“A very possible recession within the next three years would take place in the US, economists warned at a meeting of US governors. There is a “100 per cent certainty” of a recession between now and 2018, President of Fosler Group Gail Fosler said on Saturday at the 2016 National Governors Association Winter Meeting, which gathers more than 30 state governors across the country.”
should this be a reason for worry???
Where can I see the math used to make these assumptions based on option prices? In other words, anyone know the best source to learn black-scholes manipulation and theory.
I’m not sure, but I think that this kind of prediction usually comes from the size of the volatility smile and the degree to which put implied volatility is higher than similar moneyness call implied volatility.
Most likely they map a jump diffusion model onto the observed option prices. The call implied volatility is the ordinary normal volatility and the jump would explain the difference between the put and the call IVs. But a jump model does require both a jump size and a jump probability, so I’m not sure where they would get that. I suspect there is enough info if you take a whole chain of option prices and try to fit it.
I’ve seen this technique used in currencies and it’s called “risk reversal” and basically examines the IVs of similar moneyness puts and calls. However it’s often enough just to monitor the difference in IVs. To get the actual probability of a crash, you’d have to try to do the distribution mapping I mentioned above.
bchad is correct. Essentially, puts will generally have higher IV (using SPX as the instrument) due to dividend distribution - S&P futures trade at a discount - and the fact that most, but not all, jumps (big moves) are to the downside while the rest of the price movements follow more of a continuous random walk. Long story, short, they take the bid/ask to get a mid price and find out what the implied volatility is for that strike. Then they plug it into an equation and get the probability of touching for whatever expiration it is.
Easy way: on thinkorswim set your option chain to show probability of touching and the math will be done for you.
One other thing to note is you’ll see probability of touching %s on the call side as well as the put side. Since they’re both being calculated off market prices and the IVs that come w/ that, use puts for the downside and calls for the upside, as these will be more liquid than their counterparts and have “better” pricing to base the calculations off of.
ok, I found that “probablity of touching”… cool
My curiosity was especially peaked becasue of the equation I was playing around with lately (that I learned on tastytrade!) It uses ATM IV to calculate expected price range. Apparently it yields one standard deviation. I thought it may be related to this discussion because once you have the standard deviation, you can solve for the probablity of various strikes.
exp move = spot * IV * (days to exp/365)^1/2
Now, if I may go off on a tangent, I want to discuss what I have found out about this equation. The guys (on tastytrade) were using this to calculate profit targets for scalping trades buy solving it for one day. I have started to practice scalp type trades in paper money, but I am using the equation to pick target entry points. If one of the extemes is met, I take a contrarian position. It has been amazing to observed the consolidation, or even overt rejection, at these calculated range values! I suspect this equation is often used by traders. Anyway, I wanted to back test the effectivness of this equation to pick extremes that were actually contained by the actual price action of the day (thereby resulting in profitable scalping trades). I wrote some code to run on a day candle chart, print the theoretical range predicted by the equation on each candle, and tally up the times when the calculated range was wrong. What I found is that is was wrong more than 32% of the time meaning I doubt that it truly yields one standard deviation. That being said, I also ran the program to look at the calculated range’s ability to contain just the day’s closing price. This yields a much more favorable statistic. Since a scalper just wants to be profitable by the end of the day, I would say that this IS a very effective tool for day trading (if anyone cares).
******sorry OP! right, recession…just keep an eye on the data! ( not the talking heads)
I didn’t follow that too closely, but it basically sounds like a Bollinger Band technical strategy, just using option IV instead of RV?
are you taking contrary positions in the option or in the underlying here?
^^^ exactly bchad… but basing a range on current IV has a big advantage in that it is forward looking, and bollinger bands only have basis in the past. I am taking positions in stocks, futures, and indexes. It’s working!!! (in my 2 days of implementation)
Have you compared your backtests to the same backtests using RV instead of IV? I would think that if RV < IV, then your trade points would be wider under IV than they would be under RV. That would suggest that people using traditional Bollinger band strategies would exit the trade before you and perhaps scoop your profits before you are able to.
If you are using 1 SD, however, maybe that’s not as big an issue, because I think most Bollinger band strategies shoot for something like 2 or 1.5 SDs. On the other hand, you’re going to get more noise by taking a 1SD approach.
