[USER (data scientist)]: Awesome! Now, how about finding the confidence intervals for the average win/loss ratio of each surface type? And please note that calculating the confidence intervals uses the t-distribution since population standard deviation is unknown and the confidence_level is 0.95. Please generate the code with the output in [tuple, tuple, tuple] type.
My template of code snippet is:

---BEGIN CODE TEMPLATE---

import pandas as pd
import scipy.stats as stats
import pickle
from decision_company import read_csv_file, df_copy, dropna, get_dummies, corr, f_oneway, avg, sem, t_ppf

atp_tennis = read_csv_file('atp_tennis.csv')

# YOUR SOLUTION BEGIN:
<code1>
[GENERATE YOUR CODE]
</code1>
# YOUR SOLUTION END

print(hard_ci)

# save data
pickle.dump(hard_ci,open("./ref_result/hard_ci.pkl","wb"))

# YOUR SOLUTION BEGIN:
<code2>
[GENERATE YOUR CODE]
</code2>
# YOUR SOLUTION END

print(clay_ci)

# save data
pickle.dump(clay_ci,open("./ref_result/clay_ci.pkl","wb"))

# YOUR SOLUTION BEGIN:
<code3>
[GENERATE YOUR CODE]
</code3>
# YOUR SOLUTION END

print(grass_ci)

# save data
pickle.dump(grass_ci,open("./ref_result/grass_ci.pkl","wb"))

---END CODE TEMPLATE---

[YOU (AI assistant)]: No problem! You can use the t-distribution to calculate the confidence intervals. Here's the code for that: 

# MY SOLUTION BEGIN:
