Hey Rob, thanks for including us in your video! We noticed a couple of things we’d love to clarify and would be happy to walk you through how everything works. No black boxes here-just full transparency. Let us know if you're interested!
Perhaps full transparency would be to just note the "couple of things" in your comment for everyone to see? If projectionlab needs to clarify a walk through with a smart guy like Rob, then what does that say about users or potential users like me that are on a secondary education level. Seems like a pretty good OODA loop feedback to improve projection lab?
@@hanwagu9967 I can’t speak for ProjectionLab, but as a user, I can see that Rob wasn’t actually using the default settings like he claimed. They offered full transparency about how the tool works, but I can see why they might not want to point out his mistakes.
Hi Rob - thanks for doing this video and including us. Happy to jump on and answer your questions and/or come on and talk to your audience if that is helpful. All feedback is always welcome (positive or negative) and help us make the platform better. We've built Boldin with our community and that will always be the core of our work.
01.07.25 Hi @stephenchen2098 I just signed up on boldin and would like to learn/understand how to use the platform before my trial period ends. Is there a place or boldin provide free tutorial? One feedback right off the start is that the platform assumes users will receive monthly social security benefits. I wish I would but I know I will not.
Fantastic topic to cover Rob! Thank you for taking this on. I have been doing something similar myself, using Boldin and ProjectionLab, and now with a Financial Planner also using Right Capital… and I have discovered similar results as you. Namely that the outputs can often be significantly different between the tools despite efforts to run similar scenarios. Time sync issues such as you found are a problem. Differing assumptions, or exposing different variables to adjust can also be a problem. And ultimately there are black box issues as you mention in your conclusion. Once scenarios become ever n more complex - as is the case with Roth conversion explorations, it makes one wonder how much stock to put in what the tools are projecting. Thank you on behalf of ALL of us DIY’ers for running tests like this and hopefully giving us all a bit more transparency and confidence into how much we can “trust” the results these tools are displaying.
Absolutely love these comparisons. Please keep them coming. Funny, but it sounds like we need a Monte Carlo simulation of various Monte Carlo results. 😄
Imagine paying a CFP thousands of dollars to print out a software generated reports full of Type II errors. Now you're either working a few more years unnecessarily or worse run out of money. Great job on this and let us know what you find out!!!
Robert - check how the Monte Carlo simulations handle future volatility. Some use historical returns for the forward projections, some use JP Morgan forward looking projections, some use vanguard forward looking projections and some maybe something else.
Perfect timing for this review Rob. I purchased Pralana Online last month and want to purchase either Boldin and/or ProjectionLab to compare results. Cost isn’t the issue but it is the investment of time becoming proficient with these tools to determine if they meet one’s needs. Hopefully the developers will respond so you can continue with your informative videos! Many thanks, looking forward to more videos.
Excellent timing, I have just been playing with multiple financial planning, software models and noticed the discrepancy in chance of success scores. I assumed it was a data entry error on my part. Thanks for investigating this issue!
Excellent video, Rob! I love the idea of comparing these three planning tools. I look forward to learning what feedback you get from each of them and I also look forward to more comparison videos.
Thanks, Rob for a great video and analysis. I like how you break things down. I've been using all four of these tools as well and have had similar thoughts, questions, and experiments. Glad to see I'm not alone :)
A phrase I use at work is "The purpose of a report is to drive decision making." From these tools, what I would want is information that informs my decisions. If I get back that I have a 75% success rate, I want to know what conditions led to the other 25%, this that is what drives the action I should take. Are those stagflation cases? equity crash? Those would lead me to different actions. That implies output that can allow introspection of the simulation cases.
Great video content! I will be sure to watch your future videos on this topic. It is very important for us to understand why the chances of success vary so much.
Boldin's youtube page has a 5 minute video that gets into some of the technical details of its MC simulation. If I understood correctly, their simulation varies rate of return on a monthly basis. They also model standard deviation based on the rate of return assumption for each account. An account with a zero percent return assumption gets zero standard deviation (cash-like) while an account with (say) a 9% return assumption would get a larger std deviation that could generate a significant negative return in a given month for a MC trial.
@@ItsEverythingElse Sure...but the key detail is how Boldin approaches the standard deviation as that is a significant assumption in any MC calc. Of course you'd need to compare Boldin's SD assumption to that of the other software to see assess if it explains some of the differences in probability of success versus the alternative software products.
@@patrickd9576 Yes, the differences between Boldin and Projection Labs (roughly 65% and 95% success rates) mimic a 1 and 2 sigma Gaussian distribution population... and the speed of the update makes me think they are both approximating the MC simulations, but using different criteria for their final success parameter.
Rob, most of the difference between the tools probably comes from how they internally model volatility within their Monte Carlo. I'd guess that volatility is injected differently into each parameter within their black boxes for rate of return, inflation, interest rates, etc. I developed a spreadsheet that lets me see how much volatility affects returns. I can model a projected flat growth rate, such as 8%, on a starting balance, or I can use the historical total returns of the S&P 500 Index from whatever period I select. I can apply its volatility to my projected growth rate as a percentage while keeping the overall growth rate for the entire period unchanged. If I use 100% of the S&P Index volatility, that is effectively like asking, "What if over the next 30 years, the average growth is 8%, but that growth is as volatile as the S&P was during some prior 30-year period? No matter which historical 30-year period I choose, the ending balance is always lower than if the growth rate had been flat. As I dial down the volatility (i.e., 50% as volatile as the S&P), the ending balance climbs until it reaches the maximum at 0% volatility. If I dial up the volatility above 100% of the historical S&P, the ending balance will be significantly lower even when I set inflation to zero, with no withdrawals or additions to the account.
@@mindyfrohlich6187 My spreadsheet uses formulas that compare the average rate of return over the chosen time span to a table of the specific returns for each year to see how far each year is from the average. If I only want to see 70% (for example) of the historic volatility then I take 70% of that difference and enter that in a new column, which I then chart. It would make more sense if I could provide specific examples, but then this response would be too long.
@@mindyfrohlich6187 He probably has a separate box on the spreadsheet that he can multiple or divide the volatile percentage by. So if something grows or declines by X percent in a given year (or time frame), it is multiplied by Y, which can be less than one (i.e. .75 or .5) to dial down the volatility, or > 1 to dial up the volatility. It's a pretty cool idea: model an actual 30 (or any) year period based on growth/decline of S&P (or any "standard"), rather than model a consistent fixed rate of return (i.e. 5%, 6%, etc). I still haven't tried it, but I intend to.
