Home Forums Main Forum The Z turn

Tagged: 

  • Author
    Posts
  • Curve Advisor
    Keymaster
    Post count: 612
    #328 |

    When I first traded at JPM back in 2000, the year-year end turn was around 6bps. Nowadays, it is closer to 0.5 to 1bp. Almost everyone I’ve worked with (whether they were on the prop side or market-making side) used some sort of “smoothing” methodology to back out what the turn should be. So the typical thing to do was to have some cell(s) in your spreadsheet you could manually adjust. I’m not sure if the way people look at the turn has changed. It’s one of those things that people generally don’t talk about, but if you trade a lot of 6 mo flies, it becomes very important. You will be long or short 4 Z turns in half of the 6 month double flies.

    I have noticed a tiny pick-up in the Z turn the past year. It’s not clear to me if it’s just a function of the shorts leading towards selling the Z contracts (to line up with the SEPs), or if this will be a trend as rates raise.

    • This topic was modified 4 years, 2 months ago by  Curve Advisor.
    • This topic was modified 4 years, 2 months ago by  Curve Advisor.
  • mjaws
    Participant
    Post count: 4

    I typically do a linear interpolation across the futures complex comparing the results between exchange listed, combination products, and synthetic combinations for sanity. I agree with your .5bp – 1bp range recently. Have you looked at the option market for queues on this adjustment ever? may be an interesting data point there?

  • Curve Advisor
    Keymaster
    Post count: 612

    Well, as I was telling the guy in the Data Set Methodology thread, I knew a huge swaps market-maker who just “eyeballed it” on his spreadsheet. That having been said, I like your approach to looking at the turn from various sides. I’ve never looked at the options markets for clues to libor. I think in these times, that is just going to be strongly flow-driven as people like taking views on year-end rates. One thing I used to pay attention to is just looking at the libor fixings as they went over year-end. You can make assumptions about how the current year’s fixings will affect future years’ fixings.

  • mjaws
    Participant
    Post count: 4

    interesting, I came across this article by burghardt. a bit dated but there’s some good information in here. Galen’s work has been very helpful to me over the years.

    http://www.jamesgoulding.com/Research_II/Eurodallar/Eurodollar%20(One%20Good%20Turn).pdf [NOTE: You have to copy and paste the link into your browser. For some reason the highlighted link does not work.]

    I also posted this into the dataset thread as well.

  • Curve Advisor
    Keymaster
    Post count: 612

    I see the “turn” in Z6 and Z7 being pretty high (about 2x as high as the others). I think it’s mostly just people positioning for rates at the end of 2016 and 2017. But it’s interesting that in those two years, we also have 3 and 4 day turns. Back when I was trading, no one seemed to pay attention to it, and I suspect that hasn’t changed, but I think it’s an interesting coincidence.

  • Curve Advisor
    Keymaster
    Post count: 612

    When the 3 month libor cash fixing went over year-end, there was nothing. So that doesn’t mean there should be zero turn, as it could be a function of “zero” rates, but as we rally, the turn should start going to zero.

  • me
    Participant
    Post count: 27

    <cite>@Curve Advisor said:</cite>
    1) you need to determine if there is any kind of “year-end turn” or other seasonal effects in the data. Some STIR contracts traditionally have a turn, and others do not. It has been a while since I looked at the other STIR markets, but you will quickly be able to tell if you graph the strip of 3 or 6 month butterflies and there is a noticeable seasonal pattern of kinks (especially in the long end, where there is less likely to be activity priced in). Also, “year-end” may not always fall on Z – for example, in Japan it is H. Once you identify and quantify the turn, you need to adjust the futures prices for them, to get a “turn-adjusted rate”. See the “Z turn” thread for more info on the turn. I would imagine the turn in most currencies right now would be very small to negligible, and for some it is just zero.

    OK – so this is a strip of the 6 month flies at a point in time, yesterday. The red dots are flies centred about the Z-contract. Optically we can see the ‘kinks’ pretty clearly.

    How do we know how much to adjust? Are we basically trying to make the kink go away? So, in scatter plot, I drew a black line between two flies that surround a fly that is centred around the Z-contract. Are we trying to eyeball a parameter that fits these three points to the black dotted line?

    We’d probably have to look at a time series of strips to make a robust estimate of the Z-adjustment?

