0
$\begingroup$

When I construct continuous futures data (Wheat futures for example), I get different results than barchart or tradingview. Examples below. The 1st image is my adjusted continuous data and the 2nd image is Tradingview's continuous futures chart.

My adjusted continuous data

Tradingview's continuous futures chart

The method that I use is the following:

Method that I use

When, however, I do not adjust the data it yields the following chart which now looks more like Tradingview's chart:

Unadjusted futures data

The thing is that I see a lot of unadjusted charts (I presume) thrown around from different analysts who use barchart or tradingview.

The problem comes when backtesting seasonality patterns for example. As in my adjusted data it seems that the market is trending down until mid 2019, but when looking at tradingview's it trends slightly up.

Or is there something wrong in the approach that I am using (like the back-adjustment method)?

$\endgroup$
1
  • $\begingroup$ AFAIK there is no roll adjustment on TradingView. "continuous" does not necessarily mean "adjusted". $\endgroup$
    – user42108
    Sep 17, 2021 at 13:21

2 Answers 2

2
$\begingroup$

There are many different approaches you can use to back adjust the price series over multiple expiries. The most common and useful (IMO) is the one similar to the one you describe. Most adjusted series do this too (eg, Bloomberg's "ED1 Comdty", etc).

However bear in mind that with this approach, each expiry adds a new value into the cumulative adjustment that you are making to the price. In other words, the further you go back, the more adjustments are being used to calculate the adjusted price. This means that the difference between the price and adjusted price can diverge significantly as you get beyond say 5 expiry adjustments. Therefore, the exact date that you use for each adjustment point matters a lot.

You'll need to create a strategy for how to pick the date, and may in some cases may have to manually tweak the dates, so as not to capture any non-standard effects of the roll market at that given point in time. Ie, if the roll market was being particularly squeezed in that day, you might not consider that indicative of the long term market process.

I worked at a place where it took one researcher an entire year to fully construct and research adjusted series for ~120 futures contracts (ED, ES etc) going back in each case 10-20 years. It's not trivial to get right. If you have a consistent source of adjusted series from a third party that you can rely on, maybe use that rather than figure it out on your own. Good luck!

$\endgroup$
1
$\begingroup$

User 42108 summed it best by saying at the time that TradingView did not offer any back adjustment option on their continuous charts. So in other words, they were just stringing together contracts and I'm not even sure what their roll method was (i.e. could be volume, or open interest, or volume and open interest, or days to expiry). To my knowledge, they didn't offer the user an option on the roll method.

Recently, TradingView introduced a back adjustment option on their 1! and 2! continuous contract time series. I believe they're using the difference or Panama adjustment method, but I didn't go to lengths to ascertain for sure and I didn't see them specify in the documentation.

The difference or Panama adjustment you're using where you add/subtract the new-old basis difference is typically the standard found on pedestrian platforms. Better platforms offer the option to select difference, ratio/proportionality and rollover/perpetual.

Rollover/perpetual is similar to how physical crude oil traders hedge around a bill of lading date so that they come out over a five day period with a hedge in futures or swaps or options in proper quantity corresponding to their cargo and at the average price they must use to actually pay for the oil not otherwise known until after the five days are complete. This method is generally regarded as the best for backtesting purposes.

The problem with the difference/Panama method is that it introduces trend bias to the calculated series and that phenom gets worse the further back you go. Prices can go negative.

The ratio/proportionality method takes the ratio between the new and old contracts and multiplies that quotient by the expiring contract price. It's an improvement over the difference method in that it preserves percentage changes for modeling returns. But point changes get thrown off. And if a strategy needs to reference levels, there's a problem.

The rollover/perpetual method determines weights from 0 to 1 to apply to the new versus old contract on each over five days such that pricing for the old contract is slowly brought to 100% by day 5. So on day 1, the price would by 80% new, 20% old. Day 2, 60% new, 40% old. Day 3, 40% new, 60% old. Day 4, 20% new, 80% old. Day 5, 100% old. This method smooths out the basis difference between old and new contracts during the rollover period and preserves more of the actual historical price levels.

No matter which method to back adjust you choose, you must determine the roll date, and that's usually a function of using one of the roll methods I described above. Typically, system traders that use the rollover/perpetual method trade their rolls the same way, slowly rolling out of the old into the new over couple days. And they've chosen a specific number of days prior to expiry to begin that roll.

You can get a feel for what the typical roll date in terms of days before expiry is by market by just observing when volume or open interest or both volume and interest in the new contract overtake the old one. And then use that number of days before expiry to be the third day or the fifth day in the five day perpetual adjustment process.

Make sure you know the First Notice Date as well and avoid holding longs, especially, past then. And don't get cute like USO did in Nymex WTI in April'20 and wait until the last minute to roll your heavy long hoping for a better price.

Just checked and surprisingly, Tradestation uses the Panama method and rolls on the typical day when open interest in the new overtakes the old. They don't specify the lookback used to have computed that. If you're using IB TWS, their continuous series is adjusted using the ratio method and rolls apparently at expiry. Check with @Chartress or anyone you know with a BBG for what they do - my guess is high optionality for the user, a non-pedestrian feature.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.