Replicated Rob's Roth IRA result in PL ($10.062M). Using a compound interest calculator (Nerd Wallet), if I compound Annually I get PL's result to the dollar ($10,062,657), excluding the $1/mo expense item. However, if I compound Monthly, the calculator returns $10,935,730, a difference of almost $900K. I couldn't find anywhere in PL where compounding schedule can be modified, so you get what you get. Guessing the compounding assumptions might vary across the different software platforms - this simple example illustrates how that can cause a nearly 10% discrepancy in an asset's future value. In fairness to PL, the information bubble in the Investment Growth Rate entry does indicate annual compounding.
The longer the time frame and the higher the interest the more discrepancy you will have between monthly and annual compounding. But it tends to balance out on the expenses side. Also, consider that your entire portfolio is not generally compounding monthly anyway. In reality the discrepancy will be far less than the 9% in Rob's simple test.
Excellent comparison. I think you hit on the major question near the end of the video. If they all have similar assets balances at the end of the projection , why different success values. It seems their projection algorithms are similar so I guess the question is; How do you define success?
Apparently no two Monte Carlo tools are the same. I can’t provide a link to the article (TH-cam doesn’t like that?) but you might wish to search for an article like this: How Different Monte Carlo Models Perform In The Real World: Assessing Quality Of Predictiveness In Retirement Income Forecasting Models The authors discuss the 4 major categories.
Thank you, Rob, this was an excellent topic to explore. The results are both surprising and concerning. The silver lining for me was to see Boldin turning out as the most pessimistic of them all, and that's the one I've been using. Hopefully that means that if Boldin projects a healthy percentage for our scenario, other programs would show an even better chance of success.
Great video Rob! Glad you're pointing out these variances - might be a way to bound one's chance of success - use the most conservative and the most optimistic. ;-)
Great comparison test! Of course, the same variance in guidance and results would happen using various financial advisors at a much, much higher cost than using any of this software. At least I now know that Boldin is the most conservative!
I have my plan in three different calculators with Boldin being the most pessimistic. One significant difference is the expected LTC costs in the Boldin plan.
@@johnfarquhar4080 This is what i simply do not understand in Boldin. The medical care costs seem absurdly high which greatly impact amount in retirement. Between my wife and I it claims nearly $700K in medical costs. $28K a year every year after retirment? LOL, i dont think so. And since i cant edit it I bailed and went to projection lab.
Great video Rob. I have all three tools and I agree, entering info in all of them is different and hard for some of them (takes time). And numbers come out different. Let us see what they have to say. I got Parlana only this week. I do like it as it shows all the numbers being computed. All these tools do need you spending time learning their ins and outs.
I am at my wits end with these tools, and beginning to realize that maybe a financial planner is in my future as my patience has worn thin. (Lol this sounds like one of those scam intros but it's not). I like to think I'm a smart guy but there's a lot of things I don't understand and I hate when I make a simple change and then suddenly my retirement chance of success drops significantly. At least with bolden it tells you what your change's impact was, but there are things in that that just make no sense either. I currently am using projectionlab and the support is great but after using the others I can't help but to think it's not "guiding me" enough. Anyway, I love that these tools exist for us layman but I definitely have more respect for financial planners at this point.
@@Mattxwill1 If you learn to make spreadsheets dance, you can control a ton of factors & ranges, and see the effects each has (inflation, budgets, social security, IRMAA, tax brackets, rates of return, etc). It takes some time in the beginning, but the time investment is worth it, and you can see how some seemingly large changes don't affect things significantly, yet other small changes affect things a lot, especially early on & over a long period of time. I'm also frustrated by the "black box" effect of these tools, and lack of transparency. My desire to know *why* things work the way they do, and to have *control* over it made me do the spreadsheet thing. (Then again, I was in IT for a brokerage for 37+ years before retiring, so that's how my brain operates.) Good luck with your planning.
Rob, I think I know why ProjectionLab has the highest chance of success, but the lowest ending dollar value of the three programs, which you mentioned at 14:32 in the video. Both the chance of success and the portfolio ending value are based on the capital market assumptions (investment returns) used by the software. Most likely, Projection lab is using investment return assumptions for the assets that you modeled that are lower than the other two programs. This should explain the lower terminal value of the portfolio in ProjectionLab. ProjectionLab is likely using a smaller standard deviation for the investment returns than Boldin and Pralana. This results in smaller return variations year-over-year in the annual returns randomly selected in he Monte Carlo calculations. Over a 30-year time horizon, this dampens the volatility of the portfolio which in the case of your particular plan results in the chance of success score being higher than the other two programs.
This is great Rob, I have learned so much from your channel and appreciate you taking the time to ask these important questions! I had an advisor tell me I had a 58% chance and wanted to learn why that was so started figuring out how to project rates myself. I still have too many unknowns in my future to know my retirement spending but want my projection tool to be helpful when I posit the possibilities! 😅 It certainly would help to understand the basis for the different results.
As a Boldin user, I at least know it is the most conservative. The large differences in Chance of Success across the models is interesting and I look forward to learning more of the why.
I am familiar with PL and it shows the number of simulations that succeed or fail, if the others do as well, how do those compare? Trying to see how if the ending balance is similar, is the number of failures driving the difference in success rates?
@ True, however PL counts Large surplus and Just made it as success and Almost made it, and Failed in the middle as failures. SoI guess that rules that out as the source for the difference.
Wouldn't surprise me if the more pessimistic results are incorporating some version of the current CAPE ratio in their calculation. Or even if they all are, but they're using different formulas or different periods to look up that ratio.
A couple of things to look into. Are the distributions at the beginning of the period or the end? Are returns applied at the beginning or the end of the prriod? Is inflation applied every month or at the end or beginning of the period? When are taxes paid? What data is used for the Monti Carlo (btw I was there last week)? Also, you need to start at the same point, not some late 2024 the other 2025, it makes a difference!
Rob, Thanks for all your efforts in educating folks like me. I question the use case for this video. As a user of one of these retirement planning tools, why/when am I going to compare 3 different tools for the feel good (or bad) number? It may be helpful for these 3 vendors, if they've not already done this kind of comparison. What will be helpful is to create a list of specific real life use cases - Roth conversion, Withdrawal optimization to name a couple & try them out in these tools and rank them. Appreciate your effort!
I use both Boldin and Pralana. I like the look and "feel" of Boldin but I can see better what is going on in the background with Pralana. I'll eventually settle on one calculator in the future though, so I am looking forward to your followup on this video. Thanks.
