Criteria for Allocation

In order to disqualify some of the persistent (read: lucky) algo "writers" (read: optimizers), I think Quantopian should consider and implement the following criteria:

1. The algo writers must pass a financial literacy test.
2. The also writers must pass a Python programming test.
3. The algo writers must have actual real money trading experience, and they must show that they have outperformed the market, even if by a basis point
4. If 1 2 and 3 above are not satisfied, then an outstanding algo performance must be explained in a paper, and it's economic foundation must be well presented. If there is no economic foundation, then the writer must explain why and how he/she can exploit randomness, or why and how he/she can exploit un-economic factors (sun spots, moon phases, whatever).

These are my thoughts after reading comments and questions on this forum.

Question for Quantopian: of your 90K users, how many trade their own money on IB or RB using Quantopian? How many are beating the SPY?

17 responses

I trade using Robinhood, and I'm currently beating $SPY by a small margin using an ETF rotation algorithm. "I trade using Robinhood, and I'm currently beating$SPY by a small margin using an ETF rotation algorithm. "

And your point is? You fit that one criteria, but I bet 99% of the algo writers on Q do not.

You do not need to have a faux HFT algo to beat the market. If you rebalance weekly or monthly, that should be acceptable, but it seems it is not. If you go long inversely correlated assets, you should be considred "hedged" but it seems you are not, on Quantopian.

The question for Quantopian remains: of your 90,000 users, how many trade their own money on IB or RB using Quantopian? How many are beating the SPY? Thanks in advance.

The point is, just like with non-algorithmic trading, some people will be profitable, some people won't. I'd expect that the distribution of profitable to non-profitable algorithms is heavily weighted towards those that are not profitable for all algorithms that have currently been written. I'd expect that the distribution of profitable to non-profitable algorithms that are currently being live-traded looks like a bell curve.

Additionally, just because I'm curious, why is it important to you that algorithm writers be profitable or not? How is this affecting your ability to write an algorithm?

"I'd expect that the distribution of profitable to non-profitable algorithms that are currently being live-traded looks like a bell curve."

Maybe, maybe not. The overwhelming desire to stop/restart/tweak an algo when losing money is something you are not considering. Taking that in account, I would bet that most people who currently live trade have positive returns (those who lost money are no longer in the "live trading" set), and that those who still trade are lagging the market (SPY). Just a bet that Quantopian can issue data and metrics and confirm or disconfirm.

"Additionally, just because I'm curious, why is it important to you that algorithm writers be profitable or not? How is this affecting your ability to write an algorithm? "

It is not important to me PERSONALLY, but it should be important to QUANTOPIAN when allocating funds.

By definition, 99.99% of algo writers on Quantopian will not get an allocation, and most people here in the forums act like they deserve one. They act like naiive engineers and rocket scientists, recluse geniuses who don't know much about finance or Python, but learning either is a trivial task for them, since they are so bloody smart. In fact, what happens is that most of them are people who clone other's code and tinker with it until they find what they think is superior code. If Quantopian wants to prove or disprove this, they can issue one more metric - of the 90K people's algorithms, how many are cloned and tinkered with, and how many are based on brand new code? I speculate that more than 95% of the "algorithms" are clones, altered or unaltered.

So, Quantopian needs to control such behavior and test the candidates for allocation, and the candidates need to prove just how much they know about finance and Python. They need to prove that they are not cloners and copyers. If not Quantopian, then their backers (a real world actual hedge fund) should do it, and should test all of them like they are applying for a real life quant trading job.

Are the requirements I am proposing too strict or realistic for the Quantopian users or for the company itself? The question remains: of your 90,000 users, how many trade their own money on IB or RB using Quantopian? How many are beating the SPY? Thanks in advance.

I do find Grenspan's overwhelming negativity compelling yet in many ways distasteful or perhaps disturbing. It may well be that what he or she says is essentially correct nonetheless perhaps it does not need saying. Perhaps we all simply walk a sea of randomness; perhaps those who do not realize it are happiest. But the fact is walk we must, unless we wish to head for exit before the final curtain call.

On reflection we may feel better when we express positive sentiment rather negative. Perhaps we are all sailing in a ship of fools, perhaps not.

Perhaps Ms Mrs or Mrs Grenspan could cheer him/ herself up a bit by adding something constructive to the discussion? Where are we going wrong, what should we all be doing, how can we turn disaster into triumph?

