Quant Working Environment?

semut · Post by **semut** » Mon Jun 25, 2018 2:35 am

We're building a quant prop shop but haven't decided about the structure of the company.

1. Handful number of small teams working with smaller set of strategies and allocation, huge firewall for any specific strategy implementation between team, but more transparent bonus calculations and more flexible arrangements for IP ownership. Literally, you eat what your team kills.

2. Or, just a company where no team or individual has direct access to complete view of the model. Bonus will closely dependent on the company wide PnL.

As a quant, which environment do you guys prefer to have? We tend to lean toward the first one, but we'd welcome any feedback from any perspective.

goldorak · Post by **goldorak** » Mon Jun 25, 2018 3:20 am

Neither. Totally open environment with sharing of all code, information, ideas and resources is the only way to get good work done.

semut · Post by **semut** » Mon Jun 25, 2018 3:29 am

Well, that's the dream, isn't it?

Suppose you're one of the principals or managing partner, will you still impose strict NDA?

goldorak · Post by **goldorak** » Mon Jun 25, 2018 8:12 am

I am a managing partner.

goldorak · Post by **goldorak** » Mon Jun 25, 2018 8:12 am

and my business is science, not managing mercenaries.

deeds · Post by **deeds** » Mon Jun 25, 2018 12:55 pm

@semut

Lopez de Prado cites putting quant teams in silos as the first of seven reasons ML/Quant funds fail (does not appear to be specific to ML)

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3031282&download=yes

(no endorsement of idea, necessarily, details and implementation always matter, etc, but thought this should be part of dialogue)

semut · Post by **semut** » Mon Jun 25, 2018 3:38 pm

Sorry, I didn't know that you're a managing partner, but here's the thing that concerns me the most. We're starting our business in an emerging market where we somehow could get our hands on a huge dataset that no other firm that we know of has. So the base models that we have before reaching the talent pool are already producing great performance. Not a secret sauce, but still a secret that we, as a startup, have to protect while it's still highly profitable.

By your definition of "total openness", does it mean that you think it will be better off (even in the situation described above) to just share everything even though there is a risk of new recruits to leak of some parts of our models to competing firms? I'm curious, are you okay if an employee somehow share the model that he didn't even write with other firms?

To be fair, I'm totally not against having an open environment, but without any information barrier, I mean like, none whatsoever... it's a little bit uncomfortable, at least for me. But I'm still keeping my mind opens to other perspectives.

semut · Post by **semut** » Mon Jun 25, 2018 4:04 pm

@deeds

Interesting paper! It's 10 reason now Scared

Is your working environment totally open like the one described by @goldorak ?

Or if not, do you agree with the idea of full openness?

deeds · Post by **deeds** » Mon Jun 25, 2018 4:49 pm

@Semut,

must confess not in the 'scientific' realm of money management, but do quant pricing and work with exotic financial models

our folks are allowed to exchange ideas, but all so busy that that is rare

your issue seems less one of proprietary models, more one of proprietary data

Would suggest that, in the latter, if sophisticated models are used, they usually require the key modeler to be with the model to be of use...not so with datasets...conceivably can be useful to anyone who walks off with them

semut · Post by **semut** » Mon Jun 25, 2018 5:30 pm

@deeds

First of all, we believe that exchanging basic ideas, scientific knowledge/understandings, and/or technical solutions should be encouraged as it will likely to be beneficial to any kind of tech/science-based ventures, including ours. So yeah, we are not against total "openness" in that respect.

For the model vs data part, it's a bit tautological since by knowing the model, you can reverse engineer the features, then it will eventually lead to the "aha" moment where you know what to look for.

To be clearer, would you be enjoying more working in a larger part of small interdepartmental team or smaller part of large division?