    I looked at the turn adjustment thread. My head hurts at the thought of more linear interpolation… But, an in interesting phenomenon that the ‘One Good Turn’ article illustrated is that the turn premium seems to get worse as we approach the season? If we have one parameter for each year’s Z-contract, and it does not vary based on the time of the year, is that OK? Admittedly, the article shows the serial month spread, not sure whether it would be material on the quarterly.

    Attachments:
    You must be logged in to view attached files.
  • Curve Advisor
    Keymaster
    Post count: 612

    That’s a lot of questions. Let me think about the best way to present an answer and start to answer this tomorrow in the Z turn thread.

  • me
    Participant
    Post count: 27

    Well that certainly makes me feel less dense! Glad to keep giving you a challenge.

  • Curve Advisor
    Keymaster
    Post count: 612

    Q1: Are we basically trying to make the kink go away?

    A: There is a bit of chicken and the egg going on in the pricing of the Z turn. You can try and estimate what the year-end funding pressures are going to be, or you can try and infer it from the shape of the curve. Typically, people use one to infer the other. If you are a short rate market-maker, you may see flows that give you an idea of flows. For the rest of us, we have to estimate this from what is being priced into the curve. That having been said, the JPM marketmaker I knew appears to have “eyeballed” it himself off the curve. To avoid the circular reference errors, let’s just for simplicity assume the markets are efficient and all information is in the shape of the curve.

    The curve does not always “have to” be smooth. There can be cases where there can be some severe kinking in the curve, if there is some news that affects a particular date (i.e. Y2K, Fed guidance, Fed dates, etc.). However, all things being equal, you would expect the yield curve to be smooth, because the curve imbeds liftoff and hiking probabilities, and you would think the probability distributions (as well as libor spreads to FF) are reasonably smooth. However, you should always ask yourself if it makes sense that a particular part of the curve should be smooth or kinked.

    So assuming the markets are efficient and the curve should be “smooth”, then finding some adjustment to make the “kink go away” makes sense. The question is, what is the best way to get rid of that kink?

    [next: How do we know how much to adjust? … Are we trying to eyeball a parameter that fits these three points to the black dotted line?]

  • Curve Advisor
    Keymaster
    Post count: 612

    Sorry for the delay… it was a holiday weekend after all.

    Q2: How do we know how much to adjust? … Are we trying to eyeball a parameter that fits these three points to the black dotted line?

    A: As I mentioned above, if you take a probabilistic interpretation of curvature, the curve should be “smooth”. So “yes” you are trying to find the turn values that would smooth out the kink in the curve. However, I’m not sure if fitting “these three points to the black dotted line” is the best way. So this again is one of those areas where you can put as much (or as little) work into it as possible:
    – I have a preference to use the 3 mo fly curve to look at the turn – I feel like you get more granularity. There is also info in the 6mo fly curve, but it is harder to interpolate off of that, as every other 6 mo fly is affected by two turns – both M-Z-M and Z-M-Z have two turns.
    – You can use anything from “eyeballing it” to creating a complex model to back out the turns. I don’t think there is any “right” answer – use the method you feel most comfortable using. I think when you look at 3 mo flies, you are focusing on a small enough area where you can attempt to fit adjacent flies to a straight line, with a reasonable enough error margin.
    – You can use multiple methods. For example, you can use your calculation method of choice, and then manually adjust if something does not seem right to you.
    – You can also use data from the libor cash fixings (as well as overnight index swaps, or other short term vehicles) for the front turn, although it is subject to some noise. For example, if the markets are quiet, you can periodically see what happens to the cash fixings as they go over year-end. This could be a useful comparison point if you use constant turns throughout the entire curve (you adjust all Zs by the same amount).
    – You also have the option of using different Z adjustments for each contract year. This is my preference, and I will discuss this further in another post.

    The big check will be… if your 6mo fly curve looks reasonably “smooth” after you plug in the turns you decided to use. Every alternating 6mo fly will have zero turns, so you have a bunch of good reference points for what the shape of the curve should be.

