Continuous Futures Support

Awesome! Can not wait to use it! Thank You!!

hello, can someone please provide a simple sma crossover example using continuous contracts.

Can I just pass the continuous contract as shown below in onData()?

self.slow_sma = self.SMA(self._continuousContract.Symbol, 50, Resolution.Hour)
self.fast_sma = self.SMA(self._continuousContract.Symbol, 200, Resolution.Hour)

What a great update Martin Molinero! Fantastic work, team.

I'm new to futures.  Can anyone share tips on what to choose for price scaling, contract mapping and month selection?

Would love any resources  articles / tutorials /  videos / etc.

Thanks in advance.

Hey!
Just sharing a continuous future backtest using some SMAs.

oh i see ! got it ! thanks!! I was confused at first and now I know. 

cheers!

Can you provide an example of the same backrest using SCF data?

I ll try again..

There is a code on git-hub, which I mentioned in my first post, using SCF continuous future data. When I back-test it, it gives no trade, while it should , since it is a simple buy and hold strategy. I think Martin is the last editor of that code, so I post the question here, also because I would like to compare the results of different test, or in research just the time series.Testing data is of fundamental importance in any back-testing, especially when using continuous future.

Some kind of reply would be appreciated. I have tested other continuous future data in the past, and found it very important to do it before doing any trading, For example SCF and Tradestation can differ substantially in some cases. I can do some  comparison outside quantconnect , but first I would like to have a definitive answer about SCF data use on quantconnect,. In the past I noticed SCF data format changed but quantconnect did not update the import methods, and my post regarding it was removed, and I need to know if anything has changed since then.

Thanks

Wish I could help. Perhaps try asking in the slack or discord.

Hey Johnny Cash !

There is an issue reported for this, Quandl was bought by Nasdaq and their API changed and broke the Lean custom data definition. We will try to address this ASAP, you can subscribe for notifications on the issue. 

Hi Johnny Cash 

Our continuous futures support is different from the SCF Quandl Dataset. We have generated our own continuous futures mappings and compared them to third-party sources and are able to make that available for free, built into the QuantConnect Cloud. For discussion about SCF and Quandl please spin up a new thread and we can try and help there. 

Best

Jared

Hey Martin Molinero 

Thanks for having added Continuous Futures: that is a very useful way to avoid spending too much time managing the individual contracts.

However, when trying to use the Continuous futures in a Notebook, nothing is returned (it seems to be also the case for individual futures, while it worked before).

Could you let us know how we can ask for a history of prices for Continous Futures data from now on?

For instance, the following example only show an empty Dataframe in a Notebook (for “continuousContract.Symbol” and “continuousContract.Symbol.ID.Symbol”):

qb_future = QuantBook()

continuousContract = qb_future.AddFuture(Futures.Indices.SP500EMini, dataNormalizationMode = DataNormalizationMode.BackwardsRatio, dataMappingMode = DataMappingMode.OpenInterest, contractDepthOffset = 0)
history = qb_future.History([continuousContract.Symbol], 50, Resolution.Daily)
print(history)
 

Actually, I also tried to use the qb_future.GetFutureHistory (which used to work in Notebooks before the Continuous Futures implementation), but using the standard method described in the docs…

the DataFrames keep being empty (using either continuousContract.Symbol.ID.Symbol or continuousContract.Symbol):

start_time = datetime(2021, 11, 1, 9, 30)
end_time = datetime(2021, 11, 23, 16, 0)
future_history = qb_future.GetFutureHistory(continuousContract.Symbol.ID.Symbol, start_time, end_time)
historyall = future_history.GetAllData()
print(historyall)

Do I miss something obvious? Or the Continuous Futures did break the previous way to request historical data in a Notebook?

Happy Thanksgiving to you and to the whole QC Community!

I set the date to make it work. I've made a Github issue to make it set default dates (or throw with an error message) when not provided.

Thanks Jared Broad! 

I am now able to confirm that:

• Your example works also when adding the Continuous Futures calculations (dataNormalizationMode = DataNormalizationMode.BackwardsRatio, dataMappingMode = DataMappingMode.OpenInterest, contractDepthOffset = 0)… for instance
• It also works with the Second resolution in addition to the Minute resolution, which is great.

The Github you opened (Not returning silently an empty array in research Notebooks) would be very worthwhile!

Yesterday, while trying to find a way to correctly ask for data using futures, I changed the resolution to Daily (to reduce the quantity of data - so I thought!), and I should have been more careful: that was probably the issue for the History call, in addition to the lack of “SetStartDate”, as the Daily data are not available for futures!

Similar to the Github you just made, I also think that returning an error message if no data can be available in the research Notebook (Hour / Daily resolution are still not available for any future) would help avoiding such a stupid error!

At last, I tried to use the previously advised way of handling history calls (the solution I used until last week, which used to work for each individual future), as explained in the docs here: 

Similar to options, future historical data can't be accessed using qb.History. Instead, we need to use qb.GetFutureHistory, which lets us access historical futures data."

Good news is that thanks to your great work to provide Continuous Futures, the standard (and simpler) History calls works for futures from now on (so the Docs can be amended).

However, I cannot manage to make it work in the previous way, using GetFutureHistory and GetAllData.

