Pairs Cointegration code error

[Diego_Choquesillo]

Hello,

Im coding a cointegration strategy between KO and PEP the last code to graph the PNL return me and error. i put all the code and the error. i couldnt find why that error

get_ipython().system(‘pip install --upgrade yfinance’)
import yfinance as yf
import numpy as np
import pandas as pd
import statsmodels

import matplotlib.pyplot as plt
import seaborn

For cointegration

import statsmodels.api as stat
import statsmodels.tsa.stattools as ts

import warnings
warnings.filterwarnings(“ignore”)
#im going to check the cointegration at 90% confidence with dickey-fuller test
import pyexcel
from datetime import datetime

In[329]:

get_ipython().system(‘pip install pandas-datareader’)

In[330]:

x=yf.download(“KO”,start=“2019-01-01”, end=“2022-12-31”,auto_adjust=True)

y= yf.download(“PEP”, start=“2019-01-01”, end=“2022-12-31”,auto_adjust=True)

In[331]:

x.index = pd.to_datetime(x.index)
y.index = pd.to_datetime(y.index)

y.isna().sum()
x.isna().sum()
x.duplicated().value_counts()
y.duplicated().value_counts()

In[367]:

def cointegration_test(x,y):

result=stat.OLS(x["Close"],y["Close"]).fit()
x_const = stat.add_constant(x["Close"])
##adf_results=ts.adfuller(result.resid)

if adf_results[0]<=adf_results[4][“10%”]:

    ##return "Pair is cointegrated"

else:

    ##return "Pair is not cointegrated"
return ts.adfuller(result.resid)

In[368]:

print(“Cointegration test”,cointegration_test(x,y))

In[369]:

#New dataframe

In:

In[370]:

data=

headers=[“Date”,“stock1”,“stock2”,“zscore”,“signal”,“status”,“buyprice”,“sellprice”,“mtm”,“cointegration test”]

data.append(headers)

#tolerance for the stdev

threshold=1.75

start=20
end=90
prev_status= “”
mtm=“”
prev_sell_price=“”
sell_price=“”
prev_buy_price=“”
buy_price=“”
SL=-10000
TP=20000

#lotaje
#Data 1
N=5000

#data2
M=5000

In[419]:

def zscore_cal(data1, data2, start, end):

s1 = (data1["Close"].iloc[start:end].values)
s2 = (data2["Close"].iloc[start:end].values)

valid_prices = (s1 > 0) & (s2 > 0)
s1 = s1[valid_prices]
s2 = s2[valid_prices]

if len(s1) == 0 or len(s2) == 0:
    return 0

mvavg_old = np.mean(np.log(s1/s2))
std_old = np.std(np.log(s1/s2))



current_spread = np.log(
    data1["Close"][end]/data2["Close"][end])



zscore = (current_spread - mvavg_old) / \
    std_old if std_old > 0 else 0

return (zscore)

In:

In[420]:

print(type(x[“Close”]))
print(x[“Close”].shape)

In[421]:

def signal_cal(zscore,threshold, adfest):

if zscore> threshold and adfest=="Yes":
    signal="SELL"
    
elif zscore<-threshold and adfest=="Yes":
    signal="BUY"
    
else:
    signal=""
    
return signal

In[422]:

def status_cal(prev_status, mtm, SL, TP, signal, adftest):

if prev_status in["","SL","TP","CB"]:
     status = signal
        
else: 
    if adftest=="No":
            status="CB"
                  
    else:
        if mtm == "":
            status = ""
            
            
        else:
            if mtm < SL:
                status = "SL"
                
            else:
                        
                if mtm>TP:
                    status="TP"
                else:
                    status=prev_status
                    
return status                 

In[423]:

def buy_price_cal(prev_status,prev_buy_price, buy_price,signal,status,data1,data2,end):
if status ==prev_status:
buy_price = prev_buy_price

else:
    if status in ["SL", "TP", "CB", ""]:
        buy_price = ""
    
    else:
        if signal=="BUY":
            buy_price=data1["Close"][end]
        else:
            
            if signal =="SELL":
                buy_price=data2["Close"][end]
            else:
                buy_price= ""
            
return buy_price

In[424]:

def sell_price_cal(prev_status,prev_sell_price,sell_price,signal,status,data1,data2,end):
if status==prev_status:
sell_price=prev_sell_price
else:
if status in [“SL”,“TP”,“CB”,“”]:
sell_price=“”
else:
if signal==“SELL”:
sell_price=data1[“Close”][end]
else:
if signal==“BUY”:
sell_price=data2[“Close”][end]
else:
sell_price=“”
return sell_price