Feel free to message me with this if you want. I understand that you might not want to reveal too many inner details of a trading strategy on an open forum.
Well, the article says there is a 50% implied chance of SPX touching 1600 “at some point”. So you need to know the value of a “one touch” sort of option. No fancy jump diffusion is needed - just normal brownian motion. There is a close form solution for this actually, if you use the normal static vol, rates, div, sort of assumptions. Otherwise, it’s pretty standard to use a backward inductive model, unlike what that guy said in the other thread where I didn’t even know how to reply. In fact, 1y down out puts are quoted all the time in OTC markets.
Anyway, this type of analysis just shows what the current implied vol surface is pricing. It’s more like what people are paying for that scenario, not a prediction on the probability of any scenario. Of course people pay more than fair value to hedge a very bad scenario. So I think the article is misleading and frankly kind of alarmist. There are things called “crash cliquets” which pay off if say, SPX declines by 10% to 20% in one day. These can trade for 2x or more of the actual event probability, since the premiums are so low. To say the implied probability is predictive is not accurate.
I do want to continue my analysis by checking compairisons to other technically driven studies. Like I said… it’s been two days and there is A LOT I want to look at. The comparison will be mostly out of curiosity since IV is so much more powerful. IV will price in specific events like earnings and bollinger bands have no way of doing so. What is “RV” by the way? I understand bollinger bands to use a standard deviation of past close prices. I do plan to take a look at the comparison though. More likely, what I had in mind is to solve for a constant to use as a multiplier that will expand the range to higher probability of success (but offer fewer opportunities of entry)
RV = Realized Volatility (i.e. historical vol over a period)
It’s true that volatility is based on returns, and bollinger bands are based on prices, but my sense is that there should be a linear mapping between the two so that a strategy that plots buy/sell points based on vol vs bollinger bands is not likely to be all that different. If the bolinger bands are wide, it suggests that the volatility is wide too. Current price level might be an additional variable that messes up the 1:1 correspondence, but I’m still not sure they are materially different, other than some scaling factor equal to one / current price.
I bring up the issue because sometimes we get too excited about the math we do and then use it because we don’t want to have wasted the effort of discovering things. I know that I have occasionally used option implied volatility in tactical asset allocation and my conclusion from comparing it to using historical volatility was that it was a good deal more cumbersome to calculate and not materially more helpful in the strategy.
In general, if you don’t get a marked improvement in your performance by going the more complicated route, you should abandon it in favor of the simpler calculation - that’s because the more complex calculation usually makes more assumptions to make the mathematics work. More assumptions often means there is a greater chance that one of them is not going to apply in a certain situation, and then things can go haywire in unexpected ways.
It’s not always the case, of course, becasue simple models also make assumptions and sometimes the point of the more complex math is to undo some of the simplifications that have been made. I think you may be arguing it here that historical volatility assumes that the future vol looks like the past vol and that implied volatility is a better estimate of future vol because people are using it to pay for their options.
But remember that option vol is not a magical predictor of future vol’s actual value. It almost always biases higher, in part because market markers want to get paid for what they do and the model automatically attributes higher prices to higher implied volatility, and in part because lots of option buyers are scared (or hopeful) that something big is about to happen. Aside from the bias upwards, it also can also be noisier, and then you start looking at things like vol of vol, etc… Sometimes it adds information, other times it just adds noise.
The other reason that some people like to go to more complex math is because it sounds good in marketing, whether it’s one’s own skills or the fund’s performance itself. By arguing you use more complex math, it avoids the question of “why can’t we just hire someone else to use your simple formulas.”
Anyway, these things are judgment calls. I tend to advocate “simple, unless the more complex version has a demonstrably better track record.” But that’s just me.
Can I agree w/ KMD and bchad at the same time?
IV, for probability analysis, is in my opinion the superior method. This is, assuming you can easily apply it mathematically without large spreadsheets or datasets and such. Most option trading software these days can do this either visually or in easy to use datasets.
IV is superior in that it takes into account known future events such as earnings, news, releases, etc. This is vital for most equities.
For indices however, like bchad mentions, you’ll get very similar results. There aren’t any major events that are going to affect an entire market like there are in individual stocks, so you’ll see similar probability expectations. Finding a strategy between RV and IV though could prove interesting.
