In 2009 or 2010, a 79 or 80 year old man named Forrest Fenn hid some 20+ pounds of gold and gems worth between $500K and $3M+ (depending on which estimate you believe) somewhere in the Rocky Mountains. It’s yours if you solve a 24-line poem containing nine clues and go retrieve it. There are remarkable similarities and contrasts between the Fenn Treasure Hunt (or “Chase” as it is sometimes called) and the hunt for MH370. I believe the Fenn Treasure Hunt has important lessons to teach the currently baffled MH370 investigators who have spent well over $185M ($160M under ATSB-directed search and an estimated $25M for Ocean Infinity) trying to locate MH370 with no success despite a potential finder’s fee prize of $70M instituted in 2018 (but not currently active).
The Fenn Treasure was intentionally hidden and most experts believe MH370 was intentionally hidden, perhaps even by one man who went down with MH370. Searchers for the Fenn treasure formulate hypotheses at home, then go “boots on the ground (BOTG)” with hopes to find the treasure. MH370 searches have generally involved mathematically driven hypotheses based on satellite and debris drift data which drives subsequent underwater search. My estimate of search area for the Fenn Treasure is about 180,000 square kilometers. From 2014 to 2017, the ATSB directed some 120,000 square kilometers of subsurface scanning, and Ocean Infinity doubled that in 2018. The Fenn Treasure is located in a 10-inch square by 5-inch high bronze chest that may or may not be buried, and MH370 is likely located in a debris field on the floor of the South Indian Ocean near the seventh arc.
Fenn Treasure Hunters arguably apply a forensic based approach to the hunt, decoding the clues left by Forest Fenn in the poem and hints he left in his memoir called The Thrill of the Chase or TTOTC from 2010 where he first disclosed he had hidden the treasure chest. The MH370 hunt has been mathematically-based with almost no regard to forensics. I have argued since two weeks after disappearance that forensics should be applied to MH370 to no avail. In 2018, I discovered a very intimate relationship between a new branch of mathematics called Bayesian Occam’s Razor and my forensic argument and have continued to develop it.
As far as I know, mathematics has not been applied to the Fenn Treasure Hunt, but it’s instructive to apply some mathematical analysis to see how it works, if there is a chance it will be found anytime soon, and glean lessons that can be applied to the hunt for MH370. First, we collect all the publicly available metrics we can:
Correlating the various metrics to Google search interest (I have converted search interest to cumulative interest to match the other metrics), we get good to excellent linear correlations:
Google data is continuous, so we can estimate the other sparse metrics using these correlations.
The Fenn estimated number of people who have searched is the noisiest correlation, but is based on estimates by Forrest Fenn and cannot be known exactly, since nobody is required to disclose whether they have searched for the treasure or not, and the poem is published all over the internet. There may well be people searching who are not revealed by any of the metrics above. It’s not clear to me if all of these data points are his estimates of BOTG searches or armchair searches. He said he put particular care into the 65,000 data point, so it may well be that the lower bound envelope is the best estimate of the number of people who have searched. If so, the regression would be y=2.09x. That works out to 3.14 searchers looking for the pie for every book sold. Considering searchers often go out in pairs or larger parties and some people go out without the book, it sounds about right.
The number of TTOTC books sold is somewhat noisy, but that may explained in that the book is beyond it’s tenth printing and is sometimes unavailable while new copies are being produced. Also, the only official source of books is Collected Works Bookstore, but many resellers on eBay and Amazon purchase batches of books and mark them up substantially. Nevertheless, we use number of TTOTC sold as the basis for a model because Forrest Fenn has said that TTOTC contains hints and thus we can assume a book buyer is most likely to be be the poem solver. (Fenn does not make money off of book sales, and 10% goes to cancer patients).
Given the good correlations to Google search interest, we can create a synthetic TTOTC books sold proxy and then do some calculations with it:
The Forrest Fenn Treasure Hunt appears to have reached a saturation point whereby cumulative growth is linear. Given the other correlations such as website hits above, one can deduce that the Chase has reached an equilibrium where the rate of new entrants coming in roughly equals the rate of searchers leaving. With more data, we could squeeze out the the average searcher residence time in the search using Queue Theory with the growth phase prior to about 2016 possibly defined by a logistic function. Book sales should continue for the foreseeable future presumably so long as the treasure remains unfound.
Interestingly, Forrest Fenn has disclosed search progress in the past, albeit rarely. Whether by reading the thousands of emails he has received or by monitoring various Fenn Treasure forums, Fenn is aware of the progress of the search and how many clues have been solved. To my knowledge, he has only disclosed
two three (Updated: someone informed me of the obscure 26 Sep 2012 Forrest Fenn comment) data points regarding what the progress has been:
I use a small f because I’m too lazy to hit the shift button. All of this cyberspace verbiage is conspicuous by the absence of talk about where warm waters halt. Several months ago some folks correctly mentioned the first two clues to me in an email and then they went right past the other seven, not knowing that they had been so close. Alas, and dame fortune, so often a fickle and seductive wench, never spun her wheel to lure them back.
26 Sep 2012 Forrest Fenn comment to DalNietzel.com Blog (emphasis mine)
HOST: You’re a great man. Well what, look at what you’ve started. It’s a fantastic quest. In your bones do you think that somebody’s going to find it soon or do you think it’s going to elude discovery?
FENN: Well you know it’s impossible to predict it. There have been two different parties that have figured out the first two clues but the, they went right past the treasure chest and didn’t find it.
Are there signs that people are getting closer to solving your puzzle? How many clues have people solved now?
Searchers have come within about 200 feet. Some may have solved the first four clues, but I am not certain.
Plotting these points using our “TTOTC Books Sold” proxy above, we obtain a pretty remarkable clues solved power law fit (very close to square root) relationship:
This relationship is certainly subject to potential error – Fenn said two different parties had solved two clues – this may indicate one party solved two clues much earlier than indicated (Update: indeed they did per the added quote above. When I have time, I will update this analysis). The later data point may indicate something less than four clues had been solved. Differing interpretations of these data points could potentially flatten the above relationship, which would imply a much more difficult puzzle than indicated. Another recent data point would be nice, to corroborate, but I simply take the data as-is and interpret it as shown above. Given the abundance of power law relationships in nature, I wouldn’t be surprised if one applies here – more about why this may be the case below.
At first glance, the above relationship would imply the clues get harder as you progress, but we discuss why that is not the case below. The above relationship probably arises from the fact that each clue is contiguous and they must be solved in order as described by Mr. Fenn:
You cannot solve the problem by starting in the middle of the poem. You should start with the first clue and then solve the other eight in order.
Each clue could represent a screening barrier a given searcher cannot get past, and it suggests each clue may require it’s own individual way of thinking or may have a unique trick associated with it, perhaps explaining the power law relationship. We discuss this in more detail later.
Finally, we can combine our clues solved model with the books sold model to obtain a model prediction of when the treasure will be found:
We can estimate that as of today, approximately seven clues are solved and all nine should be solved sometime in the year 2023. If there are a large proportion of “silent searchers” out there that Forrest Fenn is never aware of, then the treasure could be found even earlier than our model indicates. Though with Fenn receiving some 8 or more emails for every book sold, it’s possible there are not too many. Can a treasure hunt be predicted by such a simple model? What is going on?
First, we can envision two very different extremes of how the treasure hunt works:
- Solver Model A – The treasure hunt is very hard work and requires a large quantity of time to solve. The person(s) most likely to solve are people who began in 2010 and have worked diligently solving or testing hypotheses with BOTG searches ever since. Some Fenn Treasure Hunters appear to work very diligently for years, and essentially brute-force test solutions all over the Rocky Mountains, sometimes frequently changing their solve hypotheses.
- Solver Model B – The successful treasure hunter will be one who is most aligned with Forrest Fenn’s way of thinking and while significant effort on their part may be required, relatively short solving times are required.
