Skip to content

Exploration vs. exploitation tradeoff

Alon Levy (link from Palko) looks into “Hyperloop, a loopy intercity rail transit idea proposed by Tesla Motors’ Elon Musk, an entrepreneur who hopes to make a living some day building cars,” and writes:

There is a belief within American media that a successful person can succeed at anything. He (and it’s invariably he) is omnicompetent, and people who question him and laugh at his outlandish ideas will invariably fail and end up working for him. If he cares about something, it’s important; if he says something can be done, it can. The people who are already doing the same thing are peons and their opinions are to be discounted, since they are biased and he never is. . . .

And this:

In the US, people will treat any crank seriously if he has enough money or enough prowess in another field. A sufficiently rich person is surrounded by sycophants and stenographers who won’t check his numbers against anything.

No kidding.

The basic idea is that financial resources, political resources, and respect in the news media are somewhat fungible. $ can buy you political influence; $ and political influence can get you credibility with journalists, and if you are famous and respected you can use this to raise money.

Every time a rich guy like Musk lends his reputation to something like Hyperloop that doesn’t work, he loses a bit of credibility. Fair enough. And, without the billions behind it, journalists wouldn’t treat the project so respectfully. Which is logical reasoning too, as we’d expect just about any idea to me more likely to succeed if it’s backed with big money. Plus the reasonable inference that if this guy has made all this money in business, he must have some sense of what works and what doesn’t.

Each step in the reasoning makes sense. The trouble is that each step is merely probabilistic. Rich people can have bad ideas, certain plans can’t be implemented even with big financial resources, and so on. Even if Steven Wolfram were to devote the entire resources of his company to the problem of trisecting the angle with compass and straightedge, he would not succeed (unless you’re allowed to slide the straightedge along the circle, but that’s cheating).

It’s a tough problem for journalists. Statistical reasoning would suggest giving the benefit of the doubt to business proposals coming from rich people, but this creates a perverse incentives where richies can float all sorts of ridiculous ideas, secure in the knowledge that most outlets will report their plans uncritically. Get enough of these, and on the margin these plans could even be worse than what you’d see from institutions with fewer financial resources.

Game theory eats decision theory once again.

P.S. Levy writes of his “specific problems are that Hyperloop a) made up the cost projections, b) has awful passenger comfort, c) has very little capacity, and d) lies about energy consumption of conventional HSR [high speed rail].” I won’t get into the details on that—you can follow the link and the comment thread there.


  1. Shravan says:

    What we need is a statistician who is also a journalist by profession.

  2. Anonymous says:

    …well, seems journalism is the basic issue raised here.

    A journalist is nominally a person who gathers and distributes factual information about current events.
    “Current Events” is very, very broad category; “significant” current-events is more accurate, but instantly reveals the overall effort as totally subjective (…which it is).

    Journalists fundamentally should communicate ‘truth’ to their audience, or at least honestly attempt to communicate the truth about whatever they report upon. That never happens in commercial journalism.
    Commercial journalists are loaded with personal, institutional, and cultural biases & constraints… like most people.

    Objectivity in professional journalism can only be achieved on narrow subjects. But general objectivity can never exist because journalists must always subjectively choose ‘what’ to report… from an almost infinite array of events & topics.

    Bottom Line: don’t put faith in journalism generally

    “No event is ever correctly reported in a newspaper” (– George Orwell)

    • Andrew says:


      Sure, journalism has problems (but in some segments is getting better), but my interest here was more in the perverse incentives or moral hazards involved in Bayesian or reputational reasoning. There is a logic to giving more credence to the business plans of successful businesspeople—but that creates an incentive for said businesspeople to push the envelope and promote ideas that can’t work.

      • John Hall says:

        Sounds like something Behavioral Economists could study: how does the expected profit of ventures started by successful business people compare to normal people (controlling for industry, size of project, etc.)?

        Ultimately, this is why we have a venture capital industry. There are other rich people who are willing to finance these ideas, a whole portfolio worth. Sometimes they fail, but rarely they hit a unicorn that makes it worth it.

      • To what extent is this the main thing going on in US “growth” businesses these days? I mean, for example:

        Facebook buys WhatsApp for $28 Billion, a company that provides a free service based on the Jabber protocol.

        Google buys Nest for $3.2 Billion, a company that makes thermostats for home HVAC systems, and has never once made a profit.

