Issues with TWR calculation

I am observing some results that I can't explain in the roi's TWR calculation.

With the following example ledger, the roi output looks plausible:

commodity 1,000.00 CHF 

P 2024-11-08 B 1 CHF 

2024-11-26 buying B
  assets  -20000 CHF 
  investments:B  19986 B @@ 20000 CHF 

P 2024-12-31 B 1.0018 CHF

$ hledger roi -Q --value then,CHF --inv '"investments:"' --pnl STR
+---++------------+------------++---------------+---------------+---------------+-----------++-------++------------+----------+
|   ||      Begin |        End || Value (begin) |      Cashflow |   Value (end) |       PnL ||   IRR || TWR/period | TWR/year |
+===++============+============++===============+===============+===============+===========++=======++============+==========+
| 1 || 2024-10-01 | 2024-12-31 ||             0 | 20,000.00 CHF | 20,021.97 CHF | 21.97 CHF || 1.12% ||      0.18% |    0.72% |
+---++------------+------------++---------------+---------------+---------------+-----------++-------++------------+----------+

If I now add a small other investment that doesn't move in price at all, TWR goes to almost -100%?

commodity 1,000.00 CHF 

P 2024-11-07 A 1 CHF 
P 2024-11-08 B 1 CHF 

2024-11-07 buying A
  assets  -20 CHF
  investments:A  20 A @@ 20 CHF

2024-11-26 buying B
  assets  -20000 CHF 
  investments:B  19986 B @@ 20000 CHF 

P 2024-12-31 B 1.0018 CHF

$ hledger -f test.ledger roi -Q --value then,CHF --inv '"investments:"' --pnl STR
+---++------------+------------++---------------+---------------+---------------+-----------++-------++------------+----------+
|   ||      Begin |        End || Value (begin) |      Cashflow |   Value (end) |       PnL ||   IRR || TWR/period | TWR/year |
+===++============+============++===============+===============+===============+===========++=======++============+==========+
| 1 || 2024-10-01 | 2024-12-31 ||             0 | 20,020.00 CHF | 20,041.97 CHF | 21.97 CHF || 1.12% ||    -69.95% |  -99.16% |
+---++------------+------------++---------------+---------------+---------------+-----------++-------++------------+----------

That doesn't seem right?

I compiled hledger from git (8b027a4).

@adept could you take a look ?

@TPolzer

@TPolzer this should be fixed by #2392

@simonmichael yet another "computing intra-day valuations are hard" bug :( I hope that the day will come when finally all of them are properly squashed

lib: roi: fix a bug with PnL attributed to the wrong period (fixes hl…

I've tried 306e942 and it fixes the initial example, but I think something is still off. Note the difference between these two cases (again, adding just a purchase of a fixed value investment B):

P 2024-11-07 "A" 1 CHF 
P 2024-11-29 "A" 0.972 CHF
P 2024-11-08 "B" 1 "CHF"

2024-11-07 buying A
  assets  -1 CHF 
  investments  1 "A" @@ 1 CHF 

2024-11-15 dividend A
        assets  0.03 CHF 
        equity:profit and loss
        investments  0

2024-12-01 selling A
  assets  0.97 CHF 
  investments  -1 "A" @@ 0.97 CHF

$ hledger -f test.ledger roi --value then,CHF -Q --inv inv --pnl profit
+---++------------+------------++---------------+----------+-------------+-----++-------++------------+----------+
|   ||      Begin |        End || Value (begin) | Cashflow | Value (end) | PnL ||   IRR || TWR/period | TWR/year |
+===++============+============++===============+==========+=============+=====++=======++============+==========+
| 1 || 2024-10-01 | 2024-12-31 ||             0 |        0 |           0 |   0 || 0.00% ||      0.12% |    0.46% |
+---++------------+------------++---------------+----------+-------------+-----++-------++------------+----------+

P 2024-11-07 "A" 1 CHF 
P 2024-11-29 "A" 0.972 CHF
P 2024-11-08 "B" 1 "CHF"

2024-11-07 buying A
  assets  -1 CHF 
  investments  1 "A" @@ 1 CHF 

2024-11-15 dividend A
        assets  0.03 CHF 
        equity:profit and loss
        investments  0

2024-11-26 buying B
  assets  -1 CHF
  investments  1 "B" @@ 1 CHF

2024-12-01 selling A
  assets  0.97 CHF 
  investments  -1 "A" @@ 0.97 CHF

$ hledger -f test.ledger roi --value then,CHF -Q --inv inv --pnl profit
+---++------------+------------++---------------+-----------+-------------+-----++-------++------------+----------+
|   ||      Begin |        End || Value (begin) |  Cashflow | Value (end) | PnL ||   IRR || TWR/period | TWR/year |
+===++============+============++===============+===========+=============+=====++=======++============+==========+
| 1 || 2024-10-01 | 2024-12-31 ||             0 | 1.000 CHF |   1.000 CHF |   0 || 0.00% ||      1.56% |    6.34% |
+---++------------+------------++---------------+-----------+-------------+-----++-------++------------+----------+

I dont see any issues with this at a first glance

In the first case:

Initial value is 0
1 A is purchased @ 1 CHF, rate of return for period leading up to this purchase is zero
0.03 CHF is accrued, rate of return for this period is 3%
0.03 CHR is paid out, RoR is 0%
1 A is sold for 0.97 CHF, rate of return for period leading up to this cashflow is -2.8% (we started the period with value of 1 CHF, and ended with value of 0.972 CHF), (0.972/1 - 1)*100 = -2.8%
RoR until the end of the reporting period is 0%

Chaining it all up: ((1+0)*(1+0.03)*(1+0)*(1-0.028)*(1+0)-1) = 0.00116, or multiplying by 100 and rounding up to two digits, 0.12% for 1/4 year, which gives 0.46% annualized.

In the second case:

Initial value is 0
1 A is purchased @ 1 CHF, rate of return for period leading up to this purchase is zero
0.03 CHF is accrued, rate of return for this period is 3%
0.03 CHR is paid out, RoR is 0%
1 B is bought @ 1 CHF, rate of return for this period is 0%
1 A is sold for 0.97 CHF, rate of return for period leading up to this cashflow is -1.4% (we started the period with value of 2 CHF, and ended with value of 1.972 CHF), (1.972/2-1)*100 = -1.4%
RoR until the end of the reporting period is 0%

Chaining it all up: ((1+0)*(1+0.03)*(1+0)*(1+0)*(1-0.014)*(1+0)-1) = 0.01558, or multiplying by 100 and rounding up to two digits, 1.56% for 1/4 year, which gives 6.34% annualized.

The difference between two cases is that both A and B contribute to the value of your investment, and in the second case the underperformance of A was compensated somewhat by the stable B, which leads to better TWR.

Hm, it's entirely possible that my intuition of how TWR "should" behave is off... In that case, thank you for the quick fix and explanation.

Thank you both!

lib: roi: fix a bug with PnL attributed to the wrong period (fixes #2391

Navigation Menu

Repository navigation

Issues with TWR calculation #2391

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issues with TWR calculation #2391

Description

Activity

simonmichael commented on May 24, 2025

adept commented on May 25, 2025

TPolzer commented on May 25, 2025

adept commented on May 25, 2025

TPolzer commented on May 25, 2025

simonmichael commented on May 27, 2025

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions