Share New Research 
Does QC have history data for OTC stocks? The few I've looked at don't return any history in Research or Backtesting.
QuantConnect Community Discussions
QUANTCONNECT COMMUNITY
Upvote
QuantConnectâ„¢ 2025. All Rights Reserved
Derek Melchin
Hi Brian,
We don't currently have OTC data.
Best,
Derek Melchin
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Alfred Aita
You imported but did not use MinMaxScaler. The results for this example would probably be the same.
Is this run on a GPU ? Just wondering
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Liu Jin
Hi Sherry Yang , I've tried to run your algo on a longer time period (starting from 2015) but after about a year of backtest I get the error below. I would have thought that the Train() method ( https://github.com/QuantConnect/Lean/blob/master/Algorithm.Python/TrainingExampleAlgorithm.py ) would have solved this? What do you think? Thank you!
System.TimeoutException: Algorithm took longer than 10 minutes on a single time loop. CurrentTimeStepElapsed: 11.0 minutes
at QuantConnect.Isolator.MonitorTask (System.Threading.Tasks.Task task, System.TimeSpan timeSpan, System.Func1[TRest̲]with∈Cus→mLimits,System.Int64memoryCap,System.Int32s≤epIntervalMillis)[0x002c4]∈<eefa6b447f3a4e0eb61632434c9719a7>:0atQuantCo∩ect.Isola→r.ExecuteWithTimeLimit(System.TimeSpantimeSpan,System.Func1[TResult] withinCustomLimits, System.Action codeBlock, System.Int64 memoryCap, System.Int32 sleepIntervalMillis, QuantConnect.Util.WorkerThread workerThread) [0x00092] in <eefa6b447f3a4e0eb61632434c9719a7>:0
at QuantConnect.Lean.Engine.Engine.Run (QuantConnect.Packets.AlgorithmNodePacket job, QuantConnect.Lean.Engine.AlgorithmManager manager, System.String assemblyPath) [0x0099d] in <b0a99c0f99784925a4d272e02c8243cb>:0
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Stanley Yang
great post!
I have tried to use other deep learning technique or LSTM in QC , but the problem is when you back testing longer than 3 to 5 years. you will get the following error message:(runtime error)
System.TimeoutException: Algorithm took longer than 10 minutes on a single time loop.
so I cannot file the algo in your alpha market, because your team ask for longer than 5 years backtesting record.
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Alexandre Catarino
Hi!
Alfred Aita : MinMaxScaler is used in MyLSTM class' constructor. This is not run in GPU.
Liu Jin and Stanley Yang , did you try to play with the algorithm parameters? Maybe the neural networks model is taking too long to learn with 2015 conditions.
Please check out the docs, under Machine Learning, for more details on training limits.
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Jack Simonson
Hey all,
Per a request from someone is support, we've created an example that uses a couple of technical indicators as features in an LSTM model. You can view the model in the research notebook attached below. Check it out, play with the inputs, and then find a way to put it into production!
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Hyungjun Lim
Because the prediction is made on the raw value of the next day (as opposed to change amount) this tpye of predictoin when plotted, looks usually good.
But in fact, unless the daily price jump is of noticeable size, even prodicting no motion at all would still make the graph look decent, only that the graph is shifted by a day.
In order to correctly and faily display the prediction of time series, the plot should be done w.r.t the delta value, not the raw value.
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Pangyuteng
This is a nice starting code to get our hands wets on neural nets.
Thank you for sharing Sherry. Very exciting!
I believe neural nets may be suitable for predicting trend, regime change and probably noise reduction. What I'm sharing below is doing none of the above! It is predicting next day stock return (!?) by feeding the LSTM with past Volatility Risk Premium (VRP, implied volatillity - historical volatility) and ratio of VIX and VXV (indication of Contago or Backwardation) [Tony Cooper, Easy Volatility Investing]. Custom data retreival is a copy pasta from Alex Muci's post "A simple VIX Strategy". I pretty much hacked through and crippled Sherry's code as this is just an excercise for me to get familiar with Quantconnect's framework and apis. The timeframe used here is cherry picked for a jolly PSR, training was done once as oppose to monthly, be wary of look ahead bias and other bugs. Expect timeout to occur if you enable monthly training and extend the backtesting timeframe.
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Jared Broad
Nice work Ted! Worthy of its own discussion-thread if you were inclined.