        Theranos gets widespread contracts to provide blood testing using technology that every bio-lab tech with an Associate Degree from a community college would tell you is seriously flawed, gets valuation of multi-billions of dollars and nearly IPOs before regulatory groups come down on them for the sham they really are.

        It feels like more and more the business of the US is investing large quantities of money in puffing up some “new tech” to large enough inflation levels that mutual funds are forced to purchase the stock do to their operating policy and then dumping that stock into those mutual funds thereby extracting cash from everyday suckers indirectly through their 401k

        If I set up a company and provide free services to all my friends and family while taking a paper loss every year and reducing my taxes, the IRS will come shut me down for tax fraud. If a venture capitalist does it for the whole world, they sell the company for tens of billions of dollars, making a huge profit off the poor suckers like you and me who were forced to buy the stock in their 401ks and each take a loss of $5000 when the whole thing crashes.

        It’s arbitrage vs an index fund fulcrum point plain and simple.

        • Rahul says:

          Why are mutual funds forced to buy the crappy hyped stock? I didn’t get that part of your comment.

          • The way these index funds work is that they must track some index, inclusion on that index is contingent primarily on market capitalization. If you can get your company to be big enough they must buy your companies’ stock in order to meet their stated investment strategy.

            • The importance of the “Unicorn” (or $1B+ market capitalization for a startup) is not just that $1B is a lot of money, but also that it’s the order of magnitude of what’s required to show up on these indexes. The Russel 1000 has a current median market cap of $4.6 B according to Wiki: and minimum cap of $1.8B, The Russell 3000 would be less (Russell 1000 is the top 1000 of the Russell 3000).

              So if you can make WhatsApp worth $28B it’s now firmly in the upper half of the Russell 1000 and probably going to be on the S&P 500.

              Now take for example HotDiggity iDogs (30 min Uber-based delivery of custom-ordered online Hotdogs, Bratwurst, and Polish Sausages via Android and iPhone App, delivered at half the cost of production, a company I just thought up that loses money easily as fast as Uber which lost $1.2 Billion in the first *half* of 2016) if you can make it worth around $1B you can get it firmly in the Russell 3000. As founder and early venture capitalist you could dump $2B or more of its stock on unsuspecting suckers who don’t even know they’re forced to buy it by their choice of 401k fund, pay yourself a handsome CEO salary of $10M/yr for a few years while the company runs into the ground and you continue to cash out your stock, and then move on to found UBurger an Uber based Burger delivery service…

              Once the world gets tired of UBurger you can move on to create iDefenders an online lawyer firm specializing in class action lawsuits that represents you against investor frauds using lawyers in foreign countries…. etc etc.

            • Rahul says:

              But you can always invest in funds that aren’t index funds, right? If you feel that the index routinely and majorly has crappy, hyped up stocks what forces us to blindly follow such indexes?

              • what one person does isn’t so relevant, it’s what is done in broad swaths. 401ks and everyday investors tend to invest in index funds because they have low costs associated and they are widely available, and there’s not good evidence that any alternative exists which consistently does better. Furthermore, most people are ignorant of how this all is playing out. Hedge funds and VC firms whose job it is to game the investment industry do better, but those aren’t available to the common public, only the very rich.

                Revelations like the Panama Papers show that very rich people have extremely different opportunities available to them than even the moderately wealthy (say people worth less than $5M)

                There’s also the question of the time frame. When does the puffery disappear? Certainly while we have very low interest rates and VCs can borrow money cheaper than water valuations can be kept high while resources are “invested” into “future earnings”. Will Uber eventually be making $1B/yr profit instead of $1B/yr losses? Maybe, or maybe interest rates will rise, and their recent raising of around $2 Billion in leveraged debt will become worth only $500M during their liquidation 5 years out. Yahoo is listed on Google Finance as $5 losses per share on almost 1B shares, that’s on revenue of 5 Billion, so they spend $2 for every dollar they receive. They’ve been losing $100-200M a quarter for years.

                Yahoo is pretty much the poster-child for the true long-term of all of this. A quick run-up post-IPO in which early investors get rich by dumping shares, a decline as everyone realizes the asymmetry of information, an attempt to buy up things that have value over 10 or 15 years, a series of changes of CEO, long slow decline, eventual liquidation, over a lifetime of maybe 20-25 years they’re responsible for destroying $100B in value, mostly in the form of transferring wealth from everyday people to an elite of high-end investors, and an army of silicon valley engineers who’ve converted it into little real-estate empires in inflated real-estate land (aka SF, Berkeley, Palo Alto, etc) in exchange for not much of anything.

              • Note, if you pay me $50M / quarter, I will happily “lose” all of it, thereby “manufacturing” $150M/quarter in value.

              • Rahul says:

                But if “there’s no alternative that does consistently better than index funds” (which I agree) you can’t really say that we are being “forced to buy” anything really can you? You can’t be “forced” to do something that’s really the best in league.

                i.e. If it is so obvious that these crappy stocks are hyped up and such a big problem index funds wouldn’t be doing better than all alternatives would they?

                Note, I’m not saying that those stocks aren’t crappy. Just that no one seems to be “forced” to do anything here. You can probably tell that Hyperloop is crappy from day 1 but for a majority of the credulous population it’s just an “opportunity” to make money grow fast in what seems like a “cool”, “high tech” idea.

                So, unless there’s actual fraud (which should always be vigorously punished) investors get what they deserve and a gamble they willingly got into.

              • Rahul: “forced” in the sense that if you can get the value of the stock high enough the index fund is obligated by its operating agreement to buy it. This isn’t “forced” like “we held a gun to your head” but more “forced” like “we can rely on this happening and therefore it presents an opportunity for us to create a strategy to extract money for free”

                The existence of methods to game systems doesn’t mean that the system you’re gaming isn’t the best system available to the general public, but that’s like saying “gee there’s nothing better than leather soled shoes so you’re just going to have to slide around on the floor forever”. We should seek to avoid systems that can be gamed by people who don’t produce value even if they’re currently the best thing going.

              • Carlos Ungil says:

                > [Yahoo has] been losing $100-200M a quarter for years.

                Recent results have been bad, but they had been making money for a while:


              • Rahul says:


                How difficult would it be for a Fund to change its operating agreement? How difficult would it be to start a new fund that has flexibility in its operating policy to not blindly follow an index stock that its fund managers think is hyped?

                If this were indeed a severe problem there’s contractual fixes to it. Sure I think there’s a problem in funds, but I don’t think its that funds are unwillingly being “forced” to buy stocks they see are obvious bad buys merely because their operating agreement forced them to.

              • Rahul says:


                >>>We should seek to avoid systems that can be gamed by people who don’t produce value even if they’re currently the best thing going.<<<

                What's the solution you propose?

                My ideas are: Don't buy crappy stock personally if you trust your own judgement. If your fund consistently gets hit because it always buys stock that's obviously crappy (to you) dump your fund. Go for an actively managed fund if you like so you are not slave to an index any more. Of course, pay for the higher loading. No free lunches.

                Personally, I think you are overestimating your abilities to tell apart a-priori Theranos-like stock from real money-makers. Or maybe its just that the ones you spot as crap the lay-population cannot. So they'd not like their fund not buying what looks like a great stock. In which case, they deserve the crash that follows but you shouldn't invest in the same fund the herd invests in.

                If enough people think indexes are "polluted" by such hyped up crappy stock I'm sure a market will emerge for something like a manually curated hype-stock-adjusted-index-basket that funds could base their holdings upon. Maybe there already is?

              • @Carlos: My understanding is that the main way Yahoo made money was to buy up companies that made money, not actually create things themselves. Their market capitalization is pretty much determined entirely by an ownership stake in Alibaba, and they’ve spun off their internet business and sold it to Verizon (subject to regulatory approval)

                Here’s a list of their mergers and acquisitions:


                They are, in essence, a hedge fund that’s heavily invested in publishing of online tabloid magazines. They don’t really produce much of value themselves at all they simply use their early IPO generated funds to buy up stuff other people created.

                This seems to be a consistent trend, and part of the way 401ks are being “forced” to buy crappy companies is that larger companies owned by your 401k are acting as hedge funds and buying these crappy companies. There’s a lot of inside baseball going on in silicon valley where VC firms invest in questionable startups, puff them up, then sell them to you and me by using their influence to find larger established companies (owned by your 401k) to buy them. (Nest, WhatsApp, Google Voice, etc).

                @Rahul: so, even if you don’t invest in index funds that are “forced” by their operating agreement to buy puffed up companies, if you own companies like Yahoo, Facebook, Google/Alphabet, etc they are complicit in transferring value from everyday investors to questionable small companies owned in large part by VC companies.

                The last 15 years of technology boom isn’t being driven by massive consumer demand for the products produced by these companies (Yahoo was making something like $0.20 per user per year) it’s being driven by back-room deals funneling capital out of your retirement account into VC pockets with a few “headline” companies being used as justification.


