The critique is the same for all papers and systematically impacts all estimates, so seeing more consistent results doesn't invalidate the criticism.
That said, it's quite possible the number of cases is undercounted! Someone should do a meta-analysis to obtain a better estimate, given the testing uncertainty.
For your first question: I don't know, but I speculate the majority is hedging.
Second: I get my data from CQG, but you can also get the same feeds from IB. The specific exchange you need for VIX is "CBOE"(http://www.cboe.com/vix). ampfutures.com is a broker that offers this, but there are many(possibly better) others that offer it as well.
Third: Historical data for the depth isn't available from any broker I have used. That said - I'm a retail guy, and there may be options for institutional traders that I'm unaware of. As a retail guy you can buy this data though, just not from a broker. IQFeed offers it(6mo back for market data, and 6mo for 1-tick resolution data) - but it's expensive. The last quote they gave me was $1350/month for CME and CBOE tick-by-tick resolution with 10 levels of depth.
Which broker are you using for execution? IB? Curious if there's a better one for futures/futures options since I've started getting into them recently on IB.
If it's not revealing too much, what kind of stop loss distances are you generally using in your strategies (ticks)?
Are you using any kind of "walk forward optimization" in your testing?
You mention in another comment that you intentionally do not trade during _expected_ market volatility. Do you have parameters for sitting on the sidelines during _unexpected_ volatility, e.g., VIX over certain threshold, time since last Trump tweet according to Twitter's firehose API (serious), etc.?
For execution I use Tradovate(which uses Dorman Clearing LLC). They have low commissions, and offer a membership to reduce commissions to $0(you still have to pay exchange costs though). Before Tradovate, I had used AMP for execution, but eventually left them over the cost of commissions.
I use a 8 tick stop and a variable profit target(a result of weekly re-optimizing). Last week the target was 24, and the week before the target was 16. The stop loss value I don't like to change, and 8 ticks has been enough for "good" entries. If I need more than 8 ticks, then I believe the entry price is my mistake - not the stop loss value.
I perform optimization every Friday, using that week as the training data to find the "best" thresholds for changes in depth before signaling a trade. Then the next Monday, I'll use those new values all week(and repeat the cycle that Friday). Basically - re-optimize every week. It's not a lot of work, just input the starting date, click optimize, and wait ~20m for the outputs. I like to keep track of each week's settings, with the plan to one day review their changes, and try to reason about why those changes happened(I haven't done this yet though, it's one of those one-day-I-ought-to ideas).
I don't have any algorithm parameters for unexpected volatility, the closest I have is a condition that turns the automatic entries off if the nearest 4 bids and asks(examining a total of 8 prices) contain more than 4 ticks of spread(no bids or asks), and both sides limit orders sum to less than 30. I didn't add that until sometime around Valentine's day 2018(that was not a good couple-of-days). When that happens, it disables entries and sends me an SMS.
I used to follow a lot, and drank a bunch of the kool-aid from Jigsaw Trading. Now I read very little, other than the economic calendar for US/Canada/Europe each morning. I'm looking for high volatility events(such as FOMC, GDP, etc) so I can make sure I am not in the market when they take place.
I haven't read any books about algorithmic trading(but I've read a ton about trading in general...favorite is Mark Douglas's 'Trading in the Zone'). Most of my introduction to algorithmic trading came from a trading platform(Multicharts.NET). I wasn't a .NET programmer specifically, but they include the source code for all indicators/signals that ship with the product. This made it really easy to tinker with automated trading signals, and I don't think I would have ever attempted it without that exposure.
Also, you didn't ask for this piece of advice...but: be very cautious of tips from any blog or news outlet. The people producing those are under tremendous pressure to produce "something", even when there isn't any insight to be had for a day/week/event/etc. They will produce something anyway, relevance be damned. If you needed some help/ideas, you would be better served joining a live trading group. Many of these are free, and they give you a chance to listen(and speak) to other traders, all trading at the same time, usually with the same instruments. I learned some things from GPI Trading Group(http://www.gpitradinggroup.com/) that saved me a lot of time and headache. They are focused on the ZB(bond futures, not equities), but back at that time so was I. I got a lot of feedback from their members about my indicators and automated signals I was working on, and much of it was invaluable.
