Hedging Choices (Experienced Traders)
To the experienced traders, you know who you are... Do you think the correlation or co-variance matrix is better to assist with hedging? Also let's say for some odd reason I wanted to trade front spread and back spread and they have different betas.. How would you find hedge ratio? I'm gonna assume you will have to factor in co-variance to account for volatility? Let's say I hedge, do you hedge to keep it risk neutral or at the same, do you try to find possible alpha opportunities to offset a bad trade? Example, turning a spread into a fly. Like let's say I'm trying to hedge a Cal20 strip and I use the front '19 so I build up spread risk along the curve while I'm delta neutral, how would you hedge or what opportunities can I flip this around to possibly turn into a relative value opportunity trades on the forward curve?
faceslappingcompilation Come back into my life and stay...
can be a mistake to blindly use correlation/covarance matrix (depending on the fundamental / mathematical relationship between the securities). (read the wiki)
First you need to know if your securities have a linear relationship over time. If yes, then this approach can work...but you need to inspect how they behave to determine if PCA is the right approach (i've seen things blow up when a relationship was assumed to be linear when it really wasn't...many things are linear in a very small slice of time and price...but over a larger sample they are not).
If the relationship is not linear, then using a linear approach will still give you an output, but it won't be a mean reverting process, which is what you are looking for...in which case you'll just end up with a random outright position and risk that is not what you are expecting. This is something that neural networks are good for (i suggest you could use auto-keras to find the right neural network approach).
However, before going the ai route (which i suggest you learn if you are not there yet) i always start charting the individual securities in one chart over both a longer time frame, and then i zoom in to shorter time frames, and ask myself "does this make sense?" You are looking for a way to understand the relationship...and then you can use the math to zoom in.
Co-variance itself is an intermediate step in calculating PCA...you normalize the co-variance into correlation co-efficients (correlation co-efficients are the slope of the linear relationship)
in the line formula y=m*x+b
y is the price of security A (the dependent variable...the value we are searching for, given x) x is the price of security B (the independent variable...the known thing) "m" is the correlation co-efficient...also called the slope...which is also your hedge ratio (if the relationship is indeed linear).
in the interest rate world, things are linear in a very small slice of time/price...but over a larger sample size, things are not linear. (imagine the price to yield chart...its asymptotic to both axis...so not linear....but in practical terms its usually linear at any one point in time )
I know within rates, the three biggest variances in PCA is level, slope and curvature, respectively... in commodities, what would those be? I haven't heard much about PCA being used for neutralizing certain risk factors in commodities. Personally I've never calculated PCA... trying to find cov - var matrix and using linear algebra to find the eigenvalues and eigenvector... is there any easier way to learn?
Also any feedbacks on my other questions? Like for example, let's the front spread is up 5 cents and the back end is up 1 cent.. .Obviously only an idiot would use the 5:1 hedge ratio...
Also isn't that just linear regression? I know people mistakenly think PCA is similar to linear regression.
each market trades a little different, and i'm not familiar enough with the dynamics of how your market trades (i'm assuming nat gas, but could apply to other things).
for example, if we were talking about eurodollar futures....then i'd know that the thing being measured by the market is the implied fed path...and changes to contract prices are changes to the expected fed path, plus a credit spread between fed funds and libor.
In your market, what external things can you isolate...what is the market measuring? Nag gas is also implicitly measuring weather expectations (demand), extraction, transportation and refining issues (supply)...etc. Now, if none of those things have changed, then sure, spreads can be turned into flies...otherwise you are implicitly taking a position on that externality.
regarding hedge ratios...each market is different...but generally speaking, i would start with visually inspecting the price history of the various contracts that you are considering spreading...and with your eye you can usually determine "are these linearly correlated?"
Excel has some functions to determine the stength of the linear relationship (Slope, Correll, Linest) that are useful.
faceslappingcompilation Hey question, so let's say we have a feb-dec strip and the product is illiquid and the spread is very wide so we don't know how to value these this.. so I bootstrap the outrights down the curve to find the value but I have to make a market for a GZ9 strip vs cal 20... how would you make a market on this?
I mean I can use the midpoint of the front strip but it may not be perfect as it should be off value a couple ticks but let's say i want to be spot on, arb free
Also lets I am able to make a market on this, what would you do with that residual lot? Just gtfo? The reason why i ask is because I'm trying to be active on RFQs and it was one the instrument someone was seeking on CME direct
when you have illiquid securities....you need something liquid as a reference, and then you need a valuation model...which you use to determine what the right price should be for the illiquid stuff. The valuation model is generally the hard part...and is different for every market. (this is my way of saying, i don't know enough about the markets you are trading to be able to be much help here)
Yeah.... The problem is, even the outrights aren't liquid as you go further back of the curve but the Qs and seasonal strips do get traded more but the problem is, as the strips get traded, the spreads may lag and obviously you can potentially find an arb opportunity if you guess it right and try to find which has a higher beta and how to split it evenly but I'm trying to remove from making manual values
I can use the front outright and the front spread and bootstrap it down the curve to get the second leg bid offer but even then it's still hard because the spread can be like 60 ticks wide.
FYI - this is the European natgas, which is mostly commercials trading this shit so I better be right on my values lol because it can literally move 2 cents in a second
People feel free to ask questions or chip in. I mean most people here probably don't know what face and I are talking about but it's good to ask and get other inputs
Whoever is throwing monkey shit at me is a hater.
I mean you'd think that someone who doesn't know how to combine the risk of two calendars would be more humble but then life is always full of surprises
Who mentioned anything about calendar spreads? We're talking about cal19 strip without the front that has no market vs the cal20 and finding the value in a split second and what logic to use
https://www.wallstreetoasis.com/forums/need-an-experienced-futures-trad…
Friendly reminder that you made an entire thread because you couldn't understand (A-B) + (B-C) = (A-C)
Whoever is throwing monkey shit at me is a hater.
Voluptatum voluptas fuga qui aut ut ad. Quia amet vitae nisi veniam iusto omnis similique. Neque molestiae a non voluptatem ut. Deleniti nisi delectus eum blanditiis voluptatem.
Dolor a voluptas rerum. Necessitatibus vitae rerum vero dicta repudiandae dolor eum. Eos ullam qui et iure nostrum. Est amet commodi beatae sed. Vel aliquid vel quod veritatis eaque repellendus. Quaerat ea in quas reprehenderit recusandae adipisci.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...