Anthony,
as a code-cloner and tinkerer, that was hauntingly poetic. I count myself among the time-wasting hobbyists on this website, but one thing I do like in general is the positive atmosphere among the other forum-dwellers.

OK, I am flipping everything negative in the original message into positive, see where that gets us....

In order to encourage a growing number of algo writers to become eligible for a 10Million fund allocation, Quantopian should implement the following criteria:

1. The algo writers do NOT need to pass a financial literacy test - no finance experience is required.
2. The algo writers do NOT need to pass a Python programming test - tweaking code found online is acceptable
3. The algo writers do NOT have actual real money trading experience, and they do NOT need to show that they have outperformed the market, even if by a basis point - no real life trading experience is required
4. They do NOT need to have an economic foundation for the algo they write. The allocation receiving algos could be based on random number generators, random data sets, blind machine learning, sun spots, moon phases etc.

Does that make you feel better, more accepted, and more eligible for a 10 Million allocation?

I think that sums it it pretty nicely, actually. Algorithmic trading for the everyman.

Here ya go!

About Quantopian

Quantopian is a crowd-sourced quantitative investment firm. We inspire talented people from around the world to write investment algorithms.

Quantopian provides capital, data, a research environment, and a development platform to algorithm authors (quants). We offer license agreements for algorithms that fit our investment strategy, and the licensing authors are paid based on their strategy's individual performance. We provide everything a quant needs to create a strategy and profit from it.

Quantopian's community has doubled year-over-year for the last three years and now numbers over 90,000 members. Quantopian’s members include finance professionals, scientists, developers, and students from more than 180 countries from around the world. The members collaborate in our forums and in person at regional meetups, workshops, and QuantCon, Quantopian's flagship annual event.


The question remains: of your 90,000 users, how many trade their own money on IB or RB using Quantopian? How many are beating the SPY? Thanks in advance.

C'mon, people, the guy is just a troll.

He has written 3 algorithms and performed 0 backtests.

I never understood the point of these negative topics. You don't have to spend a dime to use their software. If the people at Quantopian are foolish enough to not do their due diligence and burn their own money then so be it.

Hello Alan,

What the 90,000 users as a whole are doing, what they know, how successful they have been--none matter. Q has a commitment from Point72 of $250M in capital. Q needs a basket of good algos that will scale. My sense is that there are some clever, experienced users here, and it'll get off the ground. I respectively submit that you may be barking up the wrong tree. It kinda goes without saying that the vast majority of 90,000 users won't get an allocation, and that most of them probably don't have a clue (myself included). That said, some small fraction of them may inch up the learning curve enough to make some dough from Q, even though they start from no trading experience whatsoever. It would be interesting, though, if Q released statistics on their users, but I wouldn't hold your breath. "What the 90,000 users as a whole are doing, what they know, how successful they have been--none matter. " All of it matters, or should matter to whoever is giving them money. "Q has a commitment from Point72 of$250M in capital."

That is completely false. In fact, they got a $2+M cash injection (probably for some equity stake, which might be so low that Quantopian is ashamed to express it in terms of valuation), and a loosely worded "commitment" for UP TO 250M, and not for 250M. "that most of them probably don't have a clue" You got that one right. "some small fraction of them may inch up the learning curve enough to make some dough from Q, even though they start from no trading experience whatsoever." .01% will get an allocation, we know that, but even they should be subjected to financial literacy tests, coding tests, criminal background checks, credit checks, etc, to make sure that they know what they are doing, what they are doing is legal, and they are who they say they are. They will probably need to open up their algos, so that the code can be vetted to be sure it is not based on sun spots or voodoo. Well, 0.01%x90,000 = 9. Last I heard, the number is up to 15, with an average of$100K per algo. So, your order-of-magnitude estimate is about right. Let's say they double and the number is 30 algos. If they are actually somewhat unique/independent, then they should have a good thing, no?

They will probably need to open up their algos

My understanding is that this is voluntary, and not requested.

Anyway, you aren't going to get what you're after. I've been on Quantopian since 2012, and there is just stuff they aren't gonna share (which is the wrong approach, in my opinion, if they really want to live up fully to the crowd-sourced moniker). They tend to share information only if it is to their apparent advantage. Human nature.

I think the question is do the costs for supporting the 90,000 users justify a platform to recruit 0.01% of its users as talent? Or maybe the 90,000 are generating profits, too?

Supposedly this is an incubator.

You take the 0.01% that display raw talent and you nurture it. By definition this a high risk high, high reward kind of thing.