And I think it is enough to claim that the average of the recommendations was better than any individual stock picker.
> The problem here is that all that information is public and the amount of alpha is scarce.
Not insurmountable, no? I feel these objections are often busied by finance professors, after a student rudely assumes they wouldn't be teaching if they knew how to make money.
RenTech was doing speech recognition on foreign TV broadcasts in the 90s. Can't you think of similar features you could whip up in a 100 lines of Python and the YouTube API? Some very profitable companies hire very smart PhDs to that for them. Could you bootstrap this? Work harder? Extend to some new hip platform a well-paid quant has never heard of?
> trying to get a piece of the action, diluting or even eliminating the benefit for everyone.
So join in. You are aware of the action. It's not like we are doing this for a bit of fun. The crowd wisdom is there, the hedge funds don't have a monopoly on polling it or analyzing it. Then redistribute the wealth for the benefit of everyone. Those very hard in on the action are not going to.
> but doesn't work if they're clueless and are all trying to buy into the same bubble.
Yes, this is correct, and a big problem in crowd analytics. Or at least, it has a big negative effect (you ideally want everyone to make decisions of their own accord, using their own information). But you can also again harnass this with counter trading strategies. Over the years, it has been fairly easy to call the top of a hype, and predict the obvious correction. So for instance, if the Teletubbies twitter is tweeting about Bitcoin, you know that maybe now is time to sell some Bitcoin, and rebuy back in 6 months when all newspapers are writing about how Bitcoin is a scam and a world-wide crypto ransom attack just occured.
The problem can be overcome in a couple of ways. An interesting one is: "skin-in-the-game". If your wrong hype predictions damages your reputation or causes money loss, you are more serious about it. Another is to ask: What percentage of other people do you think got this question wrong? People who answer Sydney as the capitol of Australia, think that few got it wrong. People who give the correct answer think that many will get it wrong.
> You do realize that the stock market is exactly that, right?
Partly yes. The difference with crowdsourcing is that you are building a model on top of the other participants. Many day traders with bots do not know that hedge funds have models more complex tracking what they are doing and going to do, than the bot is complex. But many day traders also underestimate how they'd stack up against an office of suits, if they worked together, and ramen-noodle hacked it.
>The winning one did not manage to achieve results better than the baseline that was constructed as the maximum from ‘always buy’ and ‘always sell’ strategies over the testing period
Like I said, the benchmark isn't against a monkey randomly picking a stock, it's against a passive portfolio.
>Not insurmountable, no? [...] RenTech was doing speech recognition on foreign TV broadcasts in the 90s. Can't you think of similar features you could whip up in a 100 lines of Python and the YouTube API?
>So join in. You are aware of the action. [...]
That goes back to my remark about how alpha is limited. If you whipped up a script but was 10 minutes late to the party, then you'd get zero gains because someone has already beat you to the punch. The hard part isn't good data, it's getting good data that everyone else doesn't have.
>But you can also again harnass this with counter trading strategies. [...]
that almost sounds like "timing the market", which is generally known to not work for unsophisticated investors.
>The problem can be overcome in a couple of ways. An interesting one is: "skin-in-the-game". If your wrong hype predictions damages your reputation or causes money loss, you are more serious about it
The fact that they're publishing the trading strategies for anyone to use for free should tell you all you need to know about their "skin in the game". If they actually thought they had real alpha, they'd be harnessing it all for themselves, not publishing it on the internet and getting freeriders to dilute it.
>Another is to ask: What percentage of other people do you think got this question wrong? People who answer Sydney as the capitol of Australia, think that few got it wrong. People who give the correct answer think that many will get it wrong.
[insert joke about how economists have predicted 15 of the last 3 recessions]
>But many day traders also underestimate how they'd stack up against an office of suits, if they worked together, and ramen-noodle hacked it.
I don't doubt a bunch of random internet traders can outsmart an office of suits, I just doubt they can produce better returns (% wise) because their open planning (ie. posting their trades online for anyone to see) allows their alpha to be diluted and get front-runned.
> Like I said, the benchmark isn't against a monkey randomly picking a stock, it's against a passive portfolio.
At least we went from beating a fair coin with a crowd to benchmarking against a passive portfolio :)
The quote just means that the winning model was overfit to leaderboard. Not that there weren't better models submitted than the baseline, for else someone submitting the baseline would have won.
> If you whipped up a script but was 10 minutes late to the party, then you'd get zero gains
See, I really don't believe this. You can use the script features for longer timelines, then there is no rush. You are also not as much competing with the smart money, as you are joining them. I really don't think those few 1000s $ someone early on spends, are taking any action of the table. And the big players can't take it all.
> that almost sounds like "timing the market", which is generally known to not work for unsophisticated investors.
but yes, would need both sophistication and time investment, else passive way more attractive. But you can use crowdsourcing on overhyped trends as information for this strategy.
> should tell you all you need to know about their "skin in the game"
I am not saying that newspaper stock pickers solved the problem of mindless hype. But skin-in-the-game or staking is an effective counter against clueless people following trends, and moves it further away from the coin flip.
> [insert joke about how economists have predicted 15 of the last 3 recessions]
[Insert fact about how those predictions cluster around the last 3 recessions]
> allows their alpha to be diluted and get front-runned
If anything, this would happen at the real front: hedge funds would contract someone to contract an expensive New York PR firm to meme and astroturf some stock, say Palantir, to something noteworthy. Then the open planning online for anyone to see does the rest. The private Discord channel of random traders (who bought into that group with 5 BTC to have skin-in-the-game) knows what's up, and quietly rides along.
>At least we went from beating a fair coin with a crowd to benchmarking against a passive portfolio :)
I agree it's not a perfect analogy, but the underlying point holds: there's inevitably going to be some that come out on top because of luck, rather than actually skill.
>The quote just means that the winning model was overfit to leaderboard. Not that there weren't better models submitted than the baseline, for else someone submitting the baseline would have won.
So the study is inconclusive at best when it comes to forward-looking results?
>See, I really don't believe this. You can use the script features for longer timelines, then there is no rush. You are also not as much competing with the smart money, as you are joining them. I really don't think those few 1000s $ someone early on spends, are taking any action of the table. And the big players can't take it all.
The problem is that once the information has been priced in, there isn't anything to gain anymore. For instance, let's say company A shares are selling for $100 each, and you found out that company B is planning to acquire company A at $120/share. If you found out this information before anyone else and traded on it, you can make a profit of $20/share (ignoring using leverage or derivatives). However, if you were late to the party, ie. found out this information after others did, and the price already rose to $120, then there's nothing to gain from it. Buying the shares at $120 and then selling it at $120 doesn't net you any profit. This example uses an acquisition as a pricing event, because it provides an unambiguous price target, but this can extend to other events as well, such as finding out this quarter's earnings is above/below estimates.
>but yes, would need both sophistication and time investment
therein lies the problem. There's no free lunch here. You can't blindly apply the rule and get stacks of cash. Citing "countertrend" in this case is only marginally more helpful than citing "buy low sell high".
>[Insert fact about how those predictions cluster around the last 3 recessions]
A recent working paper by Zidong An, Joao Tovar Jalles, and Prakash Loungani discovered that of 153 recessions in 63 countries from 1992 to 2014, only five were predicted by a consensus of private-sector economists in April of the preceding year. And the economists tended to underestimate the magnitude of the slump until the year was almost over.
And I think it is enough to claim that the average of the recommendations was better than any individual stock picker.
> The problem here is that all that information is public and the amount of alpha is scarce.
Not insurmountable, no? I feel these objections are often busied by finance professors, after a student rudely assumes they wouldn't be teaching if they knew how to make money.
RenTech was doing speech recognition on foreign TV broadcasts in the 90s. Can't you think of similar features you could whip up in a 100 lines of Python and the YouTube API? Some very profitable companies hire very smart PhDs to that for them. Could you bootstrap this? Work harder? Extend to some new hip platform a well-paid quant has never heard of?
> trying to get a piece of the action, diluting or even eliminating the benefit for everyone.
So join in. You are aware of the action. It's not like we are doing this for a bit of fun. The crowd wisdom is there, the hedge funds don't have a monopoly on polling it or analyzing it. Then redistribute the wealth for the benefit of everyone. Those very hard in on the action are not going to.
> but doesn't work if they're clueless and are all trying to buy into the same bubble.
Yes, this is correct, and a big problem in crowd analytics. Or at least, it has a big negative effect (you ideally want everyone to make decisions of their own accord, using their own information). But you can also again harnass this with counter trading strategies. Over the years, it has been fairly easy to call the top of a hype, and predict the obvious correction. So for instance, if the Teletubbies twitter is tweeting about Bitcoin, you know that maybe now is time to sell some Bitcoin, and rebuy back in 6 months when all newspapers are writing about how Bitcoin is a scam and a world-wide crypto ransom attack just occured.
The problem can be overcome in a couple of ways. An interesting one is: "skin-in-the-game". If your wrong hype predictions damages your reputation or causes money loss, you are more serious about it. Another is to ask: What percentage of other people do you think got this question wrong? People who answer Sydney as the capitol of Australia, think that few got it wrong. People who give the correct answer think that many will get it wrong.
> You do realize that the stock market is exactly that, right?
Partly yes. The difference with crowdsourcing is that you are building a model on top of the other participants. Many day traders with bots do not know that hedge funds have models more complex tracking what they are doing and going to do, than the bot is complex. But many day traders also underestimate how they'd stack up against an office of suits, if they worked together, and ramen-noodle hacked it.