
Any opinions/comments on this paper? Optimal market making 
Nous promettons selon nos espérances, et nous tenons selon nos craintes. 



85 views no comments? I'd really appreciate an educated view on the model this paper presents. 
Nous promettons selon nos espérances, et nous tenons selon nos craintes. 

gaj


Total Posts: 103 
Joined: Apr 2018 


Skimmed through the paper. For this to be useful in a high dimensional setting, closedform approximation is indispensable. But it seems that the closedform is only given for the uninteresting case of T>infinity. Also not sure how much value there is in solving this optimal control problem vs using simple heuristics.
I feel like a lot of quant approaches fall into one of two categories: 1) model the full dynamics of the system; calibrate the parameters; then solve for the function that you need. 2) parametrize the function that you're looking for with a number of inputs; define an objective function; use supervised learning to find the optimal solution.
Andrew Ng wrote something similar in the context of classification, where he called approach 1 "generative" and approach 2 "discriminative". A lot of fancy papers are written in (1). It usually requires more assumption, harder to calibrate, and rarely gives a closedform solution. On the other hand, approach (2) by design provides a closed form solution and is easier to solve. My impression is that most practitioners use approach 2 in the real world. 




That's very interesting to learn, I was not aware it was possible to classify the problem like that. 
Nous promettons selon nos espérances, et nous tenons selon nos craintes. 
