Quant Programming Contest

[alain]

Check this quant competition for students:

http://www.quantcup.org/

I read Max Dama's blog frequently. I like what he writes.

 

[Tsotne]

Check this quant competition for students:

http://www.quantcup.org/

I read Max Dama's blog frequently. I like what he writes.

Interesting. How often does this competition takes place?

 

[alain]

I don't know. it seems this is the first time.

 

[Tsotne]

deadline:
8.15.2011

Starts in August?! Registration 5 months earlier.

 

[Andy Nguyen]

11. INTELLECTUAL PROPERTY RIGHTS: As between Max Dama and the Participant, the Participant transfers ownership of all intellectual and industrial property rights in and to the Entry that Participant had before submission, including the right to distribute the application commercially, to Max Dama. As a condition of entry, Participant grants Max Dama a perpetual, irrevocable, worldwide, royalty-free, and exclusive license to use, reproduce, publicly perform, publicly display and create a derivative work from, any Entry that Participant submits to this Challenge. Participant specifically agrees that Max Dama shall have the right to use, reproduce, publicly perform, and publicly display the Entry in connection with advertising and promotion.

 

[maxdama]

Tsotne, this is the first time running the contest. Depending on how this one goes there could be more.
Andy, I was going to ask if I could post this on QuantNet. The plan is to open source the top submissions.

 

[DominiConnor]

As well as the real dangers in singing such a contract, as someone who has more than once had to deal with intellectual property issues I will share that there is a major defect in it, possibly another which renders the agreement not to be a contract at all.

Personally I don't think it unreasonable that the winning code should be publicly displayed, in fact the real prize for winning such a contests is to be able to put on your resume "I won a big quant programming competition".

But...
You only need a licence for that, albeit one that grants Max Dama the right to spread it around a lot.

If you sell IP to someone, they have the right to stop you using it yourself, if you licence it to them, then they can do stuff and you can use it at some employer without fear of it biting you.

There have been any number of cases recently, both publicly known and privately dealt, where IP in trading has cut up rough.

If you ever believe anything I ever write, believe me when I say you only leave yourself open to an IP issue when you are comfortable with a potentially huge personal downside.

 

[tobias elbert]

I think this is a great competition for the intended target group. Personally, I dont see IP as being a big issue here, as it's quite clear as to what the intended benefits are for the person/group hosting the competition. Whilst order book modelling is definitely an interesting and important research question - the competition problem as posed is fairly "trivial" from a mathematical/ quantitative point of view.

I admit to just having glanced over it briefly, but it seems to assume perfect information as the exchanges would see it and hence who does what is known (to the exchange matching engine). As such, the hosts are (presumably) after some clever algos/ data structures that cut down on matching latency that could possibly improve upon their own internal orderbook model - which does not act on perfect information and where one has to make assumptions (ie queue position upon placement/ and any subsequent cancellations, market impact, hidden quantities,...). The real IP would be in the latter, I reckon.

 

[maxdama]

I am willing to change that section if anyone can suggest an appropriate clause that says something to the effect, "the entry will be open-sourced under the BSD license and may also be used for promotional/marketing materials."

Tobias, this is correct: "hosts are (presumably) after some clever algos/ data structures that cut down on matching latency."

 

[hardlianotion]

Just looked at the competition rubric. Why is this restricted to students at "top universities"? Should gifted folk from lesser institutions team up with someone more respectable?

 

[maxdama]

hardlianotion, It is not restricted. Your home university has no weight in the scoring criteria.

 

[maxdama]

In response to your advice, the section has been changed to:
11. INTELLECTUAL PROPERTY RIGHTS: As between Max Dama and the Participant, the Participant retains ownership of all intellectual and industrial property rights in and to the Entry that Participant had before submission, including the right to distribute the application commercially. As a condition of entry, Participant grants Max Dama a perpetual, irrevocable, worldwide, royalty-free, and non-exclusive license to use, reproduce, publicly perform, publicly display and create a derivative work from, any Entry that Participant submits to this Challenge. Participant specifically agrees that Max Dama shall have the right to use, reproduce, publicly perform, and publicly display the Entry in connection with advertising and promotion and to release the source code of the Entry under the BSD license.

 

[jimkliew]

Alain, great stuff..thanks for posting it! Hi Max, tell Mark G I said hello.

 

[DominiConnor]

I agree about 'top universities' one of my interns doesn't come from one, and would beat to death most entrants in this competition, and of course I could help him.

I'm so glad Max is smart enough to respond to feedback, and BSD is as good a licence as any, and I withdraw any objections that I voiced earlier.

Tobias is right that this is a highly simplified problem, and I think that's inevitable for an open competition run over the web.
I suspect that none of the code will be fault tolerant and am certain that the huge swathe of monitoring and compliance code a real engine needs won't be there either.
Should it reject trades that are obviously wrong, and how do you define 'obviously' ?

So let that stand.

A version 2.0 of this competition would attract some good number of entries, and pick the ten best.

Then invite them to a physical location, and add rules that they must implement within fixed human time and have them execute rapidly in machine time.
I'd define a specific hardware environment so that they could use threads, and make rational choices about certain optimisations.

That would weed out those who had 'help' , and since multi-threading a matching engine is non-trivial, the winners would look cool, which is ultimately their goal in all of this.

Also there are picky details like the data set to be tested under.
Even though I won't enter, some of my solutions would vary in speed quite considerably depending upon both the size of data and it's actual values (if you've done the CQF, you've probably worked out my default solution from that).

 

[Andy Nguyen]

Nobody else is doing this contest? It's been a month since Alain posted this and I see little (if any) change in the leaderboard.

 

[ThisGuy]

Is it just me or is the leaderboard not displaying scores? Is it working?

 

[Pranava]

Hey Andy, I am working with Max to run the contest. We haven't recieved any submissions as of yet, which is why the leader-board has not been updated. Submissions need to be sent to quantcupsubmissions@gmail.com, after which they will be graded. Thanks.

 

[ThisGuy]

I see the leaderboard has some entries now! It might be interesting to have an entry on the leaderboard for the example implementation that's given, to give a kind of baseline score to compare to.