Back to Community
deleted post

deleted post

5 responses

If they remove the backtester along with paper trading, I’ll blame you Grant! ;)

Ha. So I guess you would say that even minute bars are no longer needed. Back to old Quantopian days with EOD data.

Basically, I'm saying that the ideal tool for generating EOD signals
is not the backtester in its current form.

This is what I suggested earlier when @Thomas introduced new set of rules to get allocation in 2019. But as it would take additional engineering resources to come with another tool, @Thomas defended the use of current backtester for alpha factors (via EOD holdings) which I guess is a fair point given the amount of effort Q has put in to create this backtester.

It would be a shame if they scrapped a tool that it turns out is actually very useful towards their future ends. While their current assessment methodology is focused on EOD alpha decay, I think they need to be careful not to be too short-sighted here. Who knows -- in a year their thinking may evolve once again. They may offer other financial products/funds or improve the current one. And while they're currently dealing with the issue of liquidity by simply giving themselves a 5-day window to leg in and out of positions, perhaps in the future they can resolve the same issue via better trade execution instead. Lets not forget trading takes place during every minute between 9:30-4:00pm. Trading EOD w/a 1-5-day lag is leaving a lot on the table, especially within the context of their new signal combination paradigm.

Also, my argument for keeping the paper trade feature holds for the intraday IDE backtester as well -- it makes the Quantopian platform sticky. I come here to use it. Because it's my main backtesting tool, it's led me to learn the ins-and-outs of the Quantopian platform. I go through the hassle of setting up self-serve datafeeds so that I can access that data from within the backtesting platform. I've discovered contest- and fund- quality alpha in those datasets. Were I backtesting on my own machine, I wouldn't have had to setup CSV feeds to import my data into Quantopian, nor would I have access to Quantopian's QTU or optimizer on my local machine, nor would I be incentivized to learn how to use Quantopian as well as I have.

Having all the daily and minutely data is immensely important to the research and backtesting ideas. You have to have a superset of capabilities for good ideas to come out and lead to final algorithms. Also, the community site acts as a good education and research platform. Just whittling
it down to a contest-only platform will be a very wrong direction to go and would surely lead to a major drop in use and ultimate contributions.