4 ms·

Worth noting that about 30% of active fund managers have beat the S&P 500 going back to 1993 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4096205 https:/

Worth noting that about 30% of active fund managers have beat the S&P 500 going back to 1993 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4096205 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4096205 They may not do so for their investors- this is before fees. But it does seem significant that they are beating the index on their own, and over a consistent period of time. 30% is not nothing. Seems like a blow to strong-form EMH to me. As an FYI, I never want to hear a real-life fund or investor compared to a benchmark again. Benchmarks are theoretical investments with a 0.0% expense ratio- once you add in some of the real costs of running a passive fund, you start to get more real numbers

CuriouslyC 3y ago
Assuming a random distribution of yearly returns centered on 0 before fees, with enough fund managers you'd see plenty that would beat the market for 30 years. That past performance would also be in no way indicative of future performance given the stated process.
unusualmonkey 3y ago
A blind monkey throwing darts will beat the S&P 500 some of the time. The problem isn't that it's impossible to beat the S&P 500 (it's actually trivial), the problem is it's hard to predict which portfolio will outperform the S&P 500.
- Aachen 3y ago
  The point they're making is that fewer than 30% of monkeys would still be positive after so many years if it were random chance Idk if that's true but you're not saying what percentage one would expect to see instead I also think monkeys are the wrong example here because aren't they at even odds with the index? 50% of them, assuming they take no bananas for their service and assuming they don't get to make more trades than the index does, should have beaten the index, if my currently-half-awake brain is working correctly
  - unusualmonkey 3y ago
    > The point they're making is that fewer than 30% of monkeys would still be positive after so many years if it were random chance I'm not sure that's the argument being made, but if so, it's a terrible argument. I'll let the statisticians figure out of randomly ~50% should beat index or not.
hansvm 3y ago
That warrants further investigation, but even if the data is flawless it isn't necessarily a blow to EMH. As a very contrived example (to hopefully illustrate the "risk-adjusted" component of EMH while keeping the math simple), suppose the market consists of many equally sized firms and admits a strategy where each year 0.1% of firms will be uniformly randomly selected to have all their assets wiped out and distributed amongst the rest, and due to social pressures and incentives everyone uses the same strategy. Each year, 99.9% of firms will beat the market [0]. Forward-looking, 97% of funds starting in 1993 will have beaten the market that entire period (backward-looking is less meaningful because it depends on, among other things, how many new entrants to the market there are, not just their performance). Despite the 97% success rate on a 30-year basis, this is still very plausibly a scheme you wouldn't want to participate in (your relative valuations of different outcomes might still make it desirable, but that's a separate question), but it doesn't violate EMH because of the high risk relative to the small returns being achieved. Bringing the contrived example back to the real world a bit, that particular failure mode is common whenever a machine-learning person tries to tackle the market on their own. Even after getting over the hump of price -> bid/ask -> order book -> ... in correctly modeling what's happening, they're still prone to doing things like predicting the chance a security will go up or down and assuming that both branches have equal magnitudes. When they throw it at the real world, they find that despite low false positive and false negative rates for predicting when the price will increase, the times they were wrong were all the high-magnitude events, so they lose money on average. Another way that potentially ties back to the real world, what exactly are the incentives for an active fund manager? When they fail, can they start a new fund? Can they distribute excess losses to a couple of years when the market also did poorly to be able to say something like "every year the S&P 500 went up, we went up more" and still attract new clients? It's not obvious to me that you'd expect behavior which would result in a low chance of beating the S&P 500 over a period of time, even if strong-form EMH holds. [0] This assumes the "market" is static, but the details aren't meaningfully different when you instead benchmark against something like the S&P 500.

CuriouslyC 3y ago

Assuming a random distribution of yearly returns centered on 0 before fees, with enough fund managers you'd see plenty that would beat the market for 30 years. That past performance would also be in no way indicative of future performance given the stated process.

unusualmonkey 3y ago

A blind monkey throwing darts will beat the S&P 500 some of the time. The problem isn't that it's impossible to beat the S&P 500 (it's actually trivial), the problem is it's hard to predict which portfolio will outperform the S&P 500.

Aachen 3y ago
The point they're making is that fewer than 30% of monkeys would still be positive after so many years if it were random chance Idk if that's true but you're not saying what percentage one would expect to see instead I also think monkeys are the wrong example here because aren't they at even odds with the index? 50% of them, assuming they take no bananas for their service and assuming they don't get to make more trades than the index does, should have beaten the index, if my currently-half-awake brain is working correctly
- unusualmonkey 3y ago
  > The point they're making is that fewer than 30% of monkeys would still be positive after so many years if it were random chance I'm not sure that's the argument being made, but if so, it's a terrible argument. I'll let the statisticians figure out of randomly ~50% should beat index or not.

hansvm 3y ago

That warrants further investigation, but even if the data is flawless it isn't necessarily a blow to EMH. As a very contrived example (to hopefully illustrate the "risk-adjusted" component of EMH while keeping the math simple), suppose the market consists of many equally sized firms and admits a strategy where each year 0.1% of firms will be uniformly randomly selected to have all their assets wiped out and distributed amongst the rest, and due to social pressures and incentives everyone uses the same strategy. Each year, 99.9% of firms will beat the market [0]. Forward-looking, 97% of funds starting in 1993 will have beaten the market that entire period (backward-looking is less meaningful because it depends on, among other things, how many new entrants to the market there are, not just their performance). Despite the 97% success rate on a 30-year basis, this is still very plausibly a scheme you wouldn't want to participate in (your relative valuations of different outcomes might still make it desirable, but that's a separate question), but it doesn't violate EMH because of the high risk relative to the small returns being achieved. Bringing the contrived example back to the real world a bit, that particular failure mode is common whenever a machine-learning person tries to tackle the market on their own. Even after getting over the hump of price -> bid/ask -> order book -> ... in correctly modeling what's happening, they're still prone to doing things like predicting the chance a security will go up or down and assuming that both branches have equal magnitudes. When they throw it at the real world, they find that despite low false positive and false negative rates for predicting when the price will increase, the times they were wrong were all the high-magnitude events, so they lose money on average. Another way that potentially ties back to the real world, what exactly are the incentives for an active fund manager? When they fail, can they start a new fund? Can they distribute excess losses to a couple of years when the market also did poorly to be able to say something like "every year the S&P 500 went up, we went up more" and still attract new clients? It's not obvious to me that you'd expect behavior which would result in a low chance of beating the S&P 500 over a period of time, even if strong-form EMH holds. [0] This assumes the "market" is static, but the details aren't meaningfully different when you instead benchmark against something like the S&P 500.