In[425]:

def mtm_cal(data1, data2, prev_status, prev_sell_price, prev_buy_price, M, N, end):

if prev_status == "BUY":
    
    mtm = (prev_sell_price-data2['Close'][end])*M + (data1['Close'][end] -
                                                     prev_buy_price)*N
else:
    if prev_status == "SELL":
        
        mtm = (prev_sell_price-data1['Close'][end])*N + (data2['Close'][end]
                                                         - prev_buy_price)*M
    else:
        mtm = ""

return mtm

In[426]:

def strategy (data1, data2, threshold, start, end, prev_status, mtm, prev_sell_price,sell_price,prev_buy_price,buy_price,SL,TP,M,N):
pnl=0
i=0
while end < len(data1) and end < len(data2):

for row in data1[90:].iterrows():

    d1=data1[i:90+i]
    d2=data2[i:90+i]
    c_t=cointegration_test(d1,d2)
    if c_t[0] <= c_t[4]["10%"] and c_t[1]<=0.1:
        adftest="Yes"
    else:
        adftest="No"
        
    zscore=zscore_cal(data1,data2,start,end)
    
    signal=signal_cal(zscore,threshold,adftest)
    
    mtm=mtm_cal(data1,data2,prev_status,prev_sell_price,prev_buy_price,M,N,end)
    
    status = status_cal(prev_status, mtm, SL, TP, signal, adftest)
    
    buy_price=buy_price_cal(prev_status,prev_buy_price,buy_price,signal,status,data1,data2,end)
    
    sell_price=sell_price_cal(prev_status,prev_sell_price,sell_price,signal,status,data1,data2,end)
    
    pnl=(pnl+mtm) if status in ["TP","SL","CB"] else pnl
    
    prev_sell_price = sell_price
    prev_status = status
    prev_buy_price = buy_price

    
    data.append([
        str(data1.index[end])[:10],
        data1["Close"][end],
        data2["Close"][end],
        zscore,
        signal,
        status,
        buy_price if buy_price != "" else np.nan,
        sell_price if sell_price != "" else np.nan,
        mtm if mtm != "" else 0,
        pnl,
        adftest])
    
    end = end+1
    start = start+1
    i = i+1
    
pyexcel.save_as(array=data,dest_file_name="pair_strategy_withoutzipline-90.csv")

In:

In[427]:

fig, ax=plt.subplots(1,1, figsize=(10,7))

ax.plot(x.index.to_numpy(), x[‘Close’].to_numpy(), label=“Coke”)
ax.plot(y.index.to_numpy(), y[‘Close’].to_numpy(), label=“Pepsi”)

ax.legend()
ax.set_xlabel(“Date”, fontsize=12)
ax.set_ylabel(“Close Prices”, fontsize=12)
ax.set_title(“Coke and Pepsi”, fontsize=14)
ax.grid()

plt.show()

In[428]:

fig, ax = plt.subplots(1, 1, figsize=(10, 7))

Plot the spread

ax.plot(x.index.to_numpy(), np.log(
x.Close.values/y.Close.values), label=‘Spread’)

plt.legend()
plt.xlabel(‘Date’, fontsize=12)
plt.ylabel(‘Spread’, fontsize=12)
plt.title(‘Spread of x and Lead y Prices’, fontsize=14)
plt.grid()
plt.show()

In[429]:

strategy (x, y, threshold, start, end, prev_status, mtm, prev_sell_price,sell_price,prev_buy_price,buy_price,SL,TP,M,N)

result = pd.DataFrame(data, columns=headers)[1:]

result.set_index(‘Date’,inplace=True)

result.index = pd.to_datetime(result.index)

fig, ax = plt.subplots(1, 1, figsize=(12, 7))
ax.plot(result_df.index, result_df[‘pnl’], label=‘Profit and Loss’, color=‘blue’)
ax.axhline(y=0, color=‘black’, linestyle=‘-’) # Línea de referencia en 0
ax.set_title(“Strategy Performance - KO vs PEP Pair Trading”, fontsize=14)

plt.legend()
plt.xlabel(‘Date’, fontsize=12)
plt.ylabel(‘PnL’, fontsize=12)
plt.title(‘PnL’, fontsize=14)
plt.grid()
plt.show()

---

KeyError Traceback (most recent call last)
File ~\anaconda3\envs\quantra_py\lib\site-packages\pandas\core\indexes\base.py:3805, in Index.get_loc(self, key)
3804 try:
→ 3805 return self._engine.get_loc(casted_key)
3806 except KeyError as err:

File index.pyx:167, in pandas._libs.index.IndexEngine.get_loc()