    [next: caveats and other considerations]

  • Curve Advisor
    Keymaster
    Post count: 612

    Here are some other considerations:

    – Keep in mind you can have bad closes. So barring some system where you correct for bad closes in your turn calculation, you probably want to take some sort of x-day running average for the turn (or some other method of controlling for a bad settle).
    – You may want a turn-calculation method that you can easily recalculate/recreate, in case you want to incorporate turns into your historical database.
    – Keep in mind the year-end turn can last anywhere from 2 to 4 days (depending on which days the turn falls on). However, empirically, it rarely happens that a 4 day turn is priced 2x as much as a 2 day turn. I think it is because there are enough people who just use the same fixed estimate for every turn on the curve.
    – Turns do not move that much day to day (or even month to month). So this is not something you need to agonize over on a daily basis. You should look at it once a week or so if you have a lot of net exposure in Z.
    – You can use the same turn for all the Zs, possibly incorporating a fractional amount of day-count depending on whether it is a 2 thru 4 day turn. This also provides you with a larger data sample.
    – I prefer using independent turns for each year – mostly because the turns for each year don’t change much in the short run, and so if you have on a 6 mo fly trade involving the Zs you want there to be some consistency.
    – Keep in mind what you are measuring may not be the year-end turn, but just a lot of people wanting to take a position on Fed rates at the end of the year.

    [next: One Good Turn article]

  • me
    Participant
    Post count: 27

    First, thanks for the response. My apologies, I had no idea that you had responded already, probably because I am not ‘subscribed’ to receive email notification of replies to this thread.

    Say I’ve decided to go with the eye-balling method on my historical data-set. I’m looking at 5-10 years of data. I’m not sure that I can clearly remember the news impacting the Z contract each year. So, I opt for ‘visual accuracy’ of fitting a curve over mental accuracy of remember historical facts. As suggested, I will look at the 3 month flies and try to ‘eye-ball’ a parameter to fit to the line between the 2 flies adjacent to the one centred around the Z contract. And I will find one turn adjustment parameter for each year.

    This will probably be addressed by the subsequent post on the One Good Turn Article, but is it stable for the Z-parameter, for each yearly Z-contract, to be constant throughout the year. That is, we have the same adjustment for, say, Z5 in Feb15 as we would in Nov15? The turn effective is noticeably magnified during the holiday period as illustrated in the One Good Turn article.

    If a parameter, constant for the entire year, is good enough for our purposes, how many strips of flies for each Z-contract do I need to look at to get a robust estimate? There will be some variation in curvature each day, and there are more or less 251 trading days in a year. Should we look at strips for all 251 days, perhaps grouped or averaged by each IMM period, as the Z contract would move. This Z5 month would be ED4, but a few months ago would have been ED5.

    As I write this, I’m thinking that the Z-parameter for each contract year probably can’t be constant through the life of the contract, and thus likely also for the current year? For example, 3 years ago, EDZ5 was in the blues, further back on the curve. We were also in a very different policy environment then from today. I’m not sure we could use the same turn adjustment for Z5 that we used 3 years ago, today.

    Ultimately I think I’m trying to get to how many parameters we need to estimate. 1 parameter for each z-contract, for each year of it’s life span? So if our data set goes out to the purples, that’s 6 Z contracts x the number of years of data x the number of estimations each year. Is this correct?

  • Curve Advisor
    Keymaster
    Post count: 612

    Since the last two reader posts have both referenced the “One Good Turn” article, that is a good segue into today’s scheduled topic. I’ll finish answering the questions on the first post before moving to the second. I also probably need a little time to think about how I want to present my reply to the second post.

    Q3: the turn premium seems to get worse as we approach the season? If we have one parameter for each year’s Z-contract, and it does not vary based on the time of the year, is that OK? Admittedly, the article shows the serial month spread, not sure whether it would be material on the quarterly.

    A: Keep in mind that article was written in 1994. That’s over 20 years ago, and very close to the Golden Age of Bond Trading, that a former colleague used to describe as “any idiot who knew how to use a financial calculator could have made millions in the bond markets.” A long time ago, thinking about a 1 month rate and breaking it down to the rate for individual days was quite “revolutionary.” Now… not so much.

    While it is possible for the markets to “slightly” overestimate or underestimate the turn as the year progresses, that does not happen as a rule. Over the past five years, the year-end turn on the first contract *fractionally* decreased 3x, increased 1x and was unchanged 1x. I suspect the small decrease is probably due to the fact people had been using “0.5” for the turn as a default for the entire curve, and as the year progressed, it started converging closer to zero, as the additional liquidity in the markets makes the funding pressures negligible. So it’s okay not to make any major trend adjustments based on the time to the turn.

