Cumulative Portfolio performance by Name

Question

This is what my data.frame looks like. The rightmost column(Performance) is my desired column.

library(data.table)
    dt <- fread('
    Name      FundName     SharePrice   TotalShares   PurchaseDate   Performance
    John       A               10           500          2016-01-01       0%   
    John       A               20           1000         2016-02-01       20%     
    John       A               10           1500         2016-03-01      -25%%          
    John       B               30           500          2016-04-01      -18.18%       
    John       B               60           1000         2016-05-01       4.16%       
    Tom        A               10           500          2016-01-01       0%   
    Tom        A               20           1000         2016-02-01       20%     
    Tom        A               10           1500         2016-03-01      -25%%          
    Tom        B               30           500          2016-04-01      -18.18%       
    Tom        B               60           1000         2016-05-01       4.16%                    
      ')

• For the first row, John bought 500 shares at the price of $10. So he spent 5000 bucks and he can sell his position for 5000 bucks that day. So the performance for the first row is 0%.
• For the second row, John spent total of (10*500)+(20*1000)=25000 and he can sell his position for (20 * 1500 shares) or 30000 bucks. So his cumulative performance is (30000-25000)/25000 or 20%.
• For third row he spent total of (10*500+20*1000+10*1500)=40000 bucks and can sell for (10*3000 shares) or 30000 bucks, thereby the performance of -25%.
• Now for the 4th row, there are two funds(A&B); he can sell his A position for 30000 still as I am assuming the price of A is still $10 a share, and he bought his new B position for (30*500)=15000. So he can sell his position of both A & B for (30000+15000)=45000 but he spent(40000+15000) or 55000 so far. so his his cumulative performance for the 4th row is (55000-45000)/55000 or -18.18%.

I hope it makes sense. I am struggling with tracking the price of both funds when trying to calculate the cumulative performance. Thank you for your help.

Answer 1

I'd expand the data to cover all Date-Fund combos for each person:

dt_skel = dt[, do.call(CJ, c(.SD, unique=TRUE)), 
  by=Name, .SDcols=c("FundName", "PurchaseDate")]

dt_full = dt[dt_skel, on=names(dt_skel)]
dt_full[ is.na(TotalShares), TotalShares := 0L]
dt_full[ , SharePrice := SharePrice[1L], by=.(Name, FundName, cumsum(!is.na(SharePrice)))]

Then aggregate

res = dt_full[!is.na(SharePrice), .(
  PurchaseDate,
  spent = cumsum(TotalShares*SharePrice),
  value = cumsum(TotalShares)*SharePrice
), by=.(Name, FundName)][, .(
  value = sum(value),
  spent = sum(spent)
), by=.(Name, PurchaseDate)]


    Name PurchaseDate  value  spent
 1: John   2016-01-01   5000   5000
 2: John   2016-02-01  30000  25000
 3: John   2016-03-01  30000  40000
 4: John   2016-04-01  45000  55000
 5: John   2016-05-01 120000 115000
 6:  Tom   2016-01-01   5000   5000
 7:  Tom   2016-02-01  30000  25000
 8:  Tom   2016-03-01  30000  40000
 9:  Tom   2016-04-01  45000  55000
10:  Tom   2016-05-01 120000 115000

To add the performance metric to the original transactions table:

dt[res, ret := value/spent - 1, on=c("Name, PurchaseDate")]

---

Assuming dates are always monthly, you can make dt_skel smaller with

dt_skel = dt[, MaxDate := max(PurchaseDate), by=Name][, 
  seq(from = PurchaseDate[1L], to =MaxDate[1L], by="month"), by=.(Name, FundName)]

Of course, dates should be formatted as Date or IDate for this to work.