File index.pyx:196, in pandas._libs.index.IndexEngine.get_loc()

File pandas\_libs\hashtable_class_helper.pxi:7081, in pandas._libs.hashtable.PyObjectHashTable.get_item()

File pandas\_libs\hashtable_class_helper.pxi:7089, in pandas._libs.hashtable.PyObjectHashTable.get_item()

KeyError: 90

The above exception was the direct cause of the following exception:

KeyError Traceback (most recent call last)
Cell In[429], line 1
----> 1 strategy (x, y, threshold, start, end, prev_status, mtm, prev_sell_price,sell_price,prev_buy_price,buy_price,SL,TP,M,N)
3 result = pd.DataFrame(data, columns=headers)[1:]
5 result.set_index(‘Date’,inplace=True)

Cell In[426], line 14, in strategy(data1, data2, threshold, start, end, prev_status, mtm, prev_sell_price, sell_price, prev_buy_price, buy_price, SL, TP, M, N)
11 else:
12 adftest=“No”
—> 14 zscore=zscore_cal(data1,data2,start,end)
16 signal=signal_cal(zscore,threshold,adftest)
18 mtm=mtm_cal(data1,data2,prev_status,prev_sell_price,prev_buy_price,M,N,end)

Cell In[419], line 19, in zscore_cal(data1, data2, start, end)
13 mvavg_old = np.mean(np.log(s1/s2))
14 std_old = np.std(np.log(s1/s2))
18 current_spread = np.log(
—> 19 data1[“Close”][end]/data2[“Close”][end])
23 zscore = (current_spread - mvavg_old) /
24 std_old if std_old > 0 else 0
26 return (zscore)

File ~\anaconda3\envs\quantra_py\lib\site-packages\pandas\core\frame.py:4102, in DataFrame.getitem(self, key)
4100 if self.columns.nlevels > 1:
4101 return self._getitem_multilevel(key)
→ 4102 indexer = self.columns.get_loc(key)
4103 if is_integer(indexer):
4104 indexer = [indexer]

File ~\anaconda3\envs\quantra_py\lib\site-packages\pandas\core\indexes\base.py:3812, in Index.get_loc(self, key)
3807 if isinstance(casted_key, slice) or (
3808 isinstance(casted_key, abc.Iterable)
3809 and any(isinstance(x, slice) for x in casted_key)
3810 ):
3811 raise InvalidIndexError(key)
→ 3812 raise KeyError(key) from err
3813 except TypeError:
3814 # If we have a listlike key, _check_indexing_error will raise
3815 # InvalidIndexError. Otherwise we fall through and re-raise
3816 # the TypeError.
3817 self._check_indexing_error(key)

KeyError: 90

[QuantInsti]

Hi Diego,

Debugging is an essential part of coding—so I won’t give away the full answer, but here’s a nudge in the right direction.

In your def strategy(): function, the issue lies in the following lines:

    d1=data1[i:90+i]
    d2=data2[i:90+i]

Quick Hint:

x.index = pd.to_datetime(x.index)
y.index = pd.to_datetime(y.index)

Try to figure it out from here and let us know how it goes!

Thanks,
Ajay

[Diego_Choquesillo]

Thanks Ajay,

I undestand that im using dates as index, so the best way to slicing is using .iloc in the code d1=data1[i:90+i]
d2=data2[i:90+i]
and i put. iloc in all of the [“Close”].iloc[end] because is necesary to find the position.

but the error now is

ValueError Traceback (most recent call last)
~\AppData\Local\Temp\ipykernel_5516\1863136859.py in ?()
----> 1 strategy (x, y, threshold, start, end, prev_status, mtm, prev_sell_price,sell_price,prev_buy_price,buy_price,SL,TP,M,N)
2
3 result = pd.DataFrame(data, columns=headers)[1:]
4

~\AppData\Local\Temp\ipykernel_5516\2126476378.py in ?(data1, data2, threshold, start, end, prev_status, mtm, prev_sell_price, sell_price, prev_buy_price, buy_price, SL, TP, M, N)
12
13 zscore=zscore_cal(data1,data2,start,end)
14
15
—> 16 signal=signal_cal(zscore,threshold,adftest)
17
18 mtm=mtm_cal(data1,data2,prev_status,prev_sell_price,prev_buy_price,M,N,end)
19

~\AppData\Local\Temp\ipykernel_5516\114183041.py in ?(zscore, threshold, adfest)
1 def signal_cal(zscore,threshold, adfest):
2
3
4
----> 5 if zscore> threshold and adfest==“Yes”:
6 signal=“SELL”
7
8 elif zscore<-threshold and adfest==“Yes”:

~\anaconda3\envs\quantra_py\lib\site-packages\pandas\core\generic.py in ?(self)
1575 @final
1576 def nonzero(self) → NoReturn:
→ 1577 raise ValueError(
1578 f"The truth value of a {type(self).name} is ambiguous. "
1579 “Use a.empty, a.bool(), a.item(), a.any() or a.all().”
1580 )

ValueError: The truth value of a Series is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all().

