Buy/Sell Quantopian Algorithms in an open way

I would be both a potential buyer and potential seller of signals if it were possible.

Your idea has much merit. The problem in tying oneself to a flagship is that any changes are beyond your control. Same problem with Quantconnect of course which I am getting up to speed on to use for auto trading my account.

Are you proposing that buyers/sellers disclose the code?

My suspicion would be that from the regulatory front you would be safer selling code than black boxes. If you sell a black box you are giving investment advice; even arguably managing money.

If you are selling code you can not be accused of breaching the regulations. You are selling mere educational material, not financial advice.

That's the way it works in the UK anyway.

vladimir prelovac

@zenothestoic yes you would be buying/selling an algo with the code.

Viridian Hawk

I would be more interested in being able to license the daily positions generated by a strategy than the actual code. Code is of no use to me since a $30k/year subscription to FactSet would be too much of a drag on returns.

As for regulatory hurdles, these can be overcome. In the US I believe the distinction is that you cannot give personalized investment advice (without the proper credentials), but general advice is fine. Otherwise nobody could show up on CNBC and talk about what they think the market or some stock is about to do (especially since they're always wrong!). There is also a clear distinction between "I recommend you invest in [x]" vs "Today I'm buying [x]" in that the former is investment advice and the latter is protected free speech. I think "My system is buying [x] today" could arguably fall into the latter.

luc prieur

I'd be interested in buying. I would want the code, not the signal. One problem I see is how do you negotiate the price? On the backtest performance alone? As well, from the buyers point of view, how would you know if it is just not over fitted?

One approach would be to sell the algo with back test and live test data, but then, that means the code needs to run for a certain time live. Another way would be to have a selling price with future performance targets, which if not met, the price is discounted. E.g. it is guaranteed that the algo will maintain a 6 month rolling Sharpe ratio of 1.0 or above..

I have little input for the platform on which the transactions would occur. It could be as simple as creating a thread on Q forum where buyers and sellers convene. The transaction would be direct amongst the sellers.

luc prieur

An escrow service could be used, but that would need to be investigated. I.e. the buyer deposit the amount in escrow with seller as beneficiary with a release schedule. If targets are not met, the escrow returns the money. I am sure some escrow service specialize in software transaction with performance criteria.

vladimir prelovac

Ok lets see if there are any sellers around.

Want To Buy:

A Quantopian algo with these requirements:

Using QTradableStocksUS()
Sharpe 1+ on a backtest since 2005, using default commission/spread
Long or short (short more valuable to me)
Trades once a month or less often
3-10 positions held

Ready to pay $1k-$10k. Price depends on quality (higher sharpe is better, lower drawdown is better, short is more valuable than long). Terms to be negotiated. Contact me at [email protected] if interested.

vidal boreal

I would be both buyer and seller. Sometimes I am stuck with no new idea to improve strategies. People's workable ideas may help.

@vladimir
I wonder how you could live trade with the bought algo since I still couldnt find substitute of Quantopian Pipeline.
I only have a live trading algo that specify few stocks, which could not filter large amount of stock datas like Pipeline.

Peter Bakker

Zipline-live2 works with zipline and pipeline. However, live trading doesn't need pipeline often as you cannot have forward-looking bias in live trading. I trade live and where I needed pipeline in the past I get away with other datasets in Quandl and I dont bother putting it in a pipeline: I just query as is. In this situation you cannot backtest with your system, you need to backtest in Qantopoian.

Peter Bakker

As an answer to the original question: I have tried to set up a 2 sided marketplace for Algo's and the main issue is the lack of reliable tradeable algo's (none of the algo's I looked at to buy would perform in paper trading as well as in the backtest, live trading also showed a lot of other .... biases and off course the uncertainty around curve fitting). I'm still trying to create a simpler version of this marketplace where the platform purchases the algo's (if needed, I have a neat collection already) and sells the signals via an app/api integration on the broker. This model seems to resonate more with investing public. The app is easier as you don't need big regulatory involvement. The integration on broker accts runs into regulators' zealousness and arbitrary judgemental stupidity. Happy to continue this discussion with more detail in private as I have learned a lot in the last 2 years. Ping me

and sells the signals

In many jurisdictions you will encounter legal difficulties in selling signals.

I still couldn't find substitute of Quantopian Pipeline.

Surely Quantconnect fits the bill?

Kyle Oates

Dec 31, 2019

+1 on buy side.

Arun

Dec 31, 2019

Alpaca has zipline, pipeline, fundamental data and a brokerage. It's free. You'll have to write the optimizer. You can link your marketplace to them.