Also, IV can be very dependent on having a liquid options market. This obviously does not exist for all underlying instruments. Additionally, most VaR or similar risk models will use historical volatility to account for risk, but there are pros and cons to this as well.
I get why people like IV: it’s more forward-looking, and that can be an advantage, provided that the forward-looking expectations are based on rational analysis and not simply market panic. But often times IV is nothing more than RV + some spread that changes fairly slowly except in panicks.
All I’m saying is that one should compare the backtests with trading strategies that use RV and IV and confirm you are actually getting superior results with IV. And if you aren’t, then maybe rethink whether IV is truly worth the effort and expense of collecting.
My inclination would be to use historical volatility, and then incorporate the (IV - RV) spread as a separate variable in your model. This would have the advantage of allowing you to figure out how much information content is coming out of RV and how much is coming out of IV. However, this would require more of a regression-based approach so I’d have to think of how you would translate it into something like bollinger band type buy/sell strategy.
EDIT: After thinking about it, you would look at Historical Vol, the Implied Vol Spread, and time to expiry at various dates and use that to try to predict volatility between those dates and the option expiration. This would then go into a model to predict future volatility. You would then use the predicted volatility from the model to establish your entry/exit points (the way you use IV currently). This assumes you get statistically significant values for all relevant variables and assuming that it performs better than the simpler model using RV or IV.
ALSO: position sizing is very important with probabilistic trades like this. If you’re not familiar with the Kelly Criterion, become familiar with it.
i’m assuming you guys are not actually scalping and are confusing scalping with basic day trading. i scalped for several months. the target return for scalping is $0.01 to $0.05. my niche was $0.01 to $0.02. my time horizon was 10 seconds to 2 minutes. there is no way you could produce statistically significant data for that time period unless you traded the most volatile and highest volume stock on the exchange and your execution time was in the nanoseconds.
also, it’s not about being right by the end of the day, scalping is as much about never losing money as it is about making anything. you can’t hold a position all day as a scalper or you’ve just wasted a whole day of earning potential. at the first sign that the trade is moving against you, you need to get out and reposition.
scalping is as much about earning liquidity provider credits as it is about making full pennies. these days, scalping is entirely about outsmarting the machines, until the machines adapt, at which time you have to outsmart them again. and so the cycle goes.
MLA… I am experimenting with “scalping” but was not aware there was actually a difference between that and day trading. My thinking here is that I wanted to start working with a basic model based in probability that I could then apply with consistancy and discapline. I like the idea of taking contrarian trades on a group of liquid underlinings when/if my targets are met. I was aiming for entry targets that would give about a 68% chance of success given I hold till the end of the day (if needed). Thats why I see the close as the bottom line for my uses. That being said, in my short experience with doing this in paper trade, I have been truly “scalping”. As I said before, I have noticed legitamate consolidation at these calculated targets. I have been closing positions the size one future contract or a few hundred shares of stock for about $100 profit within a few minutes. So, I am back testing, practicing, logging results… having a ball really! EDIT… MLA, where did you learn/ how did you develop your scalping techniques? I am a total sponge right now
i worked in a prop shop right out of school for 4 months or so. my techniques were developed solely through my own trial and error. this was back in '08 so when i got an offer for a salaried job that required CFA credentials, i took it. i was quite profitable when i left though. was pulling in $300-$1000 during the hours of 11:00-12:00 and 1:00-3:00 in my last few days on fairly small size. the market was amazing for scalping back then as you’d get massive volumes executed on both the bid and the ask with very little bid-ask rollover. i’m sure it’s more difficult these days with HFT and lower volumes.
to answer your question, as the market make-up changes so much these days, probably on a day-to-day basis as algorithms adapt, the best way to learn is trial and error and constant trading. if you’re truly scalping (i.e. trading in and out every minute or so), it’s best to stick to one stock and learn how it trades and learn how the algorithms act on that stock. my tip is to never hesitate and if you get a gut-feel about the bid-ask rolling over, trust it. the gut-feel is always always your brain recognizing marginal yet important price movement faster than it can recognize that it is recognizing marginal yet important price movement. this gut-feel thing is why day trading is so much different than investing. gut-feel is terrible for investors but essential for day traders, particularly scalpers.
^^thanks, Matt!