I know everyone likes to reference the Trinity study but I think blindly using the success rate percentages from that study without considering asset allocation is a bit foolish. The AA used in that study under performs a more aggressive AA significantly. The 40-year success rate of an all-stocks portfolio is 95% when withdrawing 4%, and I think a 90/10 outperforms that by a hair.
Actually according to cfiresim it's 91.67% both for the 100% equities and 90/10 portfolio (using the default parameters with a 4%WR). It's for the 30-year cycle that a 90/10 allocation provides a higher success rate (96.61%) than the 100% equities scenario (94.42%). Still around 92% is not bad and I also believe that a high equity calculations is key for a successful early retirement.
If you want to sit there and stare at Sol's chart and nothing else, and claim that the 40-year success rate is 10% lower than the 30-year success rate, that's absolutely correct. I feel bad for the person that makes his decision to continue working or not by nothing but Sol's chart, because that's only one tiny piece of data, and the success rate doesn't have to be that low. That doesn't even count all the human intervention factors that would alter the failure course because you know withing the first decade of retirement if you have a problem.
Let's be clear here: That chart was thrown at me as a counterargument to my position. I used it to point out that it in fact backed it up.
I would totally agree that anyone who is looking at retiring really understand the ramifications of the scenario and run their own simulations. They should also run some simulations for an adverse scenario (working extra years, cutting consumption) to see how much that really would help them. And after that they come to a plan of action based on their own perception of and capacity for risk. And perhaps they should also have a predefined "statement of action" at what point they do what. E.g. a series of statements "If my portfolio goes below x% in real teams, I am going to cut my consumption by y%".
And finally they should realize that that a lot of successful cycles (and some failure cycles) do not apply, because we actually have a bit more knowledge: We know current equity and debt valuations and it is reasonable to conclude that a 2017 retiree would have an outcome below the median. Thus the conditional probability of success likely is also lower.
And once the prospective retiree realizes all this, then they can come to an intelligent decision for their course of action.
If this thread has shown me anything, it is that a certain percentage of good-intentioned folks on this forum have a rather loose command of the facts. (And I'm not talking about 91.67% vs 95%, that's just minute stuff.) So when someone claims something, you better fact-check for yourself.
I've said it before and I'll say it again. If you're a 30 year old male in the US, you have a better chance of dying before the age of 65 than you do of running out of money (using Sol's chart). Only 82.5% of 30 year-old males make it to 65. If a 15% failure rate is enough to keep you working, what does a 17.5% chance of dying do?
You do realize that there is a logical problem with this argument, don't you? Your argument - in general terms - is that behavior x has generally a good outcome and its probability of a bad outcome is lower than a bad outcome for y, so it's reasonable to do x, because of that.
So let's go back to the Russian Roulette example with the 20 bullet magazine. If someone offered you $10M for for a "shot", you could argue that as a 30-year old male in the US you have a better chance of dying before the age of 65 than you do of dying from this action. If a 5% failure rate is enough to keep you from taking the "shot", what does a 17.5% of dying do?
Now everyone would see the problem with that line of reasoning. And since the argument is constructed the same way as your argument, it would show to me that your argument also has some logical problem.
I haven't examined this thoroughly, but at first glance it seems that the problem is that you're adding risk. Now that risk is lower than your base risk (dying before 65), but it is (in first approximation) independent and and an additional risk. Therefore the probability for a bad outcome goes up, meaning you either die before 65 or run out of money (or both).
So you are now going (using Sol's chart and again assuming that those two events are independent) from a 17.5% chance of a bad outcome (you'll die before 65) to a 29.875% [1- (1-0.15)*(1-0.175)] chance of a bad outcome (you either die before 65 or run out of money or both).
So that's not a convincing argument to me.