Even by providing a SetStartDate

Maybe because we cannot add a 4th argument for GetFutureHistory to ask for a precise resolution in the history call, while it was Resolution.Minute by default previously? 

This code used to work (showing data for the different futures, order by their expiration dates) until last week, and adding qb_future.SetStartDate(2021, 11, 1) do not help:

qb_future = QuantBook()
es = qb_future.AddFuture(Futures.Indices.SP500EMini, Resolution.Minute)
start_time = datetime(2021, 11, 1, 9, 30)
end_time = datetime(2021, 11, 9, 16, 0)
future_history = qb_future.GetFutureHistory(es.Symbol, start_time, end_time)
historyall = future_history.GetAllData()
historyall

Maybe the GetFutureHistory() will be deprecated in the future?

If yes, will there be a way to still use individual future contracts in the research Notebook, as before?

Nothing urgent here, I will use the Continuous Futures instead for my current research, using my new more powerful R2-8 node!

Happy Thanksgiving again!

Hi All! 

We tweaked the implementation this evening to let you reference the underlying contracts to trade them in live trading – #first-in-the-world.

self._continuousContract = self.AddFuture(Futures.Indices.SP500EMini,....

When you add the contract you're given the reference to the continuous-contract-security, which has unrealistic prices in many of the adjustment methods. You can't trade this - but the underlying contract is available with a new Mapped property;

self._currentContract = self.Securities[self._continuousContract.Mapped]
self.Buy(self._currentContract.Symbol, 1)

This way you can develop identical algorithms for backtesting and live trading without tweaking the code much or at all.

Please note some continuous contracts will have discontinuities in live trading! This is due to how the contracts are mapped and not something we have control over. If you're using them for technical indicators you should ideally 

 - Bind to the symbol changed event
 - Reset your indicators
 - Re-seed your indicators from a history call.

The history calls should be smooth.

Happy Trading!

Hi Jared!

Thanks for the great work and the resources! Is there any way you could give us an example of what the code could look like resetting the indicators given the change in contract? I've found this template but I'm struggling to implement it with my code for continuous futures. It would be great to see an example, perhaps using the crossovers example posted in the github above.

Cheers!

Hi,

Great work!

Can you please post a link to the github source code of the continuous future calculation implementation? I tried to look at it here , but i cannot see where is splicing performed.

Thanks

Hi Johnny Cash,
We recommend going through the different files changed after integrating this issue to find what you are looking for. Refer to the following link.
Best,
Varad Kabade

Hi,

I have created a simple notebook to  check carefully the attributes of a Future object: 

qb = QuantBook()
import pandas as pd

future = qb.AddFuture(Futures.Indices.SP500EMini, Resolution.Minute,
                dataNormalizationMode = DataNormalizationMode.BackwardsRatio,
                dataMappingMode = DataMappingMode.LastTradingDay,
                contractDepthOffset = 0)


fp=future.SymbolProperties
fp.ContractMultiplier

future.Leverage

obj=future
dict([attr, getattr(obj, attr)] for attr in dir(obj) if not attr.startswith('_'))


obj=future.MarginModel
dict([attr, getattr(obj, attr)] for attr in dir(obj) if not attr.startswith('_'))

future.SetLeverage(10)
future.Leverage

to my surprise the margin model does not have any margins set up. The only other attribute which seem to be relevant is the Leverage, which is set to 1 (strange..it should be much higher for futures), and at least in the notebook, I cannot modify it. I have also checked how contract multipliers are handled, and they are ok. One way to backtest future would be to use some indicative leverage (10 or 20 lets say) and ignore the contract specific margins. In any case the contract specific margins seem not to be available, and leverage cannot be changed from 1.

Are future strategies being properly  margined? If yes, where is the margin data?

I have also checked several strategy library future algorithms using custom data and they do not set up the contractMultiplier attribute, so I doubt the results are correct. At least the native  future QC data has the multiplier correctly set up.

PS

How can notebooks be share directly instead of copying the code as done above? Below I can only see the back-test sharing option.

Hi all!

Big Chai 

-What would be the rolling method when the current contract which we have bought expires? Does buying the _currentContact roll the futures automatically as they expire?

The positions are currently not being rolled automatically, you will have to handle this manually. Please take a look at this basic template algorithm it's rolling the position.

Ariel Nechemia 

-Is there any way you could give us an example of what the code could look like resetting the indicators given the change in contract?

Please take a look at this basic template algorithm it's rolling the position, resetting the indicators should be done in a similar pattern

Johnny Cash 

-Can you please post a link to the github source code of the continuous future calculation implementation? 

The implementation that creates the mappings and factors is closed source at the moment sorry, the implementation that applies the factors and mappings is part of Lean itself, can see this with more detail in the linked PR.

-Are future strategies being properly  margined? If yes, where is the margin data?

Future margins are in the data folder, see for example cme. They require at least one data point to be able to determine which date is it see implementation that's why you are not seeing it being loaded on research.

-I have also checked several strategy library future algorithms using custom data and they do not set up the contractMultiplier attribute, so I doubt the results are correct.

Contract multiplier is driven by the symbol properties data base, see spdb. Custom data will need specifying the symbol properties manually with the desired settings.