    I used to trade the 1 month futures once in a while, but it looks like it stopped trading (CME: EM, Globex: GLB). Theoretically, the same concepts in the article hold for 3 month libor contracts, however, something like the turn is easier to see on a shorter-term maturity contract than a longer term one. The article is still useful for the description of the theory, but you can mostly disregard the “current example” as an unusual case study.

  • Curve Advisor
    Keymaster
    Post count: 612

    Q4: We’d probably have to look at a time series of strips to make a robust estimate of the Z-adjustment?

    So how best to estimate the turn? It is not clear to me which statistical method is “optimal.” Again, this is one of those things where you can put as much or as little time into it as you want. Keep in mind the “big picture” – we are trying to back out a 2-4 day rate out of a 3 month contract – in the middle of a potentially volatile period of FOMC rate changes. And there are traders and algos looking to take their “view” positions on year-end rates (ie the Z contracts). That is probably why the open interest on the Z contracts are so high (the OI on Z6 is 955K, while the average of U6 and H7 are only 566K). So there is going to be A LOT of noise, and at the end of the day, any estimates you do for the turn will just be loose estimates.

    Even just using a constant “0.5” for all turns (and occasionally asking in the thread what a good amount to use for the turn) is not going to be terrible, and that takes almost zero effort. However, you need to realize that before you start doing that M6-Z6-M7-Z7 6mo double fly (that has 4 turns), you need to take a closer look at where the turn is, as a 0.25bp change in the turn could be 1bp on the trade. Also keep in the back of your mind which are the 2, 3, and 4 day turns – you could even consider making a slight adjustment for this.

    The next step up would be to do some sort of “straight line” approximation, where you try and smooth out the humps around the Zs. So something similar to what you have suggested – but this doesn’t involve linear interpolation, it’s just basic algebra. If you use 3mo flies, keep in mind there is a turn in the MUZ and ZHM flies as well as the UZH fly.

    The next step up would be to fit some sort of tight curve to the daily ED prices and/or the fly prices. This probably can not be done on Excel, and would require some programming.

    And as previously mentioned, there are going to be settlement issues day to day, so you probably want to take some compilation of the past week(s) data. But this could be as simple as a x-day moving average.

    [This post may change]

  • me
    Participant
    Post count: 27

    Thanks. I will think about this a bit. I think this solves the z-turn for the ‘current state’ spreadsheet. But what about for the historicals? I suppose that’s in the next post.

  • Curve Advisor
    Keymaster
    Post count: 612

    <cite>@me said:</cite>
    Thanks. I will think about this a bit. I think this solves the z-turn for the ‘current state’ spreadsheet. But what about for the historicals? I suppose that’s in the next post.

    You can just use what you do for the current state, for each day in the past. Or I suppose you could use a different methodology for the historicals than the current state, in case you want a “check.”

  • Curve Advisor
    Keymaster
    Post count: 612

    @me said:
    This will probably be addressed by the subsequent post on the One Good Turn Article, but is it stable for the Z-parameter, for each yearly Z-contract, to be constant throughout the year. That is, we have the same adjustment for, say, Z5 in Feb15 as we would in Nov15? The turn effective is noticeably magnified during the holiday period as illustrated in the One Good Turn article.

    If a parameter, constant for the entire year, is good enough for our purposes, how many strips of flies for each Z-contract do I need to look at to get a robust estimate? There will be some variation in curvature each day, and there are more or less 251 trading days in a year. Should we look at strips for all 251 days, perhaps grouped or averaged by each IMM period, as the Z contract would move. This Z5 month would be ED4, but a few months ago would have been ED5.

    As I write this, I’m thinking that the Z-parameter for each contract year probably can’t be constant through the life of the contract, and thus likely also for the current year? For example, 3 years ago, EDZ5 was in the blues, further back on the curve. We were also in a very different policy environment then from today. I’m not sure we could use the same turn adjustment for Z5 that we used 3 years ago, today.

    Ultimately I think I’m trying to get to how many parameters we need to estimate. 1 parameter for each z-contract, for each year of it’s life span? So if our data set goes out to the purples, that’s 6 Z contracts x the number of years of data x the number of estimations each year. Is this correct?

    I think I answered most of these questions in response to the previous poster. The main thing I didn’t touch on is how to look at the historical data. The value for the Z turn can change over time. I don’t have THE “answer” for what is best to use. I think it is more art than science, because while the mechanics of the turn are known, we are trying to back it out a 2-4 day rate by using much coarser and noisier 3 month data to try and back it out.