Good evening Mr Rob, sadly enough there are so many folks who simply are clueless when it comes to finance. Your videos really help those folks. For my personal investment style I have always focused on quality and growth stocks. Today at the age of 75 I still have the same investment philosophy, keep it simple, basic and make sure you know where your money is going. Be careful with those advisors who rely on animal spirits. Keep up the good work!
Very interesting comparison Rob. For someone like myself who is considering using risk-based guardrails (which increase/decrease withdrawals based on % success), that's some serious food for thought.
I am looking forward to follow-up video(s) on this important topic! I am also curious which (if any) of these tools support different asset allocation methods in post-retirement (fixed AA, rising equity, bond tent, declining equity, etc). This is something I am currently wrangling with as I'm at the start of my retirement and SORR is top of mind, especially given where market valuations are today. Portfolio Visualizer makes an attempt at via their Financial Goals feature (can accomodate rising/declining asset allocation glide paths) but its targeted to pre-retirement and lacks the specific capabilities needed for post-retirement.
This is excellent Rob and I appreciate you reaching out to them as you will get a better response than a normal user I suspect. I use Bolden and as you say, these differences are concerning. It will be helpful to gain more insight how they work and differ to help explain the dirrerent results. Larry, Central Valley, Ca.
Rob should have looked at and showed what the modeling settings are in ProjectionLab - in ProjectionLab you can use historical returns for example. Not clear to me which settings he has.
@@rob_bergergiven there was higher chance of success, but lower ending balances, I wonder if the differences will come down to volatility metrics and assumed std dev.
The assumptions that each model used at each point in history were produced systematically in ways that mirror actual approaches to modeling and CMAs available to and used by advisors today, and are as follows: Traditional Monte Carlo: One set of CMAs applies to all years in the plan: In our analysis of this approach, CMAs that are the average portfolio returns and standard deviation of returns from the preceding 30 years were applied to each point of time examined; Reduced-CMA Monte Carlo: This is similar to Traditional Monte Carlo, except that CMAs are reduced by 2% and standard deviation is reduced proportionally; Historical Analysis: Actual historical sequences of returns and inflation models the range of possible return and inflation sequences. The assumptions for this analysis used all history available up to each point in time; and Regime-Based Monte Carlo: 2 sets of capital market assumptions - one that applies to the near-term and one that applies to the long-term - are used to produce simulated return and inflation sequences. The assumptions exclude half of all prior points from each point in history based on their economic dissimilarity to the point in time being examined. From this filtered set of historical prior points, averages from the first 10 years are used to produce near-term CMAs, and averages from the next 20 years are used to produce long-term CMAs.
Having a 60% only chance of success when the withdrawal rate is 4.8% doesn't make much sense and seems to be missing the forest for the trees. Sure, the assumptions are the most likely reason but right there one would expect a higher chance of success at that withdrawal rate
I agree with the comments that methods used for MC simulation are likely very different. This video inspired me to do a test in Boldin. I set up a scenario that had a single roth 401k will 1 million and there were no withdrawals or deposits, just like in the video. At the end of the timeline (37 years) the 80% confidence interval from the MC simulation was (820k, 2.1 mil). That means there is a 10% chance of getting and ending balance below 820k and a 10% chance of being above 2.1 mil.. The I created a scenario with 4 roth 401k accounts each with 250k in them. So both scenarios had the same starting balance and return assumptions. However the second scenario 80% confidence interval was (1.1 mil, 1.8 mil). This is evidence that Boldin assumes that the investment accounts returns are not 100% correlated with each other. I would like to know from Boldin what correlation assumption they make. That has a big impact on planning. Going forward I will enter balance information into consolidated accounts in Bolding. For example, I will combine both my wife's 401k and mine into a single 401k account since they have the same distribution of assets.
Excellent idea & test. One would think that the numbers should have come out exactly the same. So there seems to be some kind of random number generator in there. Did you try running each test multiple times to see if there were any variations?
@@everlastingarms3065 I am not surprised at all.Separate accounts would provide diversification if they held different assets. The problem is that I have multiple accounts that I cannot consolidate that have about the same asset distribution. Since Boldin does hot have an concept of assets, it only has "accounts", the best we can do is to work around it. Re-running does not change the confidence intervals much at all
@@everlastingarms3065 By definition Monte Carlo is random. Unless two packages use the exact same settings and algorithm they will never produce the exact same results.
In Boldin, how do I create a test plan when I already have a baseline plan with account connections, etc? I tried creating a scenario off the Baseline plan, but it replicates all the connected accounts and won’t let me delete them from the scenario (along with other restrictions). PL is much more obvious, you just create a new plan. Is there a way to save a Boldin Baseline plan so I can paste it back in later?
Rob, based on on various videos, I have uploaded my financial soul into Boldin, trying to model a Roth Conversion strategy. Which tool do you recommend for a 2nd opinion ? I am concerned my inputs/assumption in Boldin are inaccurate - I always like to trust but verify.
Thanks. Interesting. I tried last week to vompare the Vanguard projections with the free Empower program and the paid MaxiFi. Not rasy to do and rather different results, with the empower being the most pessimistic.
I use Boldin and know that there is a default setting for long term care expense (I saw someone mentioned this already). Wonder if other software includes default expenses.
I’ve heard Moshe Milevsky say most of the companies don’t understand at all deep level how their own Monte Carlo calculators work- in terms of the derivatives and other functions they have baked in. They use “off the shelf” models that may not be applicable
Great videos!! I didn't hear you talk about pricing. I have Boldin that for the $130/yr, they are GREAT!!! How about the others? Overall, how are the others with Roth Conversion analysis, scenarios, etc?
I have PV eMoney and also use Empower. I just cancelled Boldin after a couple years and am considering Pralana (reading key parts of the manual). To me eMoney lacks flexibility due to needing an advisor to make updates. It's a little old school. The value of your PV subscription is having access to great folks and also a tracking tool for your plan. Boldin had great UI but I couldn't trust the results and QA can be lacking. Pralana may be old school as well, but it's much finer grain than eMoney. I will keep watching Rob's videos to inform my purchase.
I wonder if there is some uncertainty in how "running out of money" is defined? Maybe they are using the first or last days of the target year to determine that? If each simulation carries a balance and an uncertainty, do they define zero as the balance, or the (balance - uncertainty) ? If they use the uncertainty, is it one sigma like the previous sentence, or three sigma? The good news is that the projected balances are nearly the same. Thank-you for doing these tests... definitely interested in your follow-up results!!! PS - As a follow-up, the Boldin success rates near 65% are very close to the 1-sigma values of a Gaussian, and the Projection Lab results at 93% are close to the 2-sigma value of a Gaussian distribution. Since they give answers so quickly, my gut feeling is that they are not running 1000 Monte Carlo simulations with each update, but rather making an estimate, and using different criteria for their uncertainties... 1 and 2 sigma respectively.
