You need to read stuff like this

[DominiConnor]

It's a competitive market, now more than ever and you need to stick out more. That means when you pitch yourself at interview being able to talk about things beyond what is on the curriculum that every other bright yung thing has learned.
Game theory of price formation always plays well when talking to traders.

The Game Theory of How Algorithms Can Drive Up Prices | Quanta Magazine

Recent findings reveal that even simple pricing algorithms can make things more expensive.

www.quantamagazine.org

 

[rajivp]

interesting .. even without being a ml or game theory expert, article’s gist & conclusion felt obvious.
if everyone uses similar data i.e. same sources because there is only enough data in the world to train your model from and similar model; algorithms will just converge .. learn to raise prices together or lower (whatever works) without ever talking.
It makes those tiny arbitrage windows disappear.
Feels like the field is converging - as there are only limited kind of strategy, data, models - same math (which eventually gets replicated) unless someone brings in something totally new, like quantum physics or a new kind of math.

 

[Daniel Duffy]

It's a competitive market, now more than ever and you need to stick out more. That means when you pitch yourself at interview being able to talk about things beyond what is on the curriculum that every other bright yung thing has learned.
Game theory of price formation always plays well when talking to traders.

The Game Theory of How Algorithms Can Drive Up Prices | Quanta Magazine

Recent findings reveal that even simple pricing algorithms can make things more expensive.

www.quantamagazine.org

Nice to see you again on Quantnet, Dominic.

 

[DominiConnor]

Yes we learn from biological systems the convergence you cite. Wildly different organisms evolve highly similar ways of doing things.
It's an interesting conflict of intuitions. The complexity of even small programmes makes one expect there to be very many different of about the same efficiency.
Regulation since the crisis nearly 20 years ago has stifled innovation so as people learn from each other in a constrained environment competitive advantage is ever harder to build.
AI is useful boiler plate, assembling code for tasks already programmed, new things not so much.
Certainly new maths (or at least the rediscovery of old like discrete mathematics), may change things but regulation may well stifle that.

There is some sort of tipping point out there, can't say when or here, but if you have a culture that follows traditional ways for too long, someone following different rules will screw you big time. I learned different lessons about the way the the British and to a lesser extent Europeans were able to take over the world, committing massive thefts, genocides and famines as they went despite being outnumbered 15 or 20 to 1.
They did it because they could.
Banks compete strongly with each other, as did Native Americans nations and Indian Indians, but within tight boundaries.

Innovation is a social process, sure you need smart motivated people and banks have them by the truckload, I've read their resumes.
But as institutions they have lost the appetite for disruptive innovation and the culture that nurtures.