pyquantfin 0.0.1

package for quant finance lecture

---

nav_calc

---

It is a lib for backtesting.

multi_assets

This function is used to test strategies involving mutliple assets, like multi-factors model.

params:

• prz: dataframe, index = dates, columns = stocks

• w_tar: dataframe, target weight, index = dates, columns = stocks

It is worthwhile to note that the dates in w_tar does not need to be the same as the dates in prz, it can be a subset of the dates in prz.

• dates_reb: array, rebalance dates

If you want to balance at every w_tar's dates, set dates_reb to w_tar.index

• fee: fee rate, default = 0

return a dataframe

It has the following columns:

• reb: 1 indicate it is a rebalance date, 0 for not a rebalance date

• ret_gross: return before fee

• fee: fees to substract from ret_gross

• ret_net: return after fee

• nav: net asset value of the strategy

time_series

This function is used to test timing strategies, like CTA strategies.

params:

• prz: array

• pos: array, indicate the strategy's position

It must be the same size with prz

• fee: fee rate, default 0

• ret_ext: whether to return the extended information; if False(default), return an array which represents the nav of the strategy; if True, return a DataFrame(see below)

return a dataframe

(ret_ext = True)

It has the following columns:

• price: the price of the underlying asset

• position: the strategy's position

• return: pct_change of the price

• value: position * price

• fee: trading fees

• strategy_return_before_fee:

• strategy_return: post-fee return

• nav: nav of the strategy (post_fee)

• nav_before_fee:

• nav_price: the normalized price

---

nav_analysis

---

It is a lib for analysis nav (which maybe from nav_calc).

calc_mdd

Function to calculate max drawdown

params

• arr: an array that represents nav or price

return a dict

The dict has the following keys:

• mdd_value: the max drawdown in value

• mdd_pct: the max drawdown in percentage

calc_ann_ret

Function to calculate the annualized return, assuming daily data.

parmas

• arr: an array that represents nav or price

return a value

• annualized return

stats

Function to return both ret and mdd

params

• arr: an array that represents nav or price

return a dict which has the following keys

• ret: the total return (not annualized return)

• mdd: the max drawdown in percentage

stats_yearly

params

• s: a series that represents nav or price

The index of s should be date and must be in the form like "20181010".

return a dataframe which has the following columns:

• ret: the total return (not annualized return)

• mdd: the max drawdown in percentage

The index of the dataframe is 'year'.

stats_yearly_multi_assets

params

• df: a dataframe whose each column represents a nav of an asset or strategy.

return a dataframe which has two-level columns

Level I: the asset or strategy

Level II: ret and mdd, the same meaning as in stats_yearly

---

general

---

to_db_code

params

• ts_code: security code like '600000.SH'

return

• db_code: code format in database like 'SH_600000'

example

• to_db_code('600000.SH') returns 'SH_600000'

to_ts_code

params:

• db_code: code format in database like 'SH_600000'

return:

• ts_code: security code like '600000.SH'

example:

• to_ts_code('SH_600000') returns '600000.SH'

real_round_2

precise round to 0.01 (用于计算涨跌停价)

param

• a number

return

• number rounded

example

-read_round_2(5.347) returns 5.35

check_df_struct

check if two dataframe have identical struct: same shape, same index and same columns

params

• df1, df2

return

• True if the two dataframe have the same shape, same index and same columns, else False

example

check_df_struct(df1, df2)

---

utils

---

grouping1d

assign groupid to an array, used to group stocks according a factor

params

• arr: the factor array
• nog: number of group
• ST = 999: the number need to be specially treated

return

an array that contained groupid, nan will be assigned to the 0 group, ST will be assigned to the -1 group

example

grouping1d(np.arange(100),5)

get_clean_factor_and_forward_returns

param

• factor: a mult-index series with index = ['date', 'asset'], value = factor value

• prices: a dataframe with index = daily dates and columns = the universe of stocks interested

• bins = 5: number of group (use bins as the variable name in order to be consistence with alphalens)

• periods = (20,60): the holding periods to be tested

return

a multi-index dataframe with index = ['date', 'asset'] and columns:

• XD(i.e. 20D, 60D,...): the forward return

• factor: the factor value

• factor_quantile: the group id (also, to be consistence with alphalens)

example

factor_data = get_clean_factor_and_forward_returns(factor, prices)

reference

alphalens: http://quantopian.github.io/alphalens/index.html

convert_date_in_factor_data

convert the type of index 'date' in factor_data to datetime

param

factor_data: with be changed inplace

return

None

example

convert_date_in_factor_data(factor_data)

---

SAA

---

risk_budget

params

• data: the DataFrame consists of asset price data
• b: the risk budget (list or array)

return

• a Series representing the weights

risk_budget_weight_resample

params

• data: the DataFrame consists of asset price data
• b: the risk budget (list or array)
• lb: the lookback period
• freq: the rebalace frequency (str or int)

return

• a DataFrame representing the weights at each rebalance date

risk_parity

params

• data: the DataFrame consists of asset price data

return

• a Series representing the weights

rp_weight_resample

params

• data: the DataFrame consists of asset price data
• lb: the lookback period
• freq: the rebalace frequency (str or int)

return

• a DataFrame representing the weights at each rebalance date

equalweight_resample

params

• data: the DataFrame consists of asset price data
• freq: the rebalace frequency (str or int)

return

• a DataFrame representing the weights at each rebalance date

calc_nav

params

• data: the DataFrame consists of asset price data
• wegiths: the DataFrame representing the weights at each rebalance date
• fee: fee rate

return

• df_nav: a DataFrame the contains the backtested nav of the strategy
• df: a muli_level DataFrame that contains the raw data, the weights and the period return of that assets, at each rebalace data