Forest Fenn has numerous quotes that reference the importance of “deep thinking” and “imagination” that tend to suggest Solver Model B is closer to the truth. To my knowledge he has not said that tedious brute force solving is required, but has said the successful finder will have earned it. We can also note that the first reports of two and four clue solves have come after bursts of interest in Forrest Fenn Treasure. Did the bursts of interest in The Chase bring in new searchers in tune with Forrest Fenn who solved several clues quickly? I sense from the graph that may be the case. Mr. Fenn also has various quotes that can be interpreted to support this such as these (via tarryscant.com):
Some folks correctly mentioned the first two clues to me in an email and then they went right past the other seven, not knowing they had been so close… Some of the searchers have been within 500 feet I know… The person who finds the treasure will have studied the poem over and over, and thought, and analyzed and moved with confidence.
In this and other quotes, he seems to imply searchers solve the initial clues, then progress no further. We formulate yet another solver model:
- Solver Model C – Very much like Model B, except searchers are naturally predisposed to “get” N clues quickly in a spectrum of Fenn-similarity. Searchers more aligned with Forrest Fenn’s way of thinking will get more clues solved initially. If those solvers have solved less than nine clues, they must have the perseverance to analyze and solve the rest.
Model C seems to better fit the data and the description of who will solve the poem by the man who wrote it. As books continue to be sold, new entrants with new hypotheses enter and by random chance new “Fenn-similar” thinkers may solve many clues fairly quickly. It will simply be a matter of time until a person with the right combination of “Fenn similarity” and perseverance comes along and solves the poem. While older entrants may be at a disadvantage relative to a Fenn-similar newcomer, they may be able to make up for it by experience and perseverance. Assuming Model C, just what are the odds of a searcher solving all nine clues given a relatively quick initial N-clue solve? If we make the simplifying assumption that only TTOTC book buyers are the ones who will solve clues, then given our power model, we obtain this estimate:
|Clues||Estimated TTOTC Buyers who have solved N clues or more as of today||Estimated TTOTC Buyers who will have solved N clues or more in 2023 if trends continue|
|8||0.7 (this can be interpreted as there is a 70% chance someone has solved 8 clues)||1.3|
The above are expected minimum numbers for Clues 1 through 9, because they do not necessarily fully take into account the time-perseverance component due to the age of the data. Also, Forrest Fenn has continued to release new books, scrapbooks, personal blog articles, interviews and other potential hints in interviews and various Q&A’s since the data points our model is based on were released. Whether this information makes the clue solving easier is subject to debate within the search community.
Nevertheless, by 2023, when we expect all nine clues will have been solved, we estimate 79,954 copies of TTOTC will have been sold. Of those book buyers, at least 93 will have solved Clue 1, at least 22 will have solved Clues 1 and 2, etc. Our model predicts the threshold for solving the first clue is very steep, but Forrest Fenn also seems to allude to this:
The first clue in the poem is begin it where warm waters halt. That’s the first clue. If you don’t, if you can’t figure that clue out you don’t have anything.
Our model says only 1 in about 860 TTOTC buyers will have solved Clue 1 by 2023, but their prospects become brighter after that: about a quarter of the Clue 1 solvers will go on to solve Clue 2, about a half of the Clue 2 solvers will advance to solve Clue 3, etc. Again Forrest seems to corroborate this:
“I have some advice. Read the book, then read the poem, over and over, maybe even memorize it. And then go back and read the book again looking for hints that are in the book that are going to help you with the clues that are in the poem. That’s the best advice that I can give. You have to find out, to learn where the first clue is. They get progressively easier after you discover where the first clue is.”