    I do not mind your suggested approach of using some sort of historical statistical approach to back it out. I think you need to use a shorter than a rolling 1 year time horizon, unless you somehow weight the more recent observations higher. Because on the off-chance there is a change in the turn pricing, you need to be able to see it change sooner rather than later. Using one or more years of equally weighted data will obscure more current trends in the turn. There haven’t been many significant changes in the turn pricing, but I an speaking hypothetically *if* there is one.

    Looking at constant contracts would correct for the number of turns in a particular year for specific Z contracts (i.e. looking solely at Z6). So on the margin, this could be helpful because you would be comparing a 3 day turn to the same 3-day-turn contract. However, I think it is slightly more important to look at at the levels of the turn on a given day, and how that changes over time (ie what is the level of turn this month as compared to 6 months ago). So I would try and somehow incorporate the latter into your model.

    But each of these things would incorporate a lot more work, and as I mentioned previously, you can put as much or as little time into this. Because the magnitude of the turn has been consistently small, it is not clear to me spending an enormous amount of time on this is time well-spent. But developing an accurate “true” estimate (assuming this is possible) may allow you to eke out an extra 0.25bp on every Z contract you do. This may not sound like a lot, but it is over the long run, and you would have additional confidence on the Z-related trades that you do.

    Keep in mind the overall goal is to develop a reasonable-looking graph with some of the “noise” (including turns) removed so that when you look at historicals, you will get a reasonable idea of what the trading environment is like for various curve structures.

  • Anonymous
    Post count: 2

    Hi Joseph,

    I’ve been following all of your insightful postings, and think I have a good handle on the interpolation, and the days counts between the IMM dates (personally I smoothed them out non-91 day clusters by eyeballing them, which is frustratingly unsystematic. But I’m no quant).

    What I’m struggling with is the Z turn. I get why it’s there, and how it pushes the Z contracts to a small premium. But how does this get transformed into P&L? Initially I thought the kinks only fell out of the curve (and got banked) once the Z contract expired at the front of the curve. But from some of the examples you put forward this can’t be the case unless you have a holding period of many years. Clearly you do not. And thinking about it further, I must’ve been wrong.

    So is it more that you are using constant maturity to observe a fixed point on the curve and then using the “ripple” caused by the Z contract passing through the structure to make money? Taking a fly for example. If the Z has a turn of 1bp, does it theoretically push the fly down 1bp when it is either the exact first or the last contract in the fly. And push it up 2bp when it is in the exact middle? (excluding all other factors, obviously).

    And if this is the case, would you try to avoid selling a fly that currently has Z as the last month?

    Many thanks,
    RT.

  • Curve Advisor
    Keymaster
    Post count: 612

    The reason you need to adjust for the Z turn is because when you look at historicals, or even the current curve, you want to be able to compare the similar structures with no turn, in an apples to apples way with the structures with a turn. For example, buying the M9-Z9-Z0 fly at 1 (which is like 0 turn-adjusted) and selling at 2 (which would be 1 turn-adjusted). You need the turn adjustment to see that the structure at a price of 1 is like a turn adjusted price of 0. So when you look at the turn-adjusted constant maturity historicals, you can see there is value at 0 (turn-adjusted) and you can also see that a turn-adjusted target of 1 is very reasonable. But there is no P&L impact. Buying at 0 and selling at 1 (both turn-adjusted) would be a P&L of 1bp. But when you actually go to do the trade, you would buy the fly at 1 and sell at 2 (both not turn-adjusted), which would also be a P&L of 1bp.

    Now there is a little uncertainty with what the Z turn is, and so if a trade was borderline enough, you can avoid the uncertainty by not doing the trade with a lot of turns. But there will be other times where you like the idea of a trade AND the turn that is priced in may also be favorable, so in such a case, you may look forward to doing the trade.

    Let me know if the above did not answer your questions.

  • Curve Advisor
    Keymaster
    Post count: 612

    When you trade a lot of flies, you should do the following every so often:

    * Keep track of the net number of turns you are long or short for each Z contract.
    * Sum your exposure over all Z contracts. If you are “excessively” long or short Z contracts, make sure that is how you intend to lean. “Excessive” depends on your particular situation and view.
    * You can reduce some of your exposure if your Z positioning does not