### Sample outputs

Strategy prints its recommendation and key stats.

310810-022 8.0 41.0 air-jordan-13-retro-low-chutney , Air Jordan 13 Retro Low Chutney Release date: 2017-06-10 23:59:59 du listing price: 2009.00 CNY 287.17 USD du transaction price: 2009.00 CNY 287.17 USD du transaction time: 2019-12-25T03:20:56.453461Z stockx bid: 147.00 USD stockx ask: 160.00 USD stockx annual high: 188.00 USD stockx annual low: 111.00 USD stockx volatility: 0.11 stockx sale last 72h: 0 profit ratio (bid to last): 37.68 % profit value (bid to last): 66.30 USD profit ratio (mid to last): 32.77 % profit value (mid to last): 59.80 USD profit ratio (ask to last): 28.21 % profit value (ask to last): 53.30 USD Du Transactions: First Date: 2019-08-03T03:21:02.509691 Number of Sales: 8 Sales / Day: 0.06 High: 2019.00 CNY 288.60 USD Low: 1259.00 CNY 179.96 USD First: 1469.00 CNY 209.98 USD Last: 2009.00 CNY 287.17 USD Average: 1691.50 CNY 241.78 USD Stdev: 273.30 CNY 39.07 USD Plot command: ./du_analyzer.py --style_id 310810-022 --size 8.0 --mode plot ...

Analytics produce stats and historical transaction prices plot.

![Historical transaction prices](docs/BQ6623-800.9.5.png)

Feed gathers static product details.

And price, transaction readings.

{ "du": { "prices": [ { "time": "20191225-210334", "bid_price": 211900, "ask_price": null }, ... ], "transactions": [ { "price": 336900, "time": "2019-11-07T21:03:34.791482", "id": "ada3faa5c5f8ed86ff12dfe2a16fd482" }, { "price": 329900, "time": "2019-10-26T21:03:34.791569", "id": "4d95b696577b974a965a1ced6f1bd397" }, ... ] }, "stockx": { "prices": [ { "time": "2019-12-27T02:41:39.802Z", "bid_price": 277, "ask_price": 450, "annual_high": 420, "annual_low": 245, "volatility": 0.105128, "sale_72_hours": 0 }, ... ] } }

### Lessons, progress, and TODO

Tracked in [issues](https://github.com/zhehaowang/sneaky/issues).
Some interesting items:
* [My take on trading sneakers on du](https://github.com/zhehaowang/sneaky/issues/55)
* [Post-mortem on one item we bought](https://github.com/zhehaowang/sneaky/issues/54)
* [Engineering realizations](https://github.com/zhehaowang/zhehao.me/blob/master/essays/engineering.md)

### [Playbook](docs/playbook.md)

### License: GNU Lesser General Public License v3.0

### Credits

* [Unofficial NodeJS stockx API](https://github.com/matthew1232/stockx-api)
* [Unofficial Python stockx API](https://pypi.org/project/stockx-py-sdk/) (in v1)
* [Du app sign mechanism](https://github.com/luo1994/du-app-sign)

### Contact

Zhehao Wang wangzhehao410305@gmail.com