Indeed, our model indicates they do get easier for the collective group of clue-solving searchers, but any given searcher may get completely stumped at any point along the way. By the time someone solves Clue 6, they have a really good (almost 50%) chance of solving all the way to Clue 9. Given someone may have already solved to Clue 7 by now per our model, the eventual finder may simply be working on the last two clues. Alternatively, a new Fenn-similar solver could swoop in, quickly solving up to Clue 8 and take the lead. Or similarly, perhaps a veteran searcher could suddenly rethink everything and have an 8 clue eureka moment. Our model has no opinion on these characteristics of the successful solver, but it does suggest it’s likely very lonely at the top now and until 2023.
If the Fenn estimate of searchers regression is relatively accurate, then we can estimate some 670,000 searchers will have searched for the treasure by the time it is found (if found in 2023). If each of these searchers spends conservatively $500 per trip (some spend much more), and two trips on average are taken, then searchers will have expended some $670M on the hunt, far exceeding the roughly $200M spent to search for MH370. Even if the Fenn-estimate lower bound described in the beginning of this article applies, we would reduce the numbers to 250,000 searchers and $250M spent.
However, the “progressively easier” nature of the clues, if true, presents a dilemma: is it reasonable to assume a searcher could have seven clues solved as of 2019, but end up spending the next four years solving the last two clues if they are that “Fenn-similar”? Given our model predicts that a Clue 7 solver has a better than 50% chance of continuing on to solving Clue 9, it does seem unreasonable. Our power law clue solving model may well accelerate towards the upper clue numbers. It could be that a solution to all nine clues is relatively imminent. With snow covering much of the northern and high elevation Rocky Mountains as of this writing, it could simply be a matter of the snow clearing to open the 2019 search season. It is unknown whether a searcher would know how many clues they have solved, but if they know they have seven clues solved, they have incentive to work very hard and accelerate the find, which is likely not captured by the model. Then again, whoever may have solved up to Clue 7 may be permanently stumped and never advance to Clue 9 without outside help.
Then again, the hunt could take a long while. A wildcard is that a searcher must be good both mentally and at BOTG search to be successful. Presumably living near the Rocky Mountains is an advantage depending on how much BOTG effort is required. Four people have lost their lives looking for the treasure, but at least some of them may not have not heeded Forrest Fenn’s warnings that he was 79 or 80 years old when he hid the treasure and likely was not attempting to go places or do things similar to what they were. Certainly any type of recreation in the Rocky Mountains can involve loss of life.
Douglas Preston, NYT bestselling author and friend of Forrest Fenn wrote this in a foreward to Fenn’s most recent TTOTC follow-on books:
He had worked out all the logistics but one: how he could pull this off by himself, without help. He did not feel he could entrust anyone else to assist him. “Two people can keep a secret,” he said, “only if one of them is dead.” He had already written the poem, and he now brought it out and read it to me. It was similar to the poem he later published in his book, The Thrill of the Chase, but not, if I recollect, exactly the same. He tweaked it many times over the years, making it harder.
I said that there were a lot of smart people out there and I feared the poem would be deciphered quickly and the treasure found in a week. But he assured me that the poem, while absolutely reliable if the nine clues were followed in order, was extremely difficult to interpret—so tricky that he wouldn’t be surprised if it took nine hundred years before someone cracked it.
The limited data seem to suggest otherwise. With the MH370 search in disarray, and 2019 going down as possibly the first year without subsurface search or apparent official investigation, I predict the Fenn Treasure will be found before MH370. This is despite my analysis that searching some key hotspots by Ocean Infinity that would take less than a season to complete would likely result in an MH370 find. Adding to my pessimism is that nobody besides me seems to apply analysis tools that find missing planes to MH370. To complete the trifecta of my MH370 pessimism, official, semi-official and even influential amateur MH370 enthusiasts barely or simply do not believe MH370 could be in the spot I have believed (and I believe I have shown mathematically) to be the most likely MH370 resting place since two weeks after disappearance.
The Fenn Treasure Hunt is relentless, and the above analysis indicates to me the end appears to be “drawing nigh.” The same cannot be said for MH370. In Part 2, I’ll list the lessons MH370 investigators can learn from the Fenn Treasure Hunt and suggest a way to get the MH370 search back on track.
Sources for correlation data: