He presents evidence that primate emotion evolved from interceptive awareness - goes into fascinating detail about what we know of the underlying, homeostatic, neuroanitomical structures and their evolution.
What happens if you combine physiological signals with machine learning to create an application that trades based on a human's "gut" along with other input, and perhaps historical data?
You can imagine a man vs man-machine competition where a trader's regular profit/loss is compared in real-time with the "enhanced" p/l.
What could be done if you network several traders' physiological signals and mix with machine learning and automated trading?
Is anyone doing this kind of frankentrading? Could even imagine going for scale with an Apple Watch physio-trading app :0
>We monitored these traders during a particularly volatile period (the tail end of the European sovereign debt crisis) so the performance of each trader reflected his ability to make money during periods of extreme uncertainty.
Shouldn't the market state be a pretty big experimental control?
If we're coming out of a downturn and the market is going up, on average anybody taking a financial risk would be rewarded for doing so.
It would be interesting to see the same group in a financial situation in which the market is at a peak and headed down. Are these traders who have greater interoceptive ability able to take the right risks since, on average, those who take on risk before a downturn lose money?
This is a great question. Bull markets (in retrospect) select for aggressive risk-takers without much regard for their sophistication, and bear markets the reverse. You'd basically want to perform this analysis on everyone doing this type of trading over a period of like 25 years.
Therein is the issue with a lot of social and behavioral science: the really really interesting stuff is just so hard to actually measure and experiment with.
My gut feeling says it depends on the trading strategy of the firm they used for the testing.
Traders don't need a bull market to make money, they just need market volatility.
If the market is going up they can bet on the value of stocks increasing, if the market is going down they can bet on the value of stocks going down.
If the firm has a policy of being long only and traders have to hold their trades for a long time then yes, the direction of the market is going to affect the performance of traders.
More information about trading strategy and limitations are necessary to really tell.
These guys were trading futures contracts. Unlike stocks, when you are long a futures contract, someone else is necessarily short, so it won't be the case that "anyone taking financial risk would be rewarded".
There is always someone 'short' on a trade, and another party 'long', but nobody is necessarily 'shorting' a future for someone else to be buying it (as the seller may be a producer selling future production).
Not necessarily - for stock index futures, the seller of the contract can simultaneously buy the index of stocks. When the future contract reaches maturity the seller has to "deliver" the basket of stocks to the buyer. Banks and market makers make money writing (selling) such index futures by buying a smaller basket of stocks than are strictly in the index to get the same hedge effect (using principle component analysis, yay!), consolidating an entire portfolio of contracts and having a single hedge, and also by having lower transaction costs to buy/sell the underlying stocks than the clients buying/selling the future would, because they already have all the expensive exchange memberships, they also may have inventory, and they have dark pools, in which their own traders get first priority.
I think that I've been able to observe a similar situation in my experience. I've had the fortune of being at the point of various crises in my career (massive system failures, natural disasters, man-made disasters, etc) in both contributor and leadership roles.
When in those type of situations, I definitely have perceived a heightened awareness, where the feeling is like being in the "zone". Decisions are easier, communications more fluid. I tend to be a more instinctual person in general, but these types of situations are far more pronounced.
What's interesting is that it's highly situational for me. I'm known as a fixer for work related situations but I'm pretty risk adverse in my personal life.
I'm an awful negotiator in my personal life for cars, etc. I leave money on the table all of the time. At work, a few guys dubbed my face over William Shatner from a Priceline "negotiator" commercial and played it for the christmas party after a particularly tough deal. :)
"Our results suggest that signals from the body ... contribute to success in the markets"
Does it suggest that physiological emotions help traders make good decisions? Or does it suggest that conscious awareness of physiological emotions helps traders compensate for gut biases and make more rational decisions?
I'm strongly guessing the latter. Panic selling is a great example of your gut reaction losing you extraordinary amounts of money. If you are aware of the fight-or-flight, it's probably easier to convince yourself to think twice.
The authors don‘t draw any conclusions either way ("Our study, being field work, could not establish causation"), but they seem to lean toward the former with this:
Traders in the financial world often speak of the importance of gut feelings for choosing profitable trades. By this they mean that subtle physiological changes in their bodies provide cues helping them rapidly select from a range of possible trades the one that just ‘feels right’. Our findings suggest that the gut feelings informing this decision are more than the mythical entities of financial lore - they are real physiological signals, valuable ones at that.
