Two questions about FireCalc

[AAFGJE]

Hello,

I have been using firecalc for a while. I appreciate the tool. I have two questions.

(1) When I use the monte carlo portfolio simulation tool (average and standard deviation return) my returns vary wildly from simulation to simulation. How many iterations of the simulation does that tool do? It seems to not work very well.

(2) When I do that "add certain dollar to portfolio" and then simulate when I can retire, what point in the year the resulting graph assume you retire? For example, if it has a 96% chance of success retiring in 2026, does that assume that I make my full contributions for 2024 (i.e., not accounting for us being 1/4 of the way through 2024) and 2025 and 2026 (and resultingly assume that I retire on December 31, 2026?

Thanks.

 

[MichaelB]

Hello,

I have been using firecalc for a while. I appreciate the tool. I have two questions.

(1) When I use the monte carlo portfolio simulation tool (average and standard deviation return) my returns vary wildly from simulation to simulation. How many iterations of the simulation does that tool do? It seems to not work very well.

(2) When I do that "add certain dollar to portfolio" and then simulate when I can retire, what point in the year the resulting graph assume you retire? For example, if it has a 96% chance of success retiring in 2026, does that assume that I make my full contributions for 2024 (i.e., not accounting for us being 1/4 of the way through 2024) and 2025 and 2026 (and resultingly assume that I retire on December 31, 2026?

Thanks.

Dory, the author of FIRECalc, made it clear when the tool was released that the Monte Carlo simulation was not a focal point or strength. See this post of his https://www.early-retirement.org/forums/f36/question-on-firecalc-monte-carlo-24202.html#post449961

As for date, FIRECalc assumes a Jan 1 retirement date.

 

[AAFGJE]

Dory, the author of FIRECalc, made it clear when the tool was released that the Monte Carlo simulation was not a focal point or strength. See this post of his https://www.early-retirement.org/forums/f36/question-on-firecalc-monte-carlo-24202.html#post449961

As for date, FIRECalc assumes a Jan 1 retirement date.

Interesting. So if fire calc says I have a 98% chance of success in 2027, that means that Firecalc assumes I will contribute for the remainder of 2024, all of 2025, and all of 2026?

Do you have any suggestions for monte carlo simulation tools that work better than Firecalc?

 

[ERD50]

... Do you have any suggestions for monte carlo simulation tools that work better than Firecalc?

Why would you want Monte Carlo simulation? To me it makes no sense for this case.

MC is so circular - when you ask the developers how they determine what limits to set on the variation of each element (stock, fixed, inflation, etc), they say they adjust them until it matches historical results.

Just use historical results (the real power of FIRECalc and FICalc.app). Forget Monte Carlo for this (it has good applications in other fields).

-ERD50

 

[AAFGJE]

Why would you want Monte Carlo simulation? To me it makes no sense for this case.

MC is so circular - when you ask the developers how they determine what limits to set on the variation of each element (stock, fixed, inflation, etc), they say they adjust them until it matches historical results.

Just use historical results (the real power of FIRECalc and FICalc.app). Forget Monte Carlo for this (it has good applications in other fields).

-ERD50

There are two reasons I want to use monte carlo.

(1) I think historical US results (especially post-WWII) are a historical anomaly. I don't trust that historical results will hold up; I believe that over my lifetime, the likely return in the market is closer to 4-5% above inflation than 6-8%. I can't backtest that against historical data.

(2) Even if (1) was not the the case, I do not believe that I can backtest against historical data anyway, because there are not enough historical time periods to test against over the length of my lifetime. I certainly do not trust testing against pre-WWII data (the world is too different), which leaves me with only 30-40 60-year periods to test against (my likely lifespan). That's hardly enough of a sample size.

So Monte Carlo makes more sense to me.

 

[calmloki]

Would raising the inflation rate do what you want? Just jack the inputs for it up 2-3%

Feel like a unique flower myself with no pension and very few social security taxed "earnings" since our income has predominately been from our rentals, so I use firecalc as a shotgun. figure if most of the pellets hit the target with my bogus inputs I'm probably good. T'aint nothing set in this world, firecalc won't fund your living expenses if you run out of ducats. Nor should it.

 

[Closet_Gamer]

I believe both Fidelity and Schwab retirement planners are rooted in Monte Carlo methodologies.

We've done these in addition to Firecalc. I found them all to have strengths/weaknesses. When all three said "good to go" in addition to my own modeling, logic, etc. it was sufficient to me.

I understand the concerns with Firecalc but I do think the brute force "would your plan have survived every period of x years over the last century" is pretty elegant and instructive.

YMMV

 

[ERD50]

There are two reasons I want to use monte carlo.

(1) I think historical US results (especially post-WWII) are a historical anomaly. I don't trust that historical results will hold up; I believe that over my lifetime, the likely return in the market is closer to 4-5% above inflation than 6-8%. I can't backtest that against historical data.

(2) Even if (1) was not the the case, I do not believe that I can backtest against historical data anyway, because there are not enough historical time periods to test against over the length of my lifetime. I certainly do not trust testing against pre-WWII data (the world is too different), which leaves me with only 30-40 60-year periods to test against (my likely lifespan). That's hardly enough of a sample size.

So Monte Carlo makes more sense to me.

The only real question with historical is: Will your future retirement period be worse than the worst scenario in history?

Since you are looking to be conservative anyway, accept nothing less than 100% with historical data. Then, the question above applies. It doesn't make any difference if the average of the future is worse than the average of the past, just that the worst is no worse than the worst of the past.

Of course, the future could be worse than the worst of the past. OK, through in some buffer on your spending or portfolio size. 10%, 20%, 30% whatever you are comfortable with. At least at that point you have some basis for your decision.

Monte Carlo is just throwing numbers around, we have no idea if that represents any reality at all, unless we set limits of variability based on.... history!


-ERD50