(30-06-2012, 11:41 PM)corydorus Wrote: [ -> ]Do not think you should omit any data as it is real profit/Lost. the super high value is due to solid performance within a short period.
Yes, mathematically the geometric calculation is not incorrect even if hard for the brain to wrap around it.
Quote:If you can consistently do this for rest of year, it is this super value. The reality is, we all can't. The problem can be easily overcome.
If is performance within a year (YTD), what you should do is to use end date 31 dec in your balance unsold stocks.
Yes, if want to find a way to include for YTD measurement, 1 way is to decompress end date of very short period transaction to year end.
Quote:If is multi-years as continuous of your investment since day 1, is not a problem either because over the course of years all your trade recorded will spreaded out, sudden spike due to profit within a short period for a stock is kind of "amortized". If you still see super high value, that's mean your performance is really solid which is unlikely and likely you have computed wrongly.
True, the impact of the very high % of the very short period return will get diluted by longer period returns, with dilution dependent on relative position sizes.
I like your determination and the way you think out of the box.
(01-07-2012, 12:41 AM)swakoo Wrote: [ -> ] (30-06-2012, 11:41 PM)corydorus Wrote: [ -> ]Do not think you should omit any data as it is real profit/Lost. the super high value is due to solid performance within a short period.
Yes, mathematically the geometric calculation is not incorrect even if hard for the brain to wrap around it.
Quote:If you can consistently do this for rest of year, it is this super value. The reality is, we all can't. The problem can be easily overcome.
If is performance within a year (YTD), what you should do is to use end date 31 dec in your balance unsold stocks.
Yes, if want to find a way to include for YTD measurement, 1 way is to decompress end date of very short period transaction to year end.
Quote:If is multi-years as continuous of your investment since day 1, is not a problem either because over the course of years all your trade recorded will spreaded out, sudden spike due to profit within a short period for a stock is kind of "amortized". If you still see super high value, that's mean your performance is really solid which is unlikely and likely you have computed wrongly.
True, the impact of the very high % of the very short period return will get diluted by longer period returns, with dilution dependent on relative position sizes.
I like your determination and the way you think out of the box.
Decompression of the end date does not sound correct...
The short buy/sell transactions should not be part of XIRR computations at all unless the buy involved capital injection (only injected amount is entried). Thus, as per corydorus mentioned, only the ending balance/date matter.
Regarding the geometric calculation... maybe it's easier if one relate it to effective interest rates.
(01-07-2012, 04:19 AM)smallcaps Wrote: [ -> ]Decompression of the end date does not sound correct...
Note my "if" in which case it is more "correct" in a sense.
Quote:The short buy/sell transactions should not be part of XIRR computations at all unless the buy involved capital injection (only injected amount is entried). Thus, as per corydorus mentioned, only the ending balance/date matter.
I know what you mean but even if a stock within portfolio is sold, and the proceeds rolled over to buy another stock which generates returns within days, geometric compounding still arises. Proceeds could have been assumed taken out otherwise.
(01-07-2012, 08:29 AM)swakoo Wrote: [ -> ] (01-07-2012, 04:19 AM)smallcaps Wrote: [ -> ]Decompression of the end date does not sound correct...
Note my "if" in which case it is more "correct" in a sense.
Quote:The short buy/sell transactions should not be part of XIRR computations at all unless the buy involved capital injection (only injected amount is entried). Thus, as per corydorus mentioned, only the ending balance/date matter.
I know what you mean but even if a stock within portfolio is sold, and the proceeds rolled over to buy another stock which generates returns within days, geometric compounding still arises. Proceeds could have been assumed taken out otherwise.
Sorry maybe confused, doesn't geometric compounding gives the correct return that we are interested in? (effective interest rate, for a certain portfolio range of dates)
It's just that we should not double count it by considering the rollovers in the XIRR computations.
Hi, I've been a lurker for quite some time in the forum (>1yr, but only registered after some time) reading all the great posts from you people. This is my first post, hope I can contribute a bit too.
There are many different yield definitions, which one to use depends on your objective and point of view. Common would probably be arithmetic yield, geometric yield, and internal rate of return (IRR).
As I understand it, usually, for past results, geometric yield is more relevant as it shows the effect of actual compounding. However, it is only accurate from the perspective of an investor who
1. puts in an initial investment at the start of the holding period
2. does not deposit or withdraw any money in-between
3. holds the initial investment until the end of the holding period
So if you are calculating returns over the holding period (eg. 5 years), and meet the above 3 criteria, then geometric yield is the correct yield for you.
However, in actual situation, we do not invest an initial sum of money, but over time put in or take out as the situation warrants. Then IRR is more accurate, as it tries to capture what and when you put in or take out from your investment portfolio. That will be your true yield, but is only specific to that case. If the cashflows and timing in-between are different, even though the initial and ending sums are the same, you will get different IRR.
Doltan, XIRR takes care of the in/out and different time periods. Is an improved version of IRR.