The suggestion here is that there is a way in which the mind/body are able to rapidly process large amounts of information, beyond conscious analysis/calculation, and report an answer as a "gut feeling", and that the people who have a stronger ability to sense their physiological functions are the people for whom this mechanism is more accurate.
This notion seems analogous with the thinking behind practices like meditation and other mind+body "tuning in" exercises that have long been used in martial arts training to develop better spatial awareness and more precise reflexes.
Interesting paper, but I have issues with some of the statistics--not helped by the large number of seemingly exploratory analyses that give the impression of a garden of forking paths.
1) The authors go through the work of doing several simpler analyses which are obviated by the final multiple regression. The analyses (in Wilkinson notation) P&L[0] ~ HBD[1] and HBD ~ covariates only provide a subset of the information in the final P&L ~ HBD + covariates. The paper would improve if the preliminary analyses were replaced by a more thorough treatment of the multiple regression.
2) From what I gather, the maximal model containing HBD and the covariates was fit and then variables were subtracted successively based on whether doing so reduced the AIC. This does not seem like the right choice at all, and could possibly have influenced the final result. Selecting an optimal model is not the goal here, yet that is what this process is designed to do. The research question is about the influence of HBD on P&L, not predicting P&L from some optimal combination of variables. So, the theoretically important covariates should have remained when testing the significance of P&L. The point of the covariates is to "control" for confounding factors or "stress test" the importance of HBD in the presence of confounds, not to increase the predictive power of the model.
3) Trader and session length are actually random variables but their error is not modeled. Instead, the results from all sessions are averaged for each trader. Taking the average has the result of undercounting the variance and misspecifying the variance structure. Something like a multilevel model with crossed random effects should have been used instead.
[0] "Average daily profit and loss" - measure of trading performance.
[1] "Heartbeat detection score" - how well the trader self-assesses their heartbeat.
It's a fairly small sample. However, in theory it's still possible to detect an effect in a population using a small sample if the effect size is large enough. Unfortunately, the paper doesn't present a measure of effect size, e.g. standardized regression coefficients.
There's a great anecdote about George Soros that this reminded me of: Despite the palindrome writing two dense books and conducting numerous interviews on his theory of 'Reflexivity'[0] as essential to his success, Soros' son thinks it really comes down to something like interoceptive ability, which Soros reasons is a combination of "theory and instinct"[1]:
.
.
According to his son, Robert, Soros’s trading was always influenced by more than reflexivity. “My father will sit down and give you theories to explain why he does this or that”, he once said, “but I remember seeing it as a kid and thinking, ‘Jesus Christ, at least half of this is bullshit’.
“I mean, you know [that] the reason he changes his position on the market or whatever is because his back starts killing him. It has nothing to do with reason. He literally goes into a spasm and it’s this early warning sign.”
Soros has admitted to relying greatly on “animal instincts”, saying the onset of acute pain was often “a signal that there was something wrong in my portfolio”.
His decisions, then, “are really made using a combination of theory and instinct”.
24 comments
[ 339 ms ] story [ 103 ms ] thread[1] https://www.bloomberg.com/view/articles/2016-09-19/tough-tar...
It lead me to http://www.nature.com/neuro/journal/v7/n2/full/nn0204-102.ht...
Which formalizes the directions I've started to think about, namely the connection and feedback loop between physical sensation and emotion.
He presents evidence that primate emotion evolved from interceptive awareness - goes into fascinating detail about what we know of the underlying, homeostatic, neuroanitomical structures and their evolution.
What happens if you combine physiological signals with machine learning to create an application that trades based on a human's "gut" along with other input, and perhaps historical data?
You can imagine a man vs man-machine competition where a trader's regular profit/loss is compared in real-time with the "enhanced" p/l.
What could be done if you network several traders' physiological signals and mix with machine learning and automated trading?
Is anyone doing this kind of frankentrading? Could even imagine going for scale with an Apple Watch physio-trading app :0
Shouldn't the market state be a pretty big experimental control?
If we're coming out of a downturn and the market is going up, on average anybody taking a financial risk would be rewarded for doing so.
It would be interesting to see the same group in a financial situation in which the market is at a peak and headed down. Are these traders who have greater interoceptive ability able to take the right risks since, on average, those who take on risk before a downturn lose money?
Therein is the issue with a lot of social and behavioral science: the really really interesting stuff is just so hard to actually measure and experiment with.
Traders don't need a bull market to make money, they just need market volatility.
If the market is going up they can bet on the value of stocks increasing, if the market is going down they can bet on the value of stocks going down.
If the firm has a policy of being long only and traders have to hold their trades for a long time then yes, the direction of the market is going to affect the performance of traders.
More information about trading strategy and limitations are necessary to really tell.
When in those type of situations, I definitely have perceived a heightened awareness, where the feeling is like being in the "zone". Decisions are easier, communications more fluid. I tend to be a more instinctual person in general, but these types of situations are far more pronounced.
I'm an awful negotiator in my personal life for cars, etc. I leave money on the table all of the time. At work, a few guys dubbed my face over William Shatner from a Priceline "negotiator" commercial and played it for the christmas party after a particularly tough deal. :)
Does it suggest that physiological emotions help traders make good decisions? Or does it suggest that conscious awareness of physiological emotions helps traders compensate for gut biases and make more rational decisions?
Traders in the financial world often speak of the importance of gut feelings for choosing profitable trades. By this they mean that subtle physiological changes in their bodies provide cues helping them rapidly select from a range of possible trades the one that just ‘feels right’. Our findings suggest that the gut feelings informing this decision are more than the mythical entities of financial lore - they are real physiological signals, valuable ones at that.
The suggestion here is that there is a way in which the mind/body are able to rapidly process large amounts of information, beyond conscious analysis/calculation, and report an answer as a "gut feeling", and that the people who have a stronger ability to sense their physiological functions are the people for whom this mechanism is more accurate.
This notion seems analogous with the thinking behind practices like meditation and other mind+body "tuning in" exercises that have long been used in martial arts training to develop better spatial awareness and more precise reflexes.
1) The authors go through the work of doing several simpler analyses which are obviated by the final multiple regression. The analyses (in Wilkinson notation) P&L[0] ~ HBD[1] and HBD ~ covariates only provide a subset of the information in the final P&L ~ HBD + covariates. The paper would improve if the preliminary analyses were replaced by a more thorough treatment of the multiple regression.
2) From what I gather, the maximal model containing HBD and the covariates was fit and then variables were subtracted successively based on whether doing so reduced the AIC. This does not seem like the right choice at all, and could possibly have influenced the final result. Selecting an optimal model is not the goal here, yet that is what this process is designed to do. The research question is about the influence of HBD on P&L, not predicting P&L from some optimal combination of variables. So, the theoretically important covariates should have remained when testing the significance of P&L. The point of the covariates is to "control" for confounding factors or "stress test" the importance of HBD in the presence of confounds, not to increase the predictive power of the model.
3) Trader and session length are actually random variables but their error is not modeled. Instead, the results from all sessions are averaged for each trader. Taking the average has the result of undercounting the variance and misspecifying the variance structure. Something like a multilevel model with crossed random effects should have been used instead.
[0] "Average daily profit and loss" - measure of trading performance.
[1] "Heartbeat detection score" - how well the trader self-assesses their heartbeat.
.
.
According to his son, Robert, Soros’s trading was always influenced by more than reflexivity. “My father will sit down and give you theories to explain why he does this or that”, he once said, “but I remember seeing it as a kid and thinking, ‘Jesus Christ, at least half of this is bullshit’.
“I mean, you know [that] the reason he changes his position on the market or whatever is because his back starts killing him. It has nothing to do with reason. He literally goes into a spasm and it’s this early warning sign.”
Soros has admitted to relying greatly on “animal instincts”, saying the onset of acute pain was often “a signal that there was something wrong in my portfolio”.
His decisions, then, “are really made using a combination of theory and instinct”.
[0]: http://www.mercenarytrader.com/2014/02/lessons-from-the-pali...
[1]: http://bclund.com/2014/08/20/nothing-can-learn-george-soros/