#!/usr/bin/env python
# coding: utf-8

# In[1]:


import pandas as pd
import numpy as np
import requests


# In[31]:


pd.set_option('display.precision',10)


# *Bid-ask spread* is a normalized value: (bid+ask)/(2*bid) - 1 

# In[37]:


tick = requests.get('https://poloniex.com/public?command=returnTicker').json()


# In[38]:


tick_df = pd.DataFrame.from_records([(k,v['highestBid'],v['lowestAsk']) for k,v in tick.items()])
tick_df.columns = ('symbol','bid','ask')
tick_df['bid'] = tick_df['bid'].astype(float)
tick_df['ask'] = tick_df['ask'].astype(float)
tick_df.head()


# In[39]:


tick_df['bid_ask_spread_norm'] = (tick_df['bid']+tick_df['ask'])/(2*tick_df['bid'])-1


# In[40]:


tick_df.sort_values('bid_ask_spread_norm').head(10)


# In[41]:


tick_df.sort_values('bid_ask_spread_norm').tail(10)


# The main issue with this metrics is that spread changes rapidly and top/tail results are different in any given time.

# **Handy liquidity** 
# 
# HL = Bid-ask spread / Bid-ask spread by X units of an asset, where
# X is 1 BTC or equivalent for day traders; 10 BTC or equivalent for investors. 

# In[124]:


def calc_liquidity(symbol):
    r = requests.get('https://poloniex.com/public?command=returnOrderBook&currencyPair={symbol}&depth=1000'.format(symbol=symbol)).json()
    asks = pd.DataFrame.from_records(r['asks']).rename(columns={0:'price',1:'amount'})
    bids = pd.DataFrame.from_records(r['bids']).rename(columns={0:'price',1:'amount'})
    asks['price'] = asks['price'].astype(float)
    bids['price'] = bids['price'].astype(float)
    asks['vol'] = asks['price']*asks['amount']
    bids['vol'] = bids['price']*bids['amount']
    asks['cumvol'] = asks['vol'].cumsum()
    bids['cumvol'] = bids['vol'].cumsum()
    
    best_bid = bids['price'].values[0]
    best_ask = asks['price'].values[0]
    best_1b_bid =  bids[bids['cumvol']>=1]['price'].values[0] if (bids['cumvol']>=1).sum()>0 else np.NaN
    best_1b_ask = asks[asks['cumvol']>=1]['price'].values[0] if (asks['cumvol']>=1).sum()>0 else np.NaN
    best_10b_bid = bids[bids['cumvol']>=10]['price'].values[0] if (bids['cumvol']>=10).sum()>0 else np.NaN
    best_10b_ask = asks[asks['cumvol']>=10]['price'].values[0] if (asks['cumvol']>=10).sum()>0 else np.NaN
    
    return (symbol, best_bid, best_ask, best_1b_bid, best_1b_ask, best_10b_bid, best_10b_ask)


# In[125]:


col_names = ('symbol','best_bid', 'best_ask', 'best_1b_bid', 'best_1b_ask', 'best_10b_bid', 'best_10b_ask')


# In[147]:


res = pd.DataFrame.from_records([calc_liquidity(symbol) for symbol in list(tick.keys())])
res.columns = col_names


# In[148]:


res.head()


# In[149]:


# filter out non BTC_XXX pairs
res = res[res['symbol'].str.startswith('BTC_')]


# In[150]:


res['bid_ask_spread_norm'] = (res['best_bid']+res['best_ask'])/(2*res['best_bid'])-1
res['bid_ask_1b_spread_norm'] = (res['best_1b_bid']+res['best_1b_ask'])/(2*res['best_1b_bid'])-1
res['bid_ask_10b_spread_norm'] = (res['best_10b_bid']+res['best_10b_ask'])/(2*res['best_10b_bid'])-1


# In[151]:


res.head()


# In[152]:


res.isnull().sum()


# In[153]:


# less then 10 btc
res[res['bid_ask_10b_spread_norm'].isnull()]


# In[154]:


res.sort_values('bid_ask_10b_spread_norm').head(10)


# In[155]:


res.dropna().sort_values('bid_ask_10b_spread_norm').tail(10)


# ** Handy liquidity 2 **
# 
# The cumulative volume that is present in an order book at levels that are away from the naive mid market price by 0.5% or less.  

# In[156]:


def calc_liquidity2(symbol):
    r = requests.get('https://poloniex.com/public?command=returnOrderBook&currencyPair={symbol}&depth=1000'.format(symbol=symbol)).json()
    asks = pd.DataFrame.from_records(r['asks']).rename(columns={0:'price',1:'amount'})
    bids = pd.DataFrame.from_records(r['bids']).rename(columns={0:'price',1:'amount'})
    asks['price'] = asks['price'].astype(float)
    bids['price'] = bids['price'].astype(float)
    asks['vol'] = asks['price']*asks['amount']
    bids['vol'] = bids['price']*bids['amount']
    
    asks = asks[asks['price'] < asks['price'].min()*1.005]
    bids = bids[bids['price'] > bids['price'].max()*0.995]
    
    asks['vol'].sum() + bids['vol'].sum()
    
    return (symbol, asks['vol'].sum() + bids['vol'].sum())


# In[157]:


col_names = ('symbol','cum_vol')


# In[158]:


res = pd.DataFrame.from_records([calc_liquidity2(symbol) for symbol in list(tick.keys())])
res.columns = col_names


# In[161]:


# filter out non BTC_XXX pairs
res = res[res['symbol'].str.startswith('BTC_')]


# In[162]:


res.head()


# In[168]:


res.sort_values('cum_vol',ascending=False).head(20)


# In[167]:


res.sort_values('cum_vol',ascending=True).head(10)


# In[ ]: