I guess, you can't pull a good analysis of investment strategy only taking into account latest year at any point in time in history.
Every year is different. If quants and value investors did poorly this year it tells exactly nothing about how will they perform the next month or year or decade.
Nobody actually knows the future so everyone else is driving not by observing the road ahead, but by looking to the side and trying to catch some glimps from the corner of their eye.
One reason for sharp dip is people trying to make a profit by selling when its going down and then buying when it has reached bottom or going up. So i guess quantum is bad at predicting human factor, which is the biggest factor, and to be fair - mostly random.
I believe the sharp dip was a "short ladder" as they call it -- hedge funds selling the stock to eachother at low prices to attempt to lower the stock price and trigger some retail investors' stop orders to trigger and free up some shares for cheap.
This is a conspiracy theory spread on social media. Shorting after it has dropped so much already is an unprofitable strategy. These are cascading liquidations and the valid strategy is to go long when the liquidations stop and you get a relief bounce that squeezes out the novice shorts who chased the move. People have different names for the strategy (wall drop long, parabolic long). Shorts are mostly covering (not initiating) during the drop. The only sophisticated traders adding to a short would be options dealers who need to gamma hedge, or delta one HFTs trying to capture a quick move based on short term order flow imbalance.
The tweet is fine and I'm sure many hedge funds and prop firms shorted the initial catalyst. What I'm referring to is the "short ladder" hypothesis mentioned in the parent post which was originally spread on WSB that Justin didn't refer to in the tweet.
As a quant, you are going to be long history and short novelty. I think a massive pandemic plus massive financial stimulus qualifies as novelty. The more history, the more confident you are that things will correct, but also the more history the more exposed you are to a regime change. Think of it like hypothesis testing, if evidence builds up over time it will take more evidence to change the test.
Having said that, timescale is relevant too. The kinds of strategies mentioned are somewhat long term buy-and-holds of certain kinds of portfolio, which require time to play out, and which have the attention of ordinary investors. If you were looking at very short term strategies like what an HFT might look at, the evidence builds up super fast in comparison.
The other big issue is you tend to look at whatever patterns you find that are repeated, but you don't know why. The sort of thoughts that lead you to buy techs during the pandemic come from some kind of idea about dynamics, which are not so easy to prove.
A stock market question. Looking at the outcomes - high-volatility stocks are up, safe stocks are down - makes me suspect this is a "risk bubble".
When watching AlphaGo play Go, you can tell when the network thinks it is ahead or behind. When it is ahead, it will attempt to reduce uncertainty in its evaluation, seeking outcomes where it certainly doesn't lose. When it is behind however, it will attempt to increase uncertainty in its evaluation, seeking outcomes where it doesn't certainly lose. Looking at the market makes me suspect something similar is at play, where the market "should" go down but a lot of investment money is chasing opportunities that "don't certainly lose." In that sense the march drop would be two sequential realizations: "we need to get out of stocks" and "everything else is also fucked", driving investors back into stocks due to a lack of somewhere else to go. Does that sound plausible?
> "we need to get out of stocks" and "everything else is also fucked", driving investors back into stocks due to a lack of somewhere else to go
layman's perspective here: well.. people with money want to invest it somehow, somewhere to increase it.
-It is stock. The stock prices go up and (e.g.) gold price goes down.
-Then some think "ooh gold is so cheap", and cash out their profits from stocks, and drop that money to gold. Stocks go down, gold goes up.
-Then some think "ooh stocks are so cheap", and cash out their profits from gold, and drop that money to stocks. Gold goes down, stocks go up.
-ad infinitum..
Now then.. the big players that move the marker, position themselves early on and ride the full wave. The semi-smart catch part of that wave 'up'. The little fish/retail/impatient tend to buy the stock/gold on its peak, and is paying for the exit-ers profit-cash-out. And then they either cut their losses, or they wait for 1-2-5 years for the next 'peak' to sell to someone else (I know I am oversimplifying it - apologies).
Where stocks, gold please replace with stock[A..Z], bond[A..Z], index[A..Z], commodity[A..Z] and so on..
AlphaGo (or rather AlphaZero) is a bit different because it trains from self-play, where you can get a(n unbiased approximation to the) true answer via Monte Carlo Tree Search.
> driving investors back into stocks due to a lack of somewhere else to go.
Favourited this as insightful, and because this agrees with my own assessment of the situation. You can't take money "out of the market" because there is no outside, just as the environmentalists say you cannot throw things away because there is no away.
There are vast hundreds of billions that cannot exit the market. Everything from oligarchs to 401ks to the Norwegian pension fund (which owns a little chunk of Gamestop!). There is nowhere else to go.
Conversely there are a lot of "outsiders" who are looking for more and more radical strategies where they don't certainly lose, fed by information of increasing volume and deteriorating quality. They're just going to keep trying stuff that doesn't normally work.
I guess it depends what you mean by "the market". A lot of money which previously would have been allocated to public equity credit or bond markets is now being allocated into private investment (PE, VC, private debt, etc etc). A key selling point for those investments is that they are "outside" the (public) markets.
I think the timescales thing you mention is important when interpreting articles like this. The only quant strategies that are sucking are the long horizon, scalable ones available to public funding, which get articles written about them. Shorter term prop quant strategies are making an absolute killing, as they usually do in elevated vol. VIRT's 10Q gives an indication of this, even though they're not quant heavy but still fit into this general category.
People think GME is about money flowing from hedge funds to retail, but in reality the biggest winners are the prop and HFT firms as more retail money enters the market.
> The other big issue is you tend to look at whatever patterns you find that are repeated, but you don't know why.
To what extent?
In my limited experience of quantery, you want to have some idea of the mechanism behind an effect before you start trading on it. How much of the industry is building hypothesis-free models and trading on them?
To a nerd of a generalist science+engineering variety who occasionally short-sells individual stocks, this analysis feels heartening.
Consider:
> quants often rely on measures of valuation, volatility, or momentum to guide their process, based on the historical success of these metrics in predicting returns, and in nearly every case following the numbers led investors astray in 2020, starting with perhaps the most time-tested of financial strategies, value investing.
So, there are volume holders who did not even consider selling when the Diamond Princess news began to trickle out.
Or a few years back, stayed in VW even as technical details of the crude engine controller gaming emerged, making it plain to any hacker how very culpable the company was.
23 comments
[ 3.0 ms ] story [ 64.1 ms ] threadEvery year is different. If quants and value investors did poorly this year it tells exactly nothing about how will they perform the next month or year or decade.
https://twitter.com/justinkan/status/1354861574163320833/pho...
It’s obviously profitable to short a stock when you know the value is going to tank in half an hour.
As a quant, you are going to be long history and short novelty. I think a massive pandemic plus massive financial stimulus qualifies as novelty. The more history, the more confident you are that things will correct, but also the more history the more exposed you are to a regime change. Think of it like hypothesis testing, if evidence builds up over time it will take more evidence to change the test.
Having said that, timescale is relevant too. The kinds of strategies mentioned are somewhat long term buy-and-holds of certain kinds of portfolio, which require time to play out, and which have the attention of ordinary investors. If you were looking at very short term strategies like what an HFT might look at, the evidence builds up super fast in comparison.
The other big issue is you tend to look at whatever patterns you find that are repeated, but you don't know why. The sort of thoughts that lead you to buy techs during the pandemic come from some kind of idea about dynamics, which are not so easy to prove.
When watching AlphaGo play Go, you can tell when the network thinks it is ahead or behind. When it is ahead, it will attempt to reduce uncertainty in its evaluation, seeking outcomes where it certainly doesn't lose. When it is behind however, it will attempt to increase uncertainty in its evaluation, seeking outcomes where it doesn't certainly lose. Looking at the market makes me suspect something similar is at play, where the market "should" go down but a lot of investment money is chasing opportunities that "don't certainly lose." In that sense the march drop would be two sequential realizations: "we need to get out of stocks" and "everything else is also fucked", driving investors back into stocks due to a lack of somewhere else to go. Does that sound plausible?
layman's perspective here: well.. people with money want to invest it somehow, somewhere to increase it.
-It is stock. The stock prices go up and (e.g.) gold price goes down. -Then some think "ooh gold is so cheap", and cash out their profits from stocks, and drop that money to gold. Stocks go down, gold goes up. -Then some think "ooh stocks are so cheap", and cash out their profits from gold, and drop that money to stocks. Gold goes down, stocks go up. -ad infinitum.. Now then.. the big players that move the marker, position themselves early on and ride the full wave. The semi-smart catch part of that wave 'up'. The little fish/retail/impatient tend to buy the stock/gold on its peak, and is paying for the exit-ers profit-cash-out. And then they either cut their losses, or they wait for 1-2-5 years for the next 'peak' to sell to someone else (I know I am oversimplifying it - apologies).
Where stocks, gold please replace with stock[A..Z], bond[A..Z], index[A..Z], commodity[A..Z] and so on..
AlphaGo and the StarCraft thing were trained on ridiculously huge datasets IIUC.
Favourited this as insightful, and because this agrees with my own assessment of the situation. You can't take money "out of the market" because there is no outside, just as the environmentalists say you cannot throw things away because there is no away.
There are vast hundreds of billions that cannot exit the market. Everything from oligarchs to 401ks to the Norwegian pension fund (which owns a little chunk of Gamestop!). There is nowhere else to go.
Conversely there are a lot of "outsiders" who are looking for more and more radical strategies where they don't certainly lose, fed by information of increasing volume and deteriorating quality. They're just going to keep trying stuff that doesn't normally work.
People think GME is about money flowing from hedge funds to retail, but in reality the biggest winners are the prop and HFT firms as more retail money enters the market.
To what extent?
In my limited experience of quantery, you want to have some idea of the mechanism behind an effect before you start trading on it. How much of the industry is building hypothesis-free models and trading on them?
Consider:
> quants often rely on measures of valuation, volatility, or momentum to guide their process, based on the historical success of these metrics in predicting returns, and in nearly every case following the numbers led investors astray in 2020, starting with perhaps the most time-tested of financial strategies, value investing.
So, there are volume holders who did not even consider selling when the Diamond Princess news began to trickle out.
Or a few years back, stayed in VW even as technical details of the crude engine controller gaming emerged, making it plain to any hacker how very culpable the company was.