Some software calculates things monthly and others calculate yearly and that can account for much of the different results. And I think they may handle tax rates quite differently so the differences can become even larger when taxes are starting being included. As far as Monte Carlo simulations go, the success scores might be based on different percentile ranges. They SHOULD be based on the 0-100 range but if they are not (might be 20-80 or something) then the score will be very different. Also, Boldin only uses 1000 iterations in their Montte Carlo so their result may vary a bit more from run to run. But that shouldn't necessarily make it wildly inaccurate.
@@GregoryLooney Right. There are so many variables. Not to mention that we are projecting 30 years into the future when we don't even know what will happen next year!
Rob, I look forward to your videos. Thank's for all you do. Can I ask one question? My wife is 75 and still works. She has not taken RMD's yet. When must she start. Should she wait until after the new year or does it matter?
Thanks Rob. I have been looking forward to these comparison videos. Also look forward to the responses from each of the developers. The specific standard deviation for each asset class can be set in Pralana. I don't believe this is the case for Boldin and I don't think they disclose what is used. Also, in Pralana you set the real rate of return and inflation. I believe Boldin sets the nominal rate of return and inflation. You didn't show these, but I assume these were set to be apples to apples.
Boldin does specify, "The standard deviation, while not explicitly disclosed, closely mirrors historical returns." But, couldn't your optimistic and pessimistic inputs in Boldin serve as your variable deviation?
@@hanwagu9967 As I understand it, in Boldin the optimistic and pessimistic inputs serve as the assumed return for a particular investment account. The Monte Carlos simulation would then overlay a certain standard deviation on that assumed rate of return. One would need to know what that standard deviation Boldin applies, to try to match it in Pralana. Also, as I said above, Pralana assigns the standard deviation by asset class and Boldin seems to do it by account.
Hi Rob, my name is Alicia, & I had an older sister pass away in 2022, who left me some money. I contacted the financial institution, Vanguard; I informed them at the time of her death, that I would not be doing any investing. I was told that the funds were not available to me for two years, & that I needed to be put all the money into an IRA. After listening to your program a few times am learning more about other investments options, can I use the funds that from an IRA for maximum cash flow?
I see some weird numbers coming out of Boldin and there's little to no transparency about how they are coming up with it. I'm gonna stick with the spreadsheets for now! 😆
Great Video Rob. Just another reason to not rely on a single resource for planning your financial future. Looks like the Cryto Bots have found your comment page. Sad to say, but you might have to start managing your commenters/subscribers as its become pointless to sift through them for additional knowledge/insight.
That’s an excellent video, my friend. As I ponder retirement, I have to consider how much risk there is with the crazy Trump future and what might happen to the stock market. If the Trump factor wasn’t in there with nervous investors than I would likely die with $1 million. I currently use projection lab and I think Kyle has done a wonderful job with this program and is adapting it to many very smart users input.
as Rob mentioned, he was using a "client" version of RC (from Kevin Lum at Foundry - I have the same one) and it doesn't allow rate of return to be set by the user (it just has various pre-configured asset mixes). The full version of RC is only available as a professional advisor tool with limited functionality available to the client so it really isn't possible for Rob (or anyone that isn't a professional advisor) to do an apples to apples comparison.
We realize once again that "Nobody knows nothing". It's simpler to use a reasonable SWR target for planning and adjust spending as needed when retired.
From a design of experiments basis I can guess that maybe the number of runs in the simulations could have varied between them as well as the parameters they use and the data they draw on to feed the simulations , It withstanding the equations that nodes the market. The conclusion one can draw from the set of 3 is none can be trusted , because two agreeing doesn’t mean they are more accurate. They can be equally wrong. For those projections to have any merit in interpreting they need to be open source and standardized. That way the community of investment analysts can comb over the simulation equations and inputs and assumptions and make sure all the needed stuff is there, and maybe standardize weighting system to get to a common metric - like the MPG rating on cars. Reality isn’t matching it but everyone is running the same test conditions. As it stands I don’t believe any of them.
Monte carlo simulation seems way too black box to me. Who knows what assumptions are made to generate the simulated data and how reliable those assumptions are? It's revealing to see how inconsistent the results are in your video! I know backtesting has major limitations, but at least it's easy to understand and reason about.
I think that maybe we are asking too much of these tools. They are not crystal balls. Having said that, they should be transparent about how the chance of success number is generated and what expected returns and volatility they are assuming. These tools are most useful to give a low-cost, broad -strokes look at your retirement spending plans to make sure you are not delusional about being able to retire without going broke before you die. Optimization just asking too much for something as uncertain as the future.
Hey Rob, thanks for including us in your video! We noticed a couple of things we’d love to clarify and would be happy to walk you through how everything works. No black boxes here-just full transparency. Let us know if you're interested!
I'm glad to see this and I have the lifetime subscription with ya'll. I'm a happy customer.
Perhaps full transparency would be to just note the "couple of things" in your comment for everyone to see? If projectionlab needs to clarify a walk through with a smart guy like Rob, then what does that say about users or potential users like me that are on a secondary education level. Seems like a pretty good OODA loop feedback to improve projection lab?
Interested as well to better understand about the delta between different solution as well ;)
Also interested in the couple of things to be included in the comments.
@@hanwagu9967 I can’t speak for ProjectionLab, but as a user, I can see that Rob wasn’t actually using the default settings like he claimed. They offered full transparency about how the tool works, but I can see why they might not want to point out his mistakes.
Hi Rob - thanks for doing this video and including us. Happy to jump on and answer your questions and/or come on and talk to your audience if that is helpful. All feedback is always welcome (positive or negative) and help us make the platform better. We've built Boldin with our community and that will always be the core of our work.
01.07.25 Hi @stephenchen2098 I just signed up on boldin and would like to learn/understand how to use the platform before my trial period ends. Is there a place or boldin provide free tutorial? One feedback right off the start is that the platform assumes users will receive monthly social security benefits. I wish I would but I know I will not.
Fantastic topic to cover Rob! Thank you for taking this on. I have been doing something similar myself, using Boldin and ProjectionLab, and now with a Financial Planner also using Right Capital… and I have discovered similar results as you. Namely that the outputs can often be significantly different between the tools despite efforts to run similar scenarios. Time sync issues such as you found are a problem. Differing assumptions, or exposing different variables to adjust can also be a problem. And ultimately there are black box issues as you mention in your conclusion. Once scenarios become ever n more complex - as is the case with Roth conversion explorations, it makes one wonder how much stock to put in what the tools are projecting. Thank you on behalf of ALL of us DIY’ers for running tests like this and hopefully giving us all a bit more transparency and confidence into how much we can “trust” the results these tools are displaying.
