FIRECalc problem?

[jdw_fire]

Dory,

I made the following advanced FIRECalc run to support said post:

A FIRECalc run using the numbers: $42,000 annual W/D; 40 yr plan length; your SS of $17,800/yr starting in 2016; wife's SS of $5900/yr starting in 2020; fixed pension of $44,088/yr starting in 2009; $40,000 current portfolio invested at the defaults; starting your retirement in 2009; shows a 70.8% success rate.

This could be improved if you put some more money into your portfolio between now and 2009, reduced your expected retirement expenses or worked longer (either part time or full time).

Later, without changing anything else, I went to the results tab and pushed the "how much you spend each year" button and got 100% success for a $6400/yr W/D.  Wanting to see what some other success rates would produce I changed the percentage in the line above to 80% then to 70% then to 60%.  The answer always came back as $6400/yr W/D had the selected success rate.  Since the run I started with produced a 70.8% success rate for a $42,000/yr W/D either I am not understanding the choices on the results tab or there is an error in the software.  Please help me out here.  Thanks.

 

[dory36]

An attempt to optimize performance is creating the misleading results. Here's what is happening...

It takes about 25 times as much server processing to find a withdrawal (or portfolio) amount that will achieve a given success rate than it does to find a success rate given the other information . . . IF I limit the withdrawal range to 0-16%. To expand that range would require server processing that would make the system unacceptably slow, especially for multiple simultaneous users.

Since the only figure being adjusted is the base withdrawal amount, and the maximum anticipated withdrawal of 16% of a starting $40,000 portfolio is $6,400, that's the highest it will attempt -- which is why you kept getting $6,400 and 100% when you put in lower success rates.

As ou can see, the "search for a withdrawal" logic wasn't designed for a case where the proposed withdrawals in a single year are a very large percentage of the total portfolio.

However, the basic functioning still works, and it just takes 3-4 manual runs to get a solution.

You found taking the 42,000 w/d proposed was getting 70%. I roughly halved the $42,000 and got 100% at $20,000, then split the difference and got 100% at $30,000. $31k drops to about 95%.

I'll look at adding a footnote somewhere...

 

[cute fuzzy bunny]

Why not just wait until nobody is doing anything on the forum, then calculate every possible scenario and outcome, and store it in a table.

Then when someone plugs in their numbers, all you have to do is pull that number out of the table and display it.

Since I have no idea what actually goes on inside the tool, I have to presume that this would be incredibly easy and take virtually no effort to implement.

Its probably also useful if I put a ridiculous time limit on the project, so can I have this before I'm done with my lunch?

 

[dory36]

Would you like jalapenos with your nacho cheese, while I am at it?

 

[cute fuzzy bunny]

Absolutely. Pickled. Sliced thin and a LOT of them, please.

 

[jdw_fire]

An attempt to optimize performance is creating the misleading results.  Here's what is happening...

It takes about 25 times as much server processing to find a withdrawal (or portfolio) amount that will achieve a given success rate than it does to find a success rate given the other information . . . IF I limit the withdrawal range to 0-16%.  To expand that range would require server processing that would make the system unacceptably slow, especially for multiple simultaneous users. 

Since the only figure being adjusted is the base withdrawal amount, and the maximum anticipated withdrawal of 16% of a starting $40,000 portfolio is $6,400, that's the highest it will attempt -- which is why you kept getting $6,400 and 100% when you put in lower success rates.

As ou can see, the "search for a withdrawal" logic wasn't designed for a case where the proposed withdrawals in a single year are a very large percentage of the total portfolio.

However, the basic functioning still works, and it just takes 3-4 manual runs to get a solution.

You found taking the 42,000 w/d proposed was getting 70%. I roughly halved the $42,000 and got 100% at $20,000, then split the difference and got 100% at $30,000. $31k drops to about 95%.

I'll look at adding a footnote somewhere...

Thanks for the explanation!