Roll config for VIX_mini

[Spacious08]

Hi everyone,

I’ve been analyzing the differences in liquidity between the VIX and VIX_mini contracts. Given my current trading capital, I prefer trading VIX_mini over VIX.

It appears that for VIX_mini, only the front contract has enough risk volume liquidity to be tradable. Because of this, I’m considering modifying Rob’s default roll configuration to trade exclusively on the front contract for VIX_mini.

I understand that it’s possible to adjust an instrument’s roll configuration using the interactive_controls-script.py. Before making any changes, I wanted to ask :

• Has anyone modified the roll configuration specifically for VIX_mini?
• Would trading only the front contract for VIX_mini introduce any risks or inefficiencies compared to the default roll configuration?

If anyone has experience or insights into this, I’d greatly appreciate your feedback. Thanks in advance!

[vishalg]

I have been thinking about it but haven't tried it yet. A few points to bear in mind

1. You will need to update the continuous historical price series as well and not just the roll calendar
2. What will be your carry contract?
3. Check the p&l by trading rules for VIX vs VIX_mini. Any carry rules may not perform as well if you are exclusively trading front contract
4. Front contract, I believe has a much higher volatility and your position will likely be smaller - hence switching to a smaller contract on the front expiry may not give as much benefit as you expect

[Spacious08]

1. You will need to update the continuous historical price series as well and not just the roll calendar
   Yes, this is exactly what the safely_modify_roll_parameters function in Rob's system handles. It recalculates the roll calendar, multiple prices, and adjusted price series based on the new configuration.

2. What will be your carry contract?
   To address this, I plan to test the following configuration to avoid trading the front contract, ensuring that liquidity and volatility remains adequate. This approach will help mitigate the higher volatility of the front contract as you mentioned:

VIX: FGHJKMNQUVXZ,-60,-1,FGHJKMNQUVXZ,19
VIX_mini: FGHJKMNQUVXZ,-30,-1,FGHJKMNQUVXZ,19

[vishalg]

Makes sense. Worth giving it a try.

[vishalg]

@Spacious08 - I have had a brief look at this today. Couple of things standout - VIX mini started trading only in 2020 but the prices in the repo go back to 2005. I am not entirely sure where the data in the repo comes from. Nevertheless, I created VIX_mini from scratch and backfilled the series with new roll strategy and the two data series are quite different - which is to be expected given the sharp carry/roll down in VIX term structure.

[vishalg]

Couple other charts. It may we worth playing around with changing the offset on VIX (which has a longer history). Interested to know if you get any other insights

[Spacious08]

I think the VIX_mini data going back to 2005 might be because Rob used VIX data to backfill the series until 2020, after which the actual VIX_mini data was used.

Thanks for sharing your comparison of VIX, Mini-Old, and Mini-New. It’s interesting to see how the backfilled series differs from the new roll strategy.

On my side, I ran a backtest using VIX_mini with both 30 RollOffsetDays and 60 RollOffsetDays configurations to calculate the multiples and adjusted prices. Technically, I used Barchart to download the historical futures prices for both VIX and VIX_mini. I then merged the datasets to create a complete historical series going back to 2005. Using these prices, I generated the roll calendar and calculated the multiple price series based on the respective contracts and roll calendar.

The results were largely similar between the two configurations, except for a more negative skew and a lower Sharpe ratio when using the 30 RollOffsetDays. This was expected due to the increased volatility from relying on more recent contracts.

Here are the results:

VIX_mini 60 RollOffsetDays :
[[('min', '-32.96'), ('max', '7.718'), ('median', '0.0132'), ('mean', '0.02571'), ('std', '1.197'), ('skew', '-5.504'), ('ann_mean', '6.581'), ('ann_std', '19.16'), ('sharpe', '0.3436'), ('sortino', '0.3312'), ('avg_drawdown', '-33.23'), ('time_in_drawdown', '0.9506'), ('calmar', '0.06524'), ('avg_return_to_drawdown', '0.1981'), ('avg_loss', '-0.65'), ('avg_gain', '0.6161'), ('gaintolossratio', '0.948'), ('profitfactor', '1.087'), ('hitrate', '0.5342'), ('t_stat', '1.508'), ('p_value', '0.1315')]

VIX_mini 30 RollOffsetDays :
[[('min', '-36.63'), ('max', '7.896'), ('median', '0.0252'), ('mean', '0.02016'), ('std', '1.213'), ('skew', '-7.066'), ('ann_mean', '5.162'), ('ann_std', '19.4'), ('sharpe', '0.2661'), ('sortino', '0.2442'), ('avg_drawdown', '-34.72'), ('time_in_drawdown', '0.951'), ('calmar', '0.05086'), ('avg_return_to_drawdown', '0.1487'), ('avg_loss', '-0.6786'), ('avg_gain', '0.6075'), ('gaintolossratio', '0.8951'), ('profitfactor', '1.067'), ('hitrate', '0.5437'), ('t_stat', '1.168'), ('p_value', '0.2428')]

[robcarver17]

"I think the VIX_mini data going back to 2005 might be because Rob used VIX data to backfill the series until 2020, after which the actual VIX_mini data was used."

Yes. I do this in other places, eg SP500 (which is the mini) and SP500_micro, both of which are backfilled with the now dead big contract.

Personally I wouldn't waste time backtesting; eyeballing the adjusted price series is enough to see what the effect of shifting is (which should entirely be due to the difference in roll config rather than small differences in prices).