Skip to content

which is the program used to execute OBPI strategy

Notifications You must be signed in to change notification settings

mick-liu/OBPI_strategy

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 

Repository files navigation

OBPI_strategy

which is the program used to execute OBPI strategy

Python code display (part)

class monte_carlo:
   
   def __init__(self, start, end):
       self.start = start
       self.end = end
          
   def monte_carlo_obpi(self, num_simulations, predicted_days, k, rf, sigma):
       returns = self.returns
       prices = self.prices
       
       last_price = prices[-1]
       monte_df = pd.DataFrame()
       obpi_df = pd.DataFrame()
       final_df = pd.DataFrame()
       
       for x in range(num_simulations):
           count = 0
           avg_daily_ret = returns.mean()
           variance = returns.var()
           
           daily_vol = returns.std()
           daily_drift = avg_daily_ret - (variance / 2)
           drift = daily_drift - 0.5 * daily_vol ** 2
           
           prices = []
           
           shock = drift + daily_vol * np.random.randn()
           last_price * math.exp(shock)
           prices.append(last_price)
           
           for i in range(predicted_days):
               if count == 251:
                   break
               shock = drift + daily_vol * np.random.randn()
               price = prices[count] * math.exp(shock)
               prices.append(price)
               
               count += 1
           
           monte_df['x'] = prices
           
           Nd1 = []
           Nd2 = []
           tt = []
           W1 = []
           W2 = []
           ss = []
           t = 1
           count = 0
           for j in range(predicted_days):
               if count == 251:
                   break
               s = monte_df['x'][count]
               d1 = math.log(s/k) + (rf + 0.5 * sigma **2) * t
               d1 = d1/(sigma * math.sqrt(t))
               nd1 = scipy.stats.norm(0, 1).cdf(d1)
               md2 = -(d1 - sigma * math.sqrt(t))
               mnd2 = scipy.stats.norm(0, 1).cdf(md2)
               w1 = (s * nd1) / (s * nd1 + k * math.exp(-rf * t) * mnd2)
               w2 = 1 - w1
               W1.append(w1)
               W2.append(w2)
               Nd1.append(nd1)
               Nd2.append(mnd2)
               tt.append(t)
               ss.append(s)
               t = t - 1/predicted_days
               count += 1
               
           obpi_df['s'] = ss
           obpi_df['w1'] = W1
           obpi_df['w2'] = W2
           obpi_df['Nd1'] = Nd1
           obpi_df['Nd2'] = Nd2
           
           V = []
           v = 1
           t = 1
           count = 0
           for l in range(predicted_days):
               if count == 199:
                   break
               v1 = (v * obpi_df['w1'][l]/obpi_df['s'][l]) * obpi_df['s'][l+1]
               v2 = v * obpi_df['w2'][l] * math.exp(-rf * t)
               v = v1 + v2
               t = t - 1/predicted_days
               V.append(v)
               count += 1
               
           final_df[x] = V     
           self.final_df = final_df
           self.predicted_days = predicted_days    

About

which is the program used to execute OBPI strategy

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published