Absolutely love these comparisons. Please keep them coming. Funny, but it sounds like we need a Monte Carlo simulation of various Monte Carlo results. 😄
Imagine paying a CFP thousands of dollars to print out a software generated reports full of Type II errors. Now you're either working a few more years unnecessarily or worse run out of money. Great job on this and let us know what you find out!!!
More like 1% AUM every year for Type II errors!
A man with 1 watch knows the time. A man with 2 is never sure 😂. Good info, keep it coming!
Robert - check how the Monte Carlo simulations handle future volatility. Some use historical returns for the forward projections, some use JP Morgan forward looking projections, some use vanguard forward looking projections and some maybe something else.
Perfect timing for this review Rob. I purchased Pralana Online last month and want to purchase either Boldin and/or ProjectionLab to compare results. Cost isn’t the issue but it is the investment of time becoming proficient with these tools to determine if they meet one’s needs. Hopefully the developers will respond so you can continue with your informative videos! Many thanks, looking forward to more videos.
Excellent timing, I have just been playing with multiple financial planning, software models and noticed the discrepancy in chance of success scores. I assumed it was a data entry error on my part. Thanks for investigating this issue!
Excellent video, Rob! I love the idea of comparing these three planning tools. I look forward to learning what feedback you get from each of them and I also look forward to more comparison videos.
Thanks, Rob for a great video and analysis. I like how you break things down. I've been using all four of these tools as well and have had similar thoughts, questions, and experiments. Glad to see I'm not alone :)
A phrase I use at work is "The purpose of a report is to drive decision making." From these tools, what I would want is information that informs my decisions. If I get back that I have a 75% success rate, I want to know what conditions led to the other 25%, this that is what drives the action I should take. Are those stagflation cases? equity crash? Those would lead me to different actions. That implies output that can allow introspection of the simulation cases.
Great video content! I will be sure to watch your future videos on this topic. It is very important for us to understand why the chances of success vary so much.
Boldin's youtube page has a 5 minute video that gets into some of the technical details of its MC simulation. If I understood correctly, their simulation varies rate of return on a monthly basis. They also model standard deviation based on the rate of return assumption for each account. An account with a zero percent return assumption gets zero standard deviation (cash-like) while an account with (say) a 9% return assumption would get a larger std deviation that could generate a significant negative return in a given month for a MC trial.
But in Rob's test he only has one account.
@@ItsEverythingElse Sure...but the key detail is how Boldin approaches the standard deviation as that is a significant assumption in any MC calc. Of course you'd need to compare Boldin's SD assumption to that of the other software to see assess if it explains some of the differences in probability of success versus the alternative software products.
@@patrickd9576 Yes, the differences between Boldin and Projection Labs (roughly 65% and 95% success rates) mimic a 1 and 2 sigma Gaussian distribution population... and the speed of the update makes me think they are both approximating the MC simulations, but using different criteria for their final success parameter.
Great video and comparison. We are basing our future on these tools and if they are not accurate it could drastically change our future success.
Rob, most of the difference between the tools probably comes from how they internally model volatility within their Monte Carlo. I'd guess that volatility is injected differently into each parameter within their black boxes for rate of return, inflation, interest rates, etc.
I developed a spreadsheet that lets me see how much volatility affects returns. I can model a projected flat growth rate, such as 8%, on a starting balance, or I can use the historical total returns of the S&P 500 Index from whatever period I select. I can apply its volatility to my projected growth rate as a percentage while keeping the overall growth rate for the entire period unchanged. If I use 100% of the S&P Index volatility, that is effectively like asking, "What if over the next 30 years, the average growth is 8%, but that growth is as volatile as the S&P was during some prior 30-year period? No matter which historical 30-year period I choose, the ending balance is always lower than if the growth rate had been flat. As I dial down the volatility (i.e., 50% as volatile as the S&P), the ending balance climbs until it reaches the maximum at 0% volatility. If I dial up the volatility above 100% of the historical S&P, the ending balance will be significantly lower even when I set inflation to zero, with no withdrawals or additions to the account.
excellent insights. thank you
This is ace, and makes perfect sense. Good job. And now I'm going to do likewise with my own spreadsheets lol. Thanks for the idea.
great insight. How do you adjust the volatility?
@@mindyfrohlich6187 My spreadsheet uses formulas that compare the average rate of return over the chosen time span to a table of the specific returns for each year to see how far each year is from the average. If I only want to see 70% (for example) of the historic volatility then I take 70% of that difference and enter that in a new column, which I then chart. It would make more sense if I could provide specific examples, but then this response would be too long.
@@mindyfrohlich6187 He probably has a separate box on the spreadsheet that he can multiple or divide the volatile percentage by. So if something grows or declines by X percent in a given year (or time frame), it is multiplied by Y, which can be less than one (i.e. .75 or .5) to dial down the volatility, or > 1 to dial up the volatility.
It's a pretty cool idea: model an actual 30 (or any) year period based on growth/decline of S&P (or any "standard"), rather than model a consistent fixed rate of return (i.e. 5%, 6%, etc).
I still haven't tried it, but I intend to.
Replicated Rob's Roth IRA result in PL ($10.062M). Using a compound interest calculator (Nerd Wallet), if I compound Annually I get PL's result to the dollar ($10,062,657), excluding the $1/mo expense item. However, if I compound Monthly, the calculator returns $10,935,730, a difference of almost $900K. I couldn't find anywhere in PL where compounding schedule can be modified, so you get what you get. Guessing the compounding assumptions might vary across the different software platforms - this simple example illustrates how that can cause a nearly 10% discrepancy in an asset's future value. In fairness to PL, the information bubble in the Investment Growth Rate entry does indicate annual compounding.
The longer the time frame and the higher the interest the more discrepancy you will have between monthly and annual compounding. But it tends to balance out on the expenses side. Also, consider that your entire portfolio is not generally compounding monthly anyway. In reality the discrepancy will be far less than the 9% in Rob's simple test.
Excellent comparison. I think you hit on the major question near the end of the video. If they all have similar assets balances at the end of the projection , why different success values. It seems their projection algorithms are similar so I guess the question is; How do you define success?