I understant that the probleme now is in zcore?

[Ajay_Pawar]

Hi Diego,

Which IDE are you currently using?

I’d recommend trying Spyder — it lets you run code line by line, making it much easier to see how each variable changes as the script executes. This can be super helpful for tracking down issues.

Also, I’d encourage installing a few helpful debugging tools:

pip install pdb
pip install ipdb
pip install pysnooper
pip install viztracer

These help you step through code and inspect variables at each point. For example:

• pdb.set_trace() pauses execution so you can inspect what’s going on.
• pysnooper gives a trace of variable changes line by line.
• viztracer is great for performance and function flow visualization.

---

About the error you encountered

That ValueError: The truth value of a Series is ambiguous is a classic issue in Pandas, and it’s a great learning opportunity.

When comparing something like if zscore > threshold, check:

• Is zscore a Series or a scalar?
• If it’s a Series, are you trying to evaluate one value or the whole Series?

This type of problem encourages us to ask:

• Should I use .iloc[-1] to get the latest value?
• Or .any() / .all() to evaluate multiple conditions?

Before diving into a fix, it’s good practice to pause and check:

type(zscore)
print(zscore)

This lets you see exactly what you’re working with. Debugging is often about asking good questions rather than immediately jumping to a fix.

---

If you can get into the habit of exploring what each variable contains, asking “what is this doing?”, and checking whether it’s giving what you expect — you’ll be able to debug most things on your own over time.

Happy debugging!

[Diego_Choquesillo]

Im tried to figuerout whats the problem in this part of the code.

I check my codes but i could find whats the problem im thinking if the status is change, because i always have 0 as result.

x.columns: MultiIndex([( ‘Close’, ‘MSFT’),
( ‘High’, ‘MSFT’),
( ‘Low’, ‘MSFT’),
( ‘Open’, ‘MSFT’),
(‘Volume’, ‘MSFT’)],
names=[‘Price’, ‘Ticker’])
y.columns: MultiIndex([( ‘Close’, ‘AAPL’),
( ‘High’, ‘AAPL’),
( ‘Low’, ‘AAPL’),
( ‘Open’, ‘AAPL’),
(‘Volume’, ‘AAPL’)],
names=[‘Price’, ‘Ticker’])
x.shape: (1510, 5)
y.shape: (1510, 5)
x.index (sample): DatetimeIndex([‘2017-01-03’, ‘2017-01-04’, ‘2017-01-05’], dtype=‘datetime64[ns]’, name=‘Date’, freq=None)
y.index (sample): DatetimeIndex([‘2017-01-03’, ‘2017-01-04’, ‘2017-01-05’], dtype=‘datetime64[ns]’, name=‘Date’, freq=None)
MTM: 0.00 | Status: SELL | Px_x: 235.04417419433594 | Px_y: 128.26846313476562
Date: 2022-12-30 | Status: SELL | Prev_status: SELL
Signal: SELL | Z-Score: 2.231 | ADF: Yes
Buy: 128.26846313476562 | Sell: 235.04417419433594
MTM: 0.0 | PnL acumulate: 0

[Ajay_Pawar]

Hi Diego,

Great job getting the logic in place and starting to debug the strategy — you’re on the right track!

I’d recommend checking out this blog by QuantInsti:
https://blog.quantinsti.com/algorithmic-trading-chatgpt/

It explains how to use ChatGPT effectively for coding and debugging in algorithmic trading workflows. The approach is quite intuitive — you describe what you want in plain English, and ChatGPT helps generate or fix the code accordingly.

This is also known as the “vibe‑coding” approach — where we express our intentions in natural language and let ChatGPT translate them into quality code, refining it through conversation.

What’s crucial in this process is validating what the AI generates — by carefully checking variable usage, logic, and flow — while letting the LLM handle syntax and error correction.

Let’s apply this:

• Prompt ChatGPT to correct the zscore, indexing, and PnL errors
• Have it produce clean, working implementations of the cointegration strategy (KO vs PEP)
• Iterate via chat to fine‑tune, test, and integrate plots and files

I’m confident this will streamline your debugging process and make the workflow more efficient and intuitive.

Let me know your thoughts!

Best,
Ajay