Thanks for sharing your experience—the third section in particular is really interesting. Relative to the central discussion: how much time and effort did you have to put in to find those edges, and do you think it was worth it? Would you encourage others to try and do something similar? Or is it more like a hobby?
Richard Meadows
Karma: 593
Your formulation is nifty, and intuitively makes sense to me. I am feeling too wiped right now to think about it carefully, but my off-the-cuff response is that it has something to do with the fact that ‘informational edge’ is a much broader category than information about the actual underlying assets.
For example, a complicated day-trading algorithm is on some level a reflection of the fact that the market is missing some information. But that information looks more like ‘there is a complex relationship between assets under XYZ specific conditions’ than ‘the EBTDA figure in Walmart’s quarterly report’. I guess this is what most anomalies represent: they are more like meta-information than information.
In which case, if you had a superintelligence with near-infinite compute, I think your intuition that it could indeed beat the market is right. I don’t know much about quants, but I imagine this is pretty much what they are trying to do, with the difference being that they have bounded resources.
Whoops, thanks, will fix that now.
Hi Scott,
If we take the 6th of March—the last trading day before the March 9 fall—then the market is down 12.2%, which is already in ‘correction’ territory, and an extremely rapid descent by historical standards.
If we take the 16th of March—the closest trading day we have to ‘mid-March’—the market is already down 29.5% per cent, which is not too far off the bottom, and well and truly into ‘bear’ territory.
To be clear, I think time has proven you correct about the EMH (and this is easy for you to say, now that the market has stabilized).
I thought the talk about EMH being dead was weird at the time, and left a comment saying as much. I also wrote a post on March 24, which later turned out to be the bottom, saying that timing the market was a really bad idea, and the buy-and-hold forever strategy was about the best anyone could hope for. I am as surprised about the speed of the rebound as anyone! I possess no predictive powers, but I have consistently been defending boring orthodoxy.
I admire and respect you very much as a person and a thinker—seriously man, you have no idea—so I feel extremely bad if I have made you feel bad. I didn’t mean to accuse you specifically of being a revisionist historian—it was more of a general vibe that seemed to be happening a lot—and although I think the passage as stated is misleading, I don’t think it’s deliberately so, and I’ve edited the post so it comes off as less accusatory.
Huh, I got that from a recent Bloomberg article which says 15:1...not sure who’s right or why the numbers are so different.
Active management in the equity market, both in the U.S. and abroad, is dominant. And not by a little: Active management in the U.S. trounces passive by a ratio of 8-to-1 in dollar investments. Expand that to include the entire world, and the ratio is closer to 15-to-1. If we include fixed income in our calculations, the ratio balloons to 60-to-1.
Second, and it relates to the first, one of the other things (different time) that was pointed out was “market” is a problematic terms. Each asset that is traded has a market and that is not the same as “the market”. I think this tends to be something of a problem area for people when the issue of EMH is in question.
Yes! This is another really great point. I think Noah Smith described it by saying something like, ‘there is no EMH—there are an infinite number of EMHs all happening at the same time’. Which is another reason the theory is vague and unfalsifiable.
I find it helpful to think about each market in the narrowest possible terms. For example, the market for AAPL stock is likely to be less efficient than the S&P 500 market as a whole, although presumably not by much. The market for a thinly-traded stock languishing on the secondary board of the Tanzanian stock exchange is likely to be much less efficient than that. The market for a private company with no disclosure obligations is much less efficient again. By the time you get down to ‘the market for the time and labour of this one guy called Richard’, there are truckloads of inefficiencies and EMH doesn’t really apply.
Thanks, nice post. I like ‘anti-inductive markets’, not least because it doesn’t come with all the confusing connotations of ‘efficient’.
Sure—I was testing a dual momentum strategy over the market as a whole, with a 12-month lookback period. The ‘dual’ refers to both absolute momentum, and relative momentum between asset classes (bonds, US stocks, non-US stocks).
I haven’t evaluated it properly yet, but the signals it generated told me to stay in stocks until the end of March, at which point I ought move into bonds, just in time to miss the recovery. Over the period I’ve tested it so far (16 months) it has returned −4%, while my benchmark is at +18%. I am still mildly interested to see how it pans out, but I ignored the signals and am now only tracking it on paper.
Yeah, active investors are providing a valuable service to everyone else, both by exploiting genuine asymmetries, and by collectively generating a signal in the Uncle George sense. People sometimes worry about the passive revolution for this reason, but the vast majority of funds under management are still active, and human nature being what it is, there will presumably always be plenty of people willing to have a crack.
I think I agree with this point and I’m glad you raised it—you might non-trivially beat the market every once in a while, but not all the time. To be clear, I don’t think of the EMH as an on/off switch, and if the post gives that impression, that’s my bad. In fact, the most successful investors I know of wait, wait, wait, and then very occasionally load up when they see a major asymmetry.
But I do still have a couple of reservations. The first is that this would still show up in the evidential standards I’ve suggested? If you match the market most of the time, but every five years you beat it, you’re eventually gonna have a pretty amazing track record, especially as you compound early successes. Note this is more or less the same frequency offered by some anomalies—for example, the momentum strategy I experimented with only outperforms during downturns, but ends up with a super impressive CAGR.
The second question is, how confident are you (‘you’ in the general sense, but also ‘you’ Vaniver if you feel inclined to answer) that knowing the difference between the edge trades and the others actually helps? If it’s a very strong feeling, why trade on all the other occasions when you know you’re guessing?
I think it’s interesting in an academic sense that these edge trades might really exist, but if they still ultimately result in an unexceptional record because of all the usual human foibles, I’m not sure how it practically justifies the suggestion that anyone should do more of this sort of thing?
Each of the ‘edge’ investments has also had another side to it, and the other side didn’t have the same feeling to guide me. For example, I correctly timed BP’s bottom during the Deepwater Horizon crisis, but then when it recovered my decision of when to sell it was essentially random. I think most of the people who predicted COVID a week early were then not able to outperform the market on the other side, and various things that I’ve seen people say about why they expect a continued edge seemed wrong to me.
This is a super important point too. Any investor who is long the stockmarket on a long horizon ultimately has to get back in, which means they have to time the market twice. Personally, I was gobsmacked by the speed of the recovery (or the height of the dead cat bounce, if that’s what it is). April was the stock market’s best month in 30 years, which is not really what you expect during a global pandemic.
Crypto and domain names might be in a different class, for sure—I don’t know how those markets work, although I’d be interested to hear more. But options contracts that involve e.g. the fortunes of the S&P 500 are not.
(I want to encourage rationalists to look for future opportunities to trade on, too. I just think those opportunities are vanishingly unlikely to come from highly liquid securities markets, and a lot of the recent talk is going to lead to misplaced effort.)
I don’t think in any of the three cases in which Wai Dai made very high returns he would have had a problem with telling other people about his investment and it would make sense to behave as you suggest a Realistic winners would (I will of course never tell anyone about it, or it will become useless).
Yeah, I’m still not exactly sure about whether anomalies can persist in public (see one of my confusions below). But if the trade is based on a one-off event, like this one, it’s obviously fine to tell people about it. If it’s a formula for an ongoing edge that persists in [stock-picking/domain name-picking/crypto] then not so much.
Confusion: Can an anomaly be an open secret, and not defensible, and still somehow persist?
For example, I’m especially interested in the momentum anomaly. The momentum folks make a very compelling case that the behavioural econ reasons that brought it into existence are so dang powerful that it somehow persists, even though there are many papers/articles/books about it that anyone can read.
I personally found this just compelling enough to run a momentum experiment, based on a strategy that has done extremely well in reducing volatility and drawdowns in the past. Momentum strategies naturally lag bull markets, but more than earn their keep during the downturns, so this was the long-awaited test.
...and it performed almost comically badly. Either the anomaly evaporated, or the specific version I was testing evaporated, or it did badly this one time but still outperforms over multiple cycles, or it was only ever an artifact of overenthusiastic back-testing.
If open secrets can persist in public… for how long? What are the factors that lead to them ultimately not working? Is it possible that an anomaly could exist in public indefinitely?
Confusion: Can multiple traders profit by bringing the exact same information to market?
I want to say yes? At least if they don’t individually have enough heft to fully price it in? That would explain how multiple traders are able to exploit the same anomaly, whereas hedge funds have to cap their fund size because getting too big hurts them.
Confusion: Is there some (incredibly weak) sense in which every person who makes a trade is bringing unique information to market?
Uncle George’s opinion of the new iPhone is reflected in sales figures and aggregate reviews. But technically, he is in possession of a unique piece of private information: ‘what does this one guy called Uncle George think about AAPL stock?’
This information is vanishingly close to being worthless, and a million miles way from a tradable edge, but it’s not quite worthless—it still collectively helps to generate a signal. What does this mean, if anything?
You should certainly try to test any edge you think you have, and look for missing information. If you develop a good edge it’s easy to make some profit, but it’s always easier to lose money if you’re careless. But don’t give up before you even try.
Yep, I agree with all of this. I guess I way I would phrase it is that we don’t start with a flat prior: we have mountains of evidence that most investors underperform, and that finding an edge is difficult. Doesn’t mean it’s not possible, and absolutely you should try, so long as you’re taking very careful steps not to fool yourself about performance, benchmarking appropriately, etc.
On the ‘open secrets’ - I’m writing a big effort-post right now, and this is one of my main points of confusion - would appreciate your input once it’s up.
Obviously some people have made money trading stocks. Does the EMH simply mean that less than 50% of people who trade stocks make money? That doesn’t seem to support the grandiose conclusions that are usually made on the basis of the EMH.
Yeah exactly—for example, something like 90% of active fund managers (professional investors with all the bells and whistles) fail to beat their benchmark, and those that do are highly unlikely to repeat the feat the next year. It makes no difference to me that EMH doesn’t cash out in some kind of precise formula—it just seems like a super useful and interesting thing to know. Sorry if we’ve been talking at cross-purposes!
Yep, true—hence high-frequency traders. I remember reading how one firm installed a direct cable between Chicago and New Jersey at some incredible expense to shave a few milliseconds off transmission time.
What would be an example of energy not being conserved in a closed system? Does the law of thermodynamics even mean anything?
I’m not sure what you’re trying to say, so it would probably be better to just state your point plainly.
Like anything else, the EMH is useful insofar as it generates testable predictions about the world. One of the most useful predictions, as johnswentworth puts it: ‘you shouldn’t expect to make money trading stocks’.
The onus is on the person making an extraordinary claim to provide the evidence, not the other way around.
If you think you’ve found an exploit in the market you should absolutely start from the position that you’re wrong, because… you almost certainly are. This is how anyone ought to behave purely out of naked self-interest—it has nothing to do with confirmation bias.
Plus one! I tried several free tools of this nature, but managed to find loopholes and self-sabotage every single time. Shelled out 20 bucks or something for freedom, and solved the problem instantly.