Apparently no two Monte Carlo tools are the same. I can’t provide a link to the article (TH-cam doesn’t like that?) but you might wish to search for an article like this:
How Different Monte Carlo Models Perform In The Real World: Assessing Quality Of Predictiveness In Retirement Income Forecasting Models
The authors discuss the 4 major categories.
Thank you, Rob, this was an excellent topic to explore. The results are both surprising and concerning. The silver lining for me was to see Boldin turning out as the most pessimistic of them all, and that's the one I've been using. Hopefully that means that if Boldin projects a healthy percentage for our scenario, other programs would show an even better chance of success.
Great video Rob! Glad you're pointing out these variances - might be a way to bound one's chance of success - use the most conservative and the most optimistic. ;-)
Great comparison test! Of course, the same variance in guidance and results would happen using various financial advisors at a much, much higher cost than using any of this software. At least I now know that Boldin is the most conservative!
But we need to understand what makes it "conservative". Math is math. Are they applying some kind of adjustments that we don't know about?
In Boldin, look at default setting for long term care
I have my plan in three different calculators with Boldin being the most pessimistic. One significant difference is the expected LTC costs in the Boldin plan.
@@johnfarquhar4080 This is what i simply do not understand in Boldin. The medical care costs seem absurdly high which greatly impact amount in retirement. Between my wife and I it claims nearly $700K in medical costs. $28K a year every year after retirment? LOL, i dont think so. And since i cant edit it I bailed and went to projection lab.
@@Mattxwill1 Nancy from Boldin here. You can always enter your own Medicare costs in Boldin. Look for the "Itemized Medicare" toggle.
Another great argument for using multiple tools. I use several tools to eliminate any strange biases of the black boxes or on my part.
Rob, aren't there multiple types of Monte Carlo simulations that can be employed by the various providers? Might this explain the differences?
Great video Rob. I have all three tools and I agree, entering info in all of them is different and hard for some of them (takes time). And numbers come out different. Let us see what they have to say. I got Parlana only this week. I do like it as it shows all the numbers being computed. All these tools do need you spending time learning their ins and outs.
I am at my wits end with these tools, and beginning to realize that maybe a financial planner is in my future as my patience has worn thin. (Lol this sounds like one of those scam intros but it's not). I like to think I'm a smart guy but there's a lot of things I don't understand and I hate when I make a simple change and then suddenly my retirement chance of success drops significantly. At least with bolden it tells you what your change's impact was, but there are things in that that just make no sense either. I currently am using projectionlab and the support is great but after using the others I can't help but to think it's not "guiding me" enough. Anyway, I love that these tools exist for us layman but I definitely have more respect for financial planners at this point.
@@Mattxwill1 If you learn to make spreadsheets dance, you can control a ton of factors & ranges, and see the effects each has (inflation, budgets, social security, IRMAA, tax brackets, rates of return, etc). It takes some time in the beginning, but the time investment is worth it, and you can see how some seemingly large changes don't affect things significantly, yet other small changes affect things a lot, especially early on & over a long period of time.
I'm also frustrated by the "black box" effect of these tools, and lack of transparency. My desire to know *why* things work the way they do, and to have *control* over it made me do the spreadsheet thing. (Then again, I was in IT for a brokerage for 37+ years before retiring, so that's how my brain operates.)
Good luck with your planning.
Great topic Rob. Look forward to the follow-up. As usual, your comments are excellent.
Rob, I think I know why ProjectionLab has the highest chance of success, but the lowest ending dollar value of the three programs, which you mentioned at 14:32 in the video. Both the chance of success and the portfolio ending value are based on the capital market assumptions (investment returns) used by the software. Most likely, Projection lab is using investment return assumptions for the assets that you modeled that are lower than the other two programs. This should explain the lower terminal value of the portfolio in ProjectionLab. ProjectionLab is likely using a smaller standard deviation for the investment returns than Boldin and Pralana. This results in smaller return variations year-over-year in the annual returns randomly selected in he Monte Carlo calculations. Over a 30-year time horizon, this dampens the volatility of the portfolio which in the case of your particular plan results in the chance of success score being higher than the other two programs.
Makes sense. One other comment mentioned that Boldin (I think) increases the "volatility" as the average rate of return increases. Good observation.
This is great Rob, I have learned so much from your channel and appreciate you taking the time to ask these important questions! I had an advisor tell me I had a 58% chance and wanted to learn why that was so started figuring out how to project rates myself. I still have too many unknowns in my future to know my retirement spending but want my projection tool to be helpful when I posit the possibilities! 😅 It certainly would help to understand the basis for the different results.
As a Boldin user, I at least know it is the most conservative. The large differences in Chance of Success across the models is interesting and I look forward to learning more of the why.
I am familiar with PL and it shows the number of simulations that succeed or fail, if the others do as well, how do those compare? Trying to see how if the ending balance is similar, is the number of failures driving the difference in success rates?
The success score is basically just that, the number of simulations that succeeded versus the number that failed.
@ True, however PL counts Large surplus and Just made it as success and Almost made it, and Failed in the middle as failures. SoI guess that rules that out as the source for the difference.
Wouldn't surprise me if the more pessimistic results are incorporating some version of the current CAPE ratio in their calculation. Or even if they all are, but they're using different formulas or different periods to look up that ratio.
A couple of things to look into. Are the distributions at the beginning of the period or the end? Are returns applied at the beginning or the end of the prriod? Is inflation applied every month or at the end or beginning of the period? When are taxes paid? What data is used for the Monti Carlo (btw I was there last week)? Also, you need to start at the same point, not some late 2024 the other 2025, it makes a difference!
Rob, Thanks for all your efforts in educating folks like me. I question the use case for this video. As a user of one of these retirement planning tools, why/when am I going to compare 3 different tools for the feel good (or bad) number? It may be helpful for these 3 vendors, if they've not already done this kind of comparison. What will be helpful is to create a list of specific real life use cases - Roth conversion, Withdrawal optimization to name a couple & try them out in these tools and rank them. Appreciate your effort!
By the way, great job. It's no small task to try and learn 1 tool let alone 3 or 4!
Great video, really looking forward to hearing about how each provider responds, and learn more about why the differences show up.
Thank you! Great topic and I'll be interested in why the differences.
I use both Boldin and Pralana. I like the look and "feel" of Boldin but I can see better what is going on in the background with Pralana. I'll eventually settle on one calculator in the future though, so I am looking forward to your followup on this video. Thanks.
