Hi Lucas,

Great suggestion, I think we should be able to make that available and it is something we are certainly very interested in understanding ourselves.

I'm not sure what you mean by "It would have been nicer if the winners were not allowed to change algorithm but it seems it is allowed"? We deploy the winning algorithm unchanged against a $100k trading account the next trading day after the contest is concluded each month.

Best regards, Jess

Hi Jessica,

Now that the contest is well along with several months of paper trading and actual trading, is there information that Quantopian can share that would help us developing our own systems? Quatiopian has assured members that it cannot look at individual members source code. Fine. However there is a lot of meta-data that Quantopian has access to that would be very helpful. For example,

For systems with the top 20% returns:

How many systems are trend following, mean reverting or market neutral?
How many systems are mostly fundamental vs technical? Determined by calls to fundamentals API...
What is the breakdown of markets/instruments traded?
What is the frequency of trading? Trades per day & trades per instrument.

Disclosure of the above info would not include specifics of each trading system or need to look at the source code. Indeed, author and name of the systems would not be needed at all.

Quantopian would benefit by decreasing the time to develop ever better trading systems if it would help avoid 'reinventing the wheel' month over month. Confidentiality of source code and member details would always be respected.

Thank you,
Mike

For me the main question is if the back-test and 1 month test is representative. Say if a strategy in 13 months testing had 4% drawdown but on the first month trading a -6% drawdown then it is probably that there is something wrong. specially when the market is behaving kinda "normal"

The main question about a backtesting system is if it is close to real-world trading.

Yeah, even I would like to get access to my own data. There was a 70bp dislocation the other day in live contest money trading which was not replicated in the ongoing paper trading.

Jessica, it would also be very helpful if you could provide access to an archive of the daily leaderboard spreadsheets.

Michael: I will check if there's a way to get at the daily leaderboard archive. It is currently possible to get at monthly archived leaderboards by going to the Winners page and clicking on the red boxes (e.g. "February Contest"). You'll be directed to the month-end leaderboard page where you can download the spreadsheet, or use these direct links:

• February Contest
  
• March Contest
  
• April Contest

Simon: Absolutely agree this needs to be easier, we have a planned API enhancement in the Research platform that will allow you to load live trading results - we will keep you posted on when that becomes available!

Lucas: You can get a very coarse picture of how live trading results are matching backtests with our first two contest winners today (Jeff's winning algo hasn't legged into it's positions yet, so isn't informative at this point).

If you look at Grant's leaderboard backtest results his algo had a backtest maximum drawdown of 7.5% and a paper trading maximum drawdown of 1.4%. Given that his algo has a pretty high Beta to the SPY (we didn't start displaying Beta until the April contest, but it does) I would characterize it's live performance as being within the bounds of expected backtested behavior.

Simon's winning algo in from the March contest had a backtest max drawdown of 2.4%, and a paper trading max drawdown of 0.44%, and he is currently leading our live trading results (Congrats Simon!) with a positive return of over $1100 to date (or +1.11%).

Eyeballing my ongoing paper trading of the same algo that won the contest, it's up +1.43% since the start of the contest (roughly), vs +1.11% of live trading. I think this is directly attributable to a 30-40bp dislocation overnight the other day, the cause of which is unknown, but I have my suspicions.

Hi Jess,

Are both Quantopian and Interactive Brokers required to keep detailed trading records, for compliance, accounting, and tax purposes? I would think that you'd be logging everything, no? And do you record the OHLCV minute bar data feeds for each algo? When are timestamps applied, and referenced to which clocks? Do you store any information as the OHLCV bars are generated, that we can't access in an algo? Etc.

A starting point would be to summarize for us, completely, what data you have (or could be obtained from Nanex or IB). Then we could start to formulate questions, should the data be made available (or if you were to analyze it and present it in summary form).

Grant