                Now HotDiggity iDogs doesn’t sound so funny.

              • Yahoo, Alphabet/Google, and FB are not the only hedge-fund-like companies, but they’re the major players in this business. Here’s what Yahoo did with its acquisitions:


                Which is to say, basically they bought up crappy companies owned by their friends and shut them down, then put their friends to work shoring up the tech behind their existing media to try to make them seem new and improved (image recognition companies headed over to Flicker, email management companies headed over to Yahoo Mail, various companies puffed up yahoo video and advertising tech, etc)

                In the end, Yahoo was spending money buying “talent” to puff up a business that rapidly declined to almost zero value (Yahoo = Alibaba stake + Everything Else, their market cap rapidly headed towards valuing “Everything Else = 0”. Finally Verizon bought “Everything Else” for $4.8 Billion but it’s a gamble on Verizon’s part, it’ll be like AOL in the end )

                With 951M shares currently outstanding, they once traded at about $100/share (split adjusted), so that their peak valuation was $95 Billion dollars. But let’s exclude the 1990’s, look at their price in say 2005 which was about $50 or so putting them at a cap of say 50 Billion, back when they were a straight-up media company selling advertising on sites.

                They bought a stake in Alibaba in 2007 for not that much money (a Billion or two?), which has now become almost the entirety of the company’s value, but of course they didn’t actually DO anything technology wise to produce value there. In other words, they turned $50 B in media assets into $5B in media assets + a lucky investment.

                In other words, acting as a hedge fund, they transferred huge quantities of your 401k into the pockets of Venture Capitalists who created one startup after another and sold them to you through Yahoo. (Yahoo, owned 76% by institutional investors like the funds in your 401k).

                I think both FB and Google/Alphabet are doing a better job of actually creating some desirable technology. In particular Google is developing the Android operating system and gets consistent revenues from search advertising. FB develops back-end technologies to distribute photos and video and generally makes the job of the National Security Agency cheaper to do, but Google still has a portion of their company whose purpose it is to convert your 401k into VC value (ie. buy up small companies doing drone delivery of burritos or whatever).

                Twitter has done some of the same:


                Other areas to look out for are things like Uber and Lyft and soforth who will try to IPO and become big hedge funds.

              • Rahul says:

                I still don’t get it. If you don’t like the way Google uses its cash, well, sell off your Google stocks. There must be *some* company whose operating decisions you like? If you like none, put your money in another instrument. Gold, maybe. Or treasury bonds. Or real estate. Buy futures, commodities whatever. (Looking at Google’s stock, apparently other investors see it do *something* right that you don’t)

                The whole point behind a liquid stocks market is that you can vote with your feet in a jiffy. You don’t like what Yahoo did? Sell!

                If you don’t like which stocks your 401k fund buys, well isn’t it easier to fix your fund’s policy? There’s probably a mis-incentive for your fund manager. Alternatively what *you* want from the fund doesn’t seem aligned with what the majority of the fund’s investors want?

              • Rahul, I don’t understand your position. You’re basically saying “ok, what I am talking about may be going on, but why don’t I just not participate?”. And I could tell you that, well, maybe I’m not participating, but I still think value is being destroyed and money is being moved to the wrong places due to back-room deals etc. It’s like the mortgage market, sure I could just stay out of that mess, but it doesn’t mean the mess isn’t happening.

              • @Rahul:

                also, this isn’t just about where I want to put my 401k money. The resulting distortion of prices causes people to waste resources at a societal level. There were ghost towns full of 4000 square foot multi-story McMansions in the CA central valley at the time the 2008 mortgage crisis hit. Sure, I didn’t own one, so I didn’t take the major hit of having it foreclosed etc, but the cost of construction labor and construction materials was inflated so regular responsible people who just wanted to replace their roof or add on a room or get their HVAC fixed all had to pay premium prices. etc etc. The repercussions at the societal level are the issue I’m concerned with.

                and, to bring it all back to the original topic. most of this is about everyday people expecting rich people (executives at Google, Facebook, VCs etc) to be right in ways that they pretty clearly aren’t (certainly they weren’t in the Mortgage debacle, and they weren’t in the first dot-com bubble, it remains to be seen if I’m right about the current tech bubble).

      • Keith O'Rourke says:

        > perverse incentives or moral hazards involved in Bayesian or reputational reasoning
        Here, the issue seems to be that past successes may not be reflecting anything much connected with reality, what the underlying conditions were and the decisions taken that were better or worse for those conditions.

        In professional baseball, batters all face a common environment, pitchers vary somewhat as do field conditions but over the season most batters will have faced a common reality and hence a better performance by a batter is less uncertainly based on what they did.

        Business is more like farming – some years there is a bumper crop others years drought. Businesses operated in very non-common environments (e.g. competition my come and go) that are not well characterized. Taking success here as reflecting an ability to do better in a likely different environment seems very shaky.

        My sense of what makes for a lot of success in business is perseverance (ability to get into an area and ride it out as long as one wishes), convincingness (ability to get others to see things in a good light), a willingness to work 12+ hours a day most days of the week, tempered greed and a willingness to make short term sacrifices for long term advantage.

        So if a successful business person is up to bat, and a lame pitch is thrown, they are very unlikely to strike out. A difficult pitch and who knows.

        Its why I find it scary when successful business people go into politics (or any new field).

        Of course, on top of this are the perverse incentives and gaming.

  3. zbicyclist says:

    Complicating this is the fact that people who become successful serial innovators are likely to have some clunkers.

    • Andrew says:


      You’ll also see this in academia. Indeed, I feel a bit of push to continue producing successful research. As a motivator, this is fine, but at the extreme you get those big-time biologists, oldsters with big prizes and massive labs and huge internal pressures to make big discoveries.

  4. “Statistical reasoning would suggest giving the benefit of the doubt to business proposals coming from rich people.” Is it really the case that “statistical reasoning” would lead us to this conclusion? Whenever I see the “person was successful in past venture X therefore they’ll be successful in future venture Y” I think of selection bias — that perhaps success in X is largely random, and we never see the people who failed at it, leading us to falsely conclude that this person’s success is indicative of some general characteristic. How can we apply “statistical reasoning” without knowing the underlying distribution that we’re basing our assessment on?

    Or more succinctly: are we, or the journalists, just victims of an unintentional perfect prediction scam?

  5. I am reminded of G. C. Chesterton’s essay “The Fallacy of Success.” Commenting on a magazine article on how to get rich (which praises Vanderbilt for his money-making instinct), he writes:

    “In such strange utterances we see quite clearly what is really at the bottom of all these articles and books [about success]. It is not mere business; it is not even mere cynicism. It is mysticism; the horrible mysticism of money. The writer of that passage did not really have the remotest notion of how Vanderbilt made his money, or of how anybody else is to make his. He does, indeed, conclude his remarks by advocating some scheme; but it has nothing in the world to do with Vanderbilt. He merely wished to prostrate himself before the mystery of a millionaire. For when we really worship anything, we love not only its clearness but its obscurity. We exult in its very invisibility.”

    • Not to mention that the actual High Speed Rail system is no poster-child for realism either. When that whole thing was voted on I was a CE PhD student at USC, and among CE students in my dept discussing the whole thing it was generally agreed that actual costs would be something like 20-50x as high as the actual ~$10 Bond people were voting on. So far estimates are upwards of $100B so it looks like we’ll be proven correct.

      Note that any day of the week you have around 10 or so flights out of Burbank into Oakland operated by Southwest Airlines that take around 2 hours including waiting in lines etc. Every one of these planes are typically near-full. The tickets cost something like $80 when purchased in advance, and maybe $200 purchased last-minute. A train operating between SF and LA that takes 2 hours and costs $200/ticket will have exactly zero passengers.

      Consider the SF-Oakland Bay Bridge replacement project. Originally estimated in 1996 at $0.25 B for repairs, by 1997 was estimated at $1.2B for replacement with the single tower suspension bridge, actual construction cost $6.4 B so far, with major major issues regarding corrosion of high strength bolts that may mean future costs of several more billion as things need massive early replacement. Some estimate it will ultimately be $13 B before it’s stable.

      I think this is one of those incentives problems. As an engineering/construction firm there is absolutely NO incentive to get your budget right. It is ALWAYS better to under-estimate to get the project approved, get people trapped in with sunk costs and then apply change-orders to milk the system until you’ve extracted every last dollar.

      See Kansai Airport for another example.

      • Rahul says:

        Can’t you structure the contracts differently? e.g. A turnkey or with penalties for cost overruns etc.?

        • honeyoak says:

          The issue here is the commitment problem. Conditional on the project being started, the end owner cannot credibly walk away from a failed project. Any financial structures overlaid on top cannot mitigate this. there is also too much risk (in the real sense) to forecast costs in advance. A financial penalty of sufficient size will not attract sufficient bids.

Leave a Reply