This was great
Thanks Rob. I was surprised they were so different
Good evening Mr Rob, sadly enough there are so many folks who simply are clueless when it comes to finance. Your videos really help those folks. For my personal investment style I have always focused on quality and growth stocks. Today at the age of 75 I still have the same investment philosophy, keep it simple, basic and make sure you know where your money is going. Be careful with those advisors who rely on animal spirits. Keep up the good work!
Very interesting comparison Rob. For someone like myself who is considering using risk-based guardrails (which increase/decrease withdrawals based on % success), that's some serious food for thought.
I am looking forward to follow-up video(s) on this important topic! I am also curious which (if any) of these tools support different asset allocation methods in post-retirement (fixed AA, rising equity, bond tent, declining equity, etc). This is something I am currently wrangling with as I'm at the start of my retirement and SORR is top of mind, especially given where market valuations are today. Portfolio Visualizer makes an attempt at via their Financial Goals feature (can accomodate rising/declining asset allocation glide paths) but its targeted to pre-retirement and lacks the specific capabilities needed for post-retirement.
This is excellent Rob and I appreciate you reaching out to them as you will get a better response than a normal user I suspect. I use Bolden and as you say, these differences are concerning. It will be helpful to gain more insight how they work and differ to help explain the dirrerent results. Larry, Central Valley, Ca.
Rob should have looked at and showed what the modeling settings are in ProjectionLab - in ProjectionLab you can use historical returns for example. Not clear to me which settings he has.
I thought I did. Anyway, I used normal distribution, not historical returns.
@@rob_bergergiven there was higher chance of success, but lower ending balances, I wonder if the differences will come down to volatility metrics and assumed std dev.
The assumptions that each model used at each point in history were produced systematically in ways that mirror actual approaches to modeling and CMAs available to and used by advisors today, and are as follows:
Traditional Monte Carlo: One set of CMAs applies to all years in the plan: In our analysis of this approach, CMAs that are the average portfolio returns and standard deviation of returns from the preceding 30 years were applied to each point of time examined;
Reduced-CMA Monte Carlo: This is similar to Traditional Monte Carlo, except that CMAs are reduced by 2% and standard deviation is reduced proportionally;
Historical Analysis: Actual historical sequences of returns and inflation models the range of possible return and inflation sequences. The assumptions for this analysis used all history available up to each point in time; and
Regime-Based Monte Carlo: 2 sets of capital market assumptions - one that applies to the near-term and one that applies to the long-term - are used to produce simulated return and inflation sequences. The assumptions exclude half of all prior points from each point in history based on their economic dissimilarity to the point in time being examined. From this filtered set of historical prior points, averages from the first 10 years are used to produce near-term CMAs, and averages from the next 20 years are used to produce long-term CMAs.
Having a 60% only chance of success when the withdrawal rate is 4.8% doesn't make much sense and seems to be missing the forest for the trees. Sure, the assumptions are the most likely reason but right there one would expect a higher chance of success at that withdrawal rate
I agree with the comments that methods used for MC simulation are likely very different. This video inspired me to do a test in Boldin. I set up a scenario that had a single roth 401k will 1 million and there were no withdrawals or deposits, just like in the video. At the end of the timeline (37 years) the 80% confidence interval from the MC simulation was (820k, 2.1 mil). That means there is a 10% chance of getting and ending balance below 820k and a 10% chance of being above 2.1 mil.. The I created a scenario with 4 roth 401k accounts each with 250k in them. So both scenarios had the same starting balance and return assumptions. However the second scenario 80% confidence interval was (1.1 mil, 1.8 mil). This is evidence that Boldin assumes that the investment accounts returns are not 100% correlated with each other. I would like to know from Boldin what correlation assumption they make. That has a big impact on planning. Going forward I will enter balance information into consolidated accounts in Bolding. For example, I will combine both my wife's 401k and mine into a single 401k account since they have the same distribution of assets.
Excellent idea & test. One would think that the numbers should have come out exactly the same. So there seems to be some kind of random number generator in there. Did you try running each test multiple times to see if there were any variations?
@@everlastingarms3065 I am not surprised at all.Separate accounts would provide diversification if they held different assets. The problem is that I have multiple accounts that I cannot consolidate that have about the same asset distribution. Since Boldin does hot have an concept of assets, it only has "accounts", the best we can do is to work around it. Re-running does not change the confidence intervals much at all
@@everlastingarms3065 By definition Monte Carlo is random. Unless two packages use the exact same settings and algorithm they will never produce the exact same results.
Thank you for PRACTICALITY. More More More
Great video, can't wait for the follow-up.
@Rob, have asked a few times but maybe it is getting lost. Could you please do an OnTrajectory vs Boldin please. I would appreciate it.
I'll add it to my list.
Data/parameters used (eg., random, historical returns ),number simulation runs used, etc in each package?
In Boldin, how do I create a test plan when I already have a baseline plan with account connections, etc? I tried creating a scenario off the Baseline plan, but it replicates all the connected accounts and won’t let me delete them from the scenario (along with other restrictions). PL is much more obvious, you just create a new plan. Is there a way to save a Boldin Baseline plan so I can paste it back in later?
Rob, based on on various videos, I have uploaded my financial soul into Boldin, trying to model a Roth Conversion strategy. Which tool do you recommend for a 2nd opinion ? I am concerned my inputs/assumption in Boldin are inaccurate - I always like to trust but verify.
Thanks. Interesting. I tried last week to vompare the Vanguard projections with the free Empower program and the paid MaxiFi. Not rasy to do and rather different results, with the empower being the most pessimistic.
I use Boldin and know that there is a default setting for long term care expense (I saw someone mentioned this already). Wonder if other software includes default expenses.
I’ve heard Moshe Milevsky say most of the companies don’t understand at all deep level how their own Monte Carlo calculators work- in terms of the derivatives and other functions they have baked in. They use “off the shelf” models that may not be applicable
Monte Carlo simulation is very straightforward math. It mostly just comes down to how the statistical parameters are set.
@ that may have been what he was referring to. The essence being they couldn’t explain to him the parameters they chose and why.
@ Dr Milevsky mentioned it in “The Rational Reminder” podcast. Episode 122. 44;40 mark if interested to hear his take.
i love watching your video it gives me motivation
Boldin also predicts a much lower probability that Ohio State will win the National Championship.
cruel...
That’s just mean…I love it!!!❤😂😊
And that's using the optimistic scenario. 🤣
Great videos!! I didn't hear you talk about pricing. I have Boldin that for the $130/yr, they are GREAT!!! How about the others? Overall, how are the others with Roth Conversion analysis, scenarios, etc?
I will do a roth conversion comparison soon. Pricing is roughly the same for all three tools.
How does your PlanVision eMoney compare to these three diy apps?