It's supposed to be, but I stare at it every day for 2 hours(for years)...and it just doesn't work out that way. Half the time moves initiate in the VIX, and only moments later does price move in equities.
Example: this morning at the open, the VIX rose 7 ticks in the first 20 seconds - there was almost no volume to the opening move(on the VIX), and price(NQ) was kicked down 20 points. When the NQ was at this bottom, the inside ask for the VIX changed from 150-ish to 700 - the NQ price moved back up 20 points, but the VIX did not move. The only thing that changed was the depth. Now the NQ is right back at the open...that same 700 ask faded back to 100-200, and the price dropped back another 20 points. Keep in mind, this is the first 5 minutes of the open. It doesn't get more chaotic or random than that period of time. However, as chaotic and random as it might seem, it's like this "most" days(say, 7 out of 10).
One of the large reasons that instruments(like the ES, NQ, or VIX) don't behave the way the way their prospectus might indicate...is humans. Another example, the NQ is supposed to reflect the Nasdaq 100 basket of stocks, weighted appropriately...yet it does not. There is a whole school of trading around "program trading", which is automated trading to fade price movements when this imbalance occurs. If the prospectus is to be believed, this imbalance should never occur. Human beings are the answer: a fund manager has a gut feeling(or some technical analysis, or signal) that equities will move up, so he shorts the VIX for 2,500 contracts, which will absolutely move equities up, at least temporarily. It's not the rules or defined behavior of either instrument - it is the intent of someone with enough margin to execute a 2,500 contract trade. That person(and persons like him/her) are the ones controlling those instruments - not an arbitrary set of rules, or the formula laid out in the instrument's prospectus.
After saying all that...also realize a lot of this is truly random. The moment in time that Trader Joe and Trader Bob both kick off 2,500(or 10,000 for that matter) contract trades is not known by anyone but those humans. So to them, it's not random, but to hundreds of thousands of other market participants, it's random as hell, and causes a lot of "whoah man! did you see that!?" moments. Those pesky humans get in the way of everything :)
From your experience, would you say the VIX futures are useful for anything other than traders taking chunks off the bid or ask moving the equities a few seconds later?
Oh certainly yes it is, but that's just not my use-case(anymore). In the past(when I had longer-lasting positions), the VIX was a useful indicator of "is this move real, or not?". Typically, if all 3 of the major equities(NQ,ES,YM) move in tandem over a 5-30s period, and the VIX inversely matches their movement - then that was my entry signal. When the VIX doesn't match, prices tended to return to their starting point moments later. If you speculate as to "why did the prices snap back?", my best guess was it was the result of profit taking or bad positions puking.
My current use of the VIX is more short-term. Over time, I found ultra-short-term positions worked out the best for me(in terms of profitability, stress, and opportunities per day - stress was #1).
Twine | Onsite | NYC or Bay Area | Data Engineer, Full-Stack Engineer
Twine (https://twinelabs.com) helps companies make better decisions about the lives and careers of their employees. We’re growing rapidly, and already serve many of the world’s best technology companies: Evernote, Mapbox, Segment, Asana.
We're hiring a Data Engineer to build our data infrastructure (ETL, schema design, APIs), and a Full-Stack Engineer for web development (data-driven, context-rich features and workflows).
We're a small, high-intensity team (and well-funded by unicorn founders and Sand Hill Road VCs) - you'll have true ownership and incredible impact. We believe thoughtfulness, self-awareness, and empathy create great teams and products.
Twine | Onsite | NYC or Bay Area | Data Engineer, Full-Stack Engineer
Twine (https://twinelabs.com) helps companies make better decisions about the lives and careers of their employees. We’re growing rapidly, and already serve many of the world’s best technology companies, like Asana, SurveyMonkey, Segment, and Evernote.
We're hiring a Data Engineer to build our data infrastructure (ETL, schema design, APIs ~ Airflow/AWS), and a Full-Stack Engineer for web development (data-heavy analytics workflows and designs ~ Django/React/Redux).
We're a small but quickly growing team - you'll have true ownership and incredible impact. We believe thoughtfulness, self-awareness, and empathy create great teams and products.
I'm CTO - you'll find my contact info on the website in my profile
