Back to Community
is this portfolio of any value?


I could not test from earlier than June 2014 because this expression returns NULL prior to june 2014.


Is the selection criterion dependent on high the sharpe ratio is?
If I had to improve this algorithm, what criterion should I aim for?


Clone Algorithm
Backtest from to with initial capital
Total Returns
Max Drawdown
Benchmark Returns
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Backtest ID: 598de084cda64150ec0dd56f
There was a runtime error.
8 responses

Nice positive alpha, low beta. Sharpe over 1.0. Basic stats look great.

Go under research -> Notebooks
Create a new new notebook
Type in bt = get_backtest('598de084cda64150ec0dd56f') and hit control-enter.
Then type in bt.create_full_tear_sheet() and hit control-enter again.

That will show you all sorts of analysis on how your algorithm performed.

You can see in the "fama-french factors" it's highly correlated to "momentum" stocks, and highly inverse-correlated to "high growth" stocks. If you want to get an allocation you'd have to create an algorithm without those kinds of correlations.

Looking at drawdown periods you can see the main area where the algorithm doesn't perform so well. It spends most of its time in drawdown periods, meaning that if you started up the algorithm at any of those peaks, you'd have to wait a year or a year and a half before you even broke even again. Most of the time you'd be under water. Looking at the underwater plot you can see deep in losses you'd be in 2016 and 2017 if you'd started the algo up at the end of 2015. That tells us that there's a high likelihood that if we start up the algo today, we're likely to head straight underwater and stay underwater for a long time.

The great thing though is that your algorithm's drawdowns aren't correlated to the stock market's drawdowns.

You can have three algorithms in the contest at any time. If I were you I'd just go ahead and get a horse in the game, so you can see how it performs out-of-sample compared to other algorithms, and then as you improve the algorithms or create betters ones swap out the losers for your more promising entries.

Keep in mind, for the contest it's based on $10m initial capital instead of $1m. That can have a big effect on how well an algorithm performs.

did you already try out the algorithm in a notebook? I mentioned that I am unable to get fundamental data prior to june 2014 for
morningstar.valuation_ratios.peg_ratio.latest , which limits my analysis across various time periods. so, how did you complete the test?
what exactly is f"ama french factors" and why is it relevant?
I expect this also to outperform SPY over the long term but not over the short term. so, I am ok with your comments about being under water for long -but if you think the gains are not significant or risks are higher, then let me know. My algorithm invests money on start of the week. so, will the contest defer allocations to monday? imagine what might happen if we hit manic monday like 1929?


which I clock notebook, it shows a spinning wheel saying "Loading Notebooks.." but doesn't do anything. Any other way i can test this?


I'm just learned this stuff myself and sharing my new knowledge. I'm sure somebody else can explain everything better than I can.

There's the notebook with your tearsheet. Just click view notebook or clone notebook.

Loading notebook preview...
Notebook previews are currently unavailable.

Your algorithm makes a huge bet (66%) that QQQ is going to continue to outperform. This has been the case up until now. So this is an example of look-ahead bias. From 2017 you can create a backtest that performs well based on what you know worked so far. But you don't know what will happen in the future. Many people are skeptical that tech will continue to be the high growth leader, and lots of people are saying there needs to be a correction in the tech sector. Who knows ,but algorithms that exploit lookahead bias and overfit bias don't tend to perform well going forward, because the backtest was skewed based on knowledge you wouldn't necessarily have had on the start date.

when I clicked "Clone Notebook:, it shows me the same "Loading Notebooks..", but does not go beyond that as if it is stuck in a loop.
Is it something to do with safari browser that I use on my mac?
My algo bets 60% on QQQ and some other shorts to balance it out. If the beta is 0, does it not indicate a risk neutral algorithm?
Is anyone else able to get valid data from datasets on jan 1 2014 for

I got a message from contest saying my leverage is > 1.1, but I am using optimise API to handle the leverage.
Is the API not being used correctly or any there safeguards I need in place to keep leverage down?