I have PV eMoney and also use Empower. I just cancelled Boldin after a couple years and am considering Pralana (reading key parts of the manual). To me eMoney lacks flexibility due to needing an advisor to make updates. It's a little old school. The value of your PV subscription is having access to great folks and also a tracking tool for your plan. Boldin had great UI but I couldn't trust the results and QA can be lacking. Pralana may be old school as well, but it's much finer grain than eMoney. I will keep watching Rob's videos to inform my purchase.
Rob, is Boldin (aka New Retirement) still a sponsor of the channel???
I don't like how some show the year on charts instead of your age. The year 2060 means nothing to me until I convert it to my age.
Great point. All my spreadsheets show both & I only look at the age, never the year lol.
I wonder if there is some uncertainty in how "running out of money" is defined? Maybe they are using the first or last days of the target year to determine that? If each simulation carries a balance and an uncertainty, do they define zero as the balance, or the (balance - uncertainty) ? If they use the uncertainty, is it one sigma like the previous sentence, or three sigma?
The good news is that the projected balances are nearly the same. Thank-you for doing these tests... definitely interested in your follow-up results!!!
PS - As a follow-up, the Boldin success rates near 65% are very close to the 1-sigma values of a Gaussian, and the Projection Lab results at 93% are close to the 2-sigma value of a Gaussian distribution. Since they give answers so quickly, my gut feeling is that they are not running 1000 Monte Carlo simulations with each update, but rather making an estimate, and using different criteria for their uncertainties... 1 and 2 sigma respectively.
Some software calculates things monthly and others calculate yearly and that can account for much of the different results. And I think they may handle tax rates quite differently so the differences can become even larger when taxes are starting being included. As far as Monte Carlo simulations go, the success scores might be based on different percentile ranges. They SHOULD be based on the 0-100 range but if they are not (might be 20-80 or something) then the score will be very different. Also, Boldin only uses 1000 iterations in their Montte Carlo so their result may vary a bit more from run to run. But that shouldn't necessarily make it wildly inaccurate.
and even the monthly expense distributions could differ with end of the month vs beginning of the month.
@@GregoryLooney Right. There are so many variables. Not to mention that we are projecting 30 years into the future when we don't even know what will happen next year!
Rob, I look forward to your videos. Thank's for all you do. Can I ask one question? My wife is 75 and still works. She has not taken RMD's yet. When must she start. Should she wait until after the new year or does it matter?
I like how PL shows success score to .00 digits of precision, lol.
Thanks Rob. I have been looking forward to these comparison videos. Also look forward to the responses from each of the developers. The specific standard deviation for each asset class can be set in Pralana. I don't believe this is the case for Boldin and I don't think they disclose what is used. Also, in Pralana you set the real rate of return and inflation. I believe Boldin sets the nominal rate of return and inflation. You didn't show these, but I assume these were set to be apples to apples.
Boldin does specify, "The standard deviation, while not explicitly disclosed, closely mirrors historical returns." But, couldn't your optimistic and pessimistic inputs in Boldin serve as your variable deviation?
@@hanwagu9967 As I understand it, in Boldin the optimistic and pessimistic inputs serve as the assumed return for a particular investment account. The Monte Carlos simulation would then overlay a certain standard deviation on that assumed rate of return. One would need to know what that standard deviation Boldin applies, to try to match it in Pralana. Also, as I said above, Pralana assigns the standard deviation by asset class and Boldin seems to do it by account.
Hi Rob, my name is Alicia, & I had an older sister pass away in 2022, who left me some money. I contacted the financial institution, Vanguard; I informed them at the time of her death, that I would not be doing any investing. I was told that the funds were not available to me for two years, & that I needed to be put all the money into an IRA. After listening to your program a few times am learning more about other investments options, can I use the funds that from an IRA for maximum cash flow?
I see some weird numbers coming out of Boldin and there's little to no transparency about how they are coming up with it. I'm gonna stick with the spreadsheets for now! 😆
Great Video Rob. Just another reason to not rely on a single resource for planning your financial future. Looks like the Cryto Bots have found your comment page. Sad to say, but you might have to start managing your commenters/subscribers as its become pointless to sift through them for additional knowledge/insight.
That’s an excellent video, my friend. As I ponder retirement, I have to consider how much risk there is with the crazy Trump future and what might happen to the stock market. If the Trump factor wasn’t in there with nervous investors than I would likely die with $1 million. I currently use projection lab and I think Kyle has done a wonderful job with this program and is adapting it to many very smart users input.
Thank You.
Could you do a vid comparing Right Capital?
as Rob mentioned, he was using a "client" version of RC (from Kevin Lum at Foundry - I have the same one) and it doesn't allow rate of return to be set by the user (it just has various pre-configured asset mixes). The full version of RC is only available as a professional advisor tool with limited functionality available to the client so it really isn't possible for Rob (or anyone that isn't a professional advisor) to do an apples to apples comparison.
@@lindsaynewell6319 .... Thank you. I hadn't watched it yet, but I use RC and saw it not included so I figured I'd request.
Very interesting!
Yeah I’ve always thought Monte Carlo is useless for this application.
If Rob can't figure it out, I'm heading to my bunker.
We realize once again that "Nobody knows nothing". It's simpler to use a reasonable SWR target for planning and adjust spending as needed when retired.
From a design of experiments basis I can guess that maybe the number of runs in the simulations could have varied between them as well as the parameters they use and the data they draw on to feed the simulations , It withstanding the equations that nodes the market.
The conclusion one can draw from the set of 3 is none can be trusted , because two agreeing doesn’t mean they are more accurate. They can be equally wrong.
For those projections to have any merit in interpreting they need to be open source and standardized. That way the community of investment analysts can comb over the simulation equations and inputs and assumptions and make sure all the needed stuff is there, and maybe standardize weighting system to get to a common metric - like the MPG rating on cars. Reality isn’t matching it but everyone is running the same test conditions.
As it stands I don’t believe any of them.
Monte carlo simulation seems way too black box to me. Who knows what assumptions are made to generate the simulated data and how reliable those assumptions are? It's revealing to see how inconsistent the results are in your video!
I know backtesting has major limitations, but at least it's easy to understand and reason about.
I think that maybe we are asking too much of these tools. They are not crystal balls.
Having said that, they should be transparent about how the chance of success number is generated and what expected returns and volatility they are assuming.
These tools are most useful to give a low-cost, broad -strokes look at your retirement spending plans to make sure you are not delusional about being able to retire without going broke before you die.
Optimization just asking too much for something as uncertain as the future.