In fact if you applied the same condition of equal period, both XIRR and IRR will yield same results.
(01-07-2012, 09:25 AM)smallcaps Wrote: [ -> ] (01-07-2012, 08:29 AM)swakoo Wrote: [ -> ] (01-07-2012, 04:19 AM)smallcaps Wrote: [ -> ]Decompression of the end date does not sound correct...
Note my "if" in which case it is more "correct" in a sense.
Quote:The short buy/sell transactions should not be part of XIRR computations at all unless the buy involved capital injection (only injected amount is entried). Thus, as per corydorus mentioned, only the ending balance/date matter.
I know what you mean but even if a stock within portfolio is sold, and the proceeds rolled over to buy another stock which generates returns within days, geometric compounding still arises. Proceeds could have been assumed taken out otherwise.
Sorry maybe confused, doesn't geometric compounding gives the correct return that we are interested in? (effective interest rate, for a certain portfolio range of dates)
It's just that we should not double count it by considering the rollovers in the XIRR computations.
Ok, my reply to corydorus was general in nature. And early this morning, I was rushing to make a long journey, so there is a possibility I could have misread your posting before firing off a quick reply. But I don't sense there is any fundamental disconnect otherwise.
(30-06-2012, 10:35 PM)swakoo Wrote: [ -> ]1 day
Geometric: (1 + 0.01)^365 - 1 = 3678%
Arithmetic: 1% x 365 = 365%
Since this is a value investing forum, guess holding periods should be long enough not to be affected by this, esp > 1 month.
But if quick kung fu over a few days, XIRR could make one very happy.
Just curious,
In a multi-stocks portfolio, what if I have multiple transactions of,
Buy B then Sell A (1 day later) where A Value > B Value
Buy C then Sell B (1 day later) where B Value > C Value
.
Buy Z then Sell Y (1 day later) where Y Value > Z Value
gives a super high XIRR value (since XIRR doesn't know that you are transacting in different stocks)? Or somehow, the initial Buy A Value and final Z Value will help mitigate the XIRR figure?
(01-07-2012, 10:16 PM)KopiKat Wrote: [ -> ] (30-06-2012, 10:35 PM)swakoo Wrote: [ -> ]1 day
Geometric: (1 + 0.01)^365 - 1 = 3678%
Arithmetic: 1% x 365 = 365%
Since this is a value investing forum, guess holding periods should be long enough not to be affected by this, esp > 1 month.
But if quick kung fu over a few days, XIRR could make one very happy.
Just curious,
In a multi-stocks portfolio, what if I have multiple transactions of,
Buy B then Sell A (1 day later) where A Value > B Value
Buy C then Sell B (1 day later) where B Value > C Value
.
Buy Z then Sell Y (1 day later) where Y Value > Z Value
gives a super high XIRR value (since XIRR doesn't know that you are transacting in different stocks)? Or somehow, the initial Buy A Value and final Z Value will help mitigate the XIRR figure?
Each transaction's XIRR value is high in 1 day, but will be normalized after annualized to 365 days.
It does not matter the number of transactions, but the "annualized return" of each transaction matter.
The best is to try with an example via excel.
(01-07-2012, 10:37 PM)CityFarmer Wrote: [ -> ] (01-07-2012, 10:16 PM)KopiKat Wrote: [ -> ] (30-06-2012, 10:35 PM)swakoo Wrote: [ -> ]1 day
Geometric: (1 + 0.01)^365 - 1 = 3678%
Arithmetic: 1% x 365 = 365%
Since this is a value investing forum, guess holding periods should be long enough not to be affected by this, esp > 1 month.
But if quick kung fu over a few days, XIRR could make one very happy.
Just curious,
In a multi-stocks portfolio, what if I have multiple transactions of,
Buy B then Sell A (1 day later) where A Value > B Value
Buy C then Sell B (1 day later) where B Value > C Value
.
Buy Z then Sell Y (1 day later) where Y Value > Z Value
gives a super high XIRR value (since XIRR doesn't know that you are transacting in different stocks)? Or somehow, the initial Buy A Value and final Z Value will help mitigate the XIRR figure?
Each transaction's XIRR value is high in 1 day, but will be normalized after annualized to 365 days.
It does not matter the number of transactions, but the "annualized return" of each transaction matter.
The best is to try with an example via excel.
Exactly. I think KopiKat has cited a very good example of what corydorus was trying to say and my clumsy replies to her.
Just to flesh this out.
Suppose an investor buys a stock for $1 and makes a gain of 1 cent in 1 day. Sells this, buys another the next day with proceeds and makes a gain of 1 cent in 1 day again. Repeats this for 1 period.
If 1 period = 1 year (365 days)
Annualise using geometric compounding (XIRR): (1 + 3.65)^1 - 1 = 365%
Annualise using arithmetic compounding: 365% x 1 = 365%
If period is shorter, the difference becomes bigger, as we already know.