That cherry-picking isn't so bad - we can call it tuned to "recent market behaviors".
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Dirk bothof
Hi guys, great resources!
This really inspired me to build some neural networks over the past weekend, however I feel that the current hardware underpinning QuantConnect is not suitable for this type of ML, a small net with relatively little input data (for a neural net) already has > 1 hour training time. To make something robust and value adding you need way more hardware.
It would be pretty great if I could add my own resources and pool that with the QuantConnect to speed things up or get access to a bigger pool of CPU's or a GPU at some extra fee. As is, I don't think it is feasible to do the propper research required, let alone trade neural net.
Does anybody have a different experience?
Best,
Dirk
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Jared Broad
HI Dirk; professional subscriptions can harness more CPUs, and have a longer training period allocations for backtesting.
Live trading should be fine recalling a stored model, or retaining the previous models on a schedule as shown in the documentation above. It is happening in real-time which is 10000x slower than backtesting which should give it plenty of time to train the model to trade?
Adding a GPU backtesting option is a cool idea though and relatively easy for us to do. If there's demand we can test that out with the community. The GPU cards are expensive though it would need to be about $100/mo subscription to cover the hardware cost.
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Colton Surdyk
Couldn't you train the model offline on your own hardware and then load the saved model/weights manually in QC?
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Jared Broad
You can train offline and load them into QC if you have the data - this would be fine for daily sources like Yahoo etc. For an intraday strategy that could be tricky.
We've created a new subscription upgrade "GPU Upgrade" - if we get >5 subscribers we'll add the necessary hardware. It's about $4,000 for the graphics card and there will be about 2-weeks setup time.
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Dirk bothof
Hi Jared,
Really cool to add GPU support this sets QC appart from competing platforms! For now I'm awaiting the L1 data to test my current stratgegies but will in parralel start reading on how to fully utilize the possibility of incorperating deep learning into a strategy!
Best,
Dirk
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Antimarket
I've modified the code to run on the "EURUSD" and a set of other G10 currencies. The result is not satisfactory and it needs further exploration.
Thanks to Sherry Yang and Ted for the basic code implementation and Alexandre Catarino for his help.
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Derek Melchin
Hi Apollos,
This algorithm's initialization process takes a while to complete because model training is computationally-intensive. Since we called the Train method, the algorithm can take up to 30 minutes to train the model before the backtest starts. To speed up the process, consider reducing the size of the neural network. For more information on model training, refer to our documentation.
Best,
Derek Melchin
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Apollos Hill
Hi Derek,
Thank you for the tip. I have been playing around with pytorch since then. Not sure why i didn't get your comment notification.
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Shile Wen
Hi Baba Gi,
1. We may look into that for the future
2. This is one of the challenges of running LEAN locally. As a part of our agreement with our data providers, our data can't leave QuantConnect
3. Training periods are longer for higher level plans. If users are getting cut short, please contact our support at support@quantconnect.com
4. This is a good idea to explore.
One solution might be to train the models using less data but more periodically. Furthermore, if you could provide more details about your models, we might be able to give more suggestions on how to increase efficiency.
Best,
Shile Wen
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
AlMoJo
Hi everyone :)
I saw the backtest made by Pangyuteng and I have to say it looks amazing in terms of return / drawdown ratio.
When I tried to launch it I get this error message and nothing happens after the backtest starts launching.
Does anyone knows how to solve that please?
Thanks a lot
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Vladimir
AlMoJo
Try changing in my_custom_data
url_vxv = "http://cache.quantconnect.com/alternative/cboe/vix3m.csv"url_vix = "http://cache.quantconnect.com/alternative/cboe/vix.csv"
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
Spacetime
AlMoJo ,
The data links for quandl were updated, so you will need to change the quandl data links in my_custom_data.py file
Please see the attached backtest project with the new quandl data links.
Hope that helps.
Quandl New Data Links:
https://cdn.cboe.com/api/global/us_indices/daily_prices/VIX_History.csv
https://cdn.cboe.com/api/global/us_indices/daily_prices/VIX3M_History.csv
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by QuantConnect. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. QuantConnect makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances. All investments involve risk, including loss of principal. You should consult with an investment professional before making any investment decisions.
To unlock posting to the community forums please complete at least 30% of Boot Camp.
You can continue your Boot Camp training progress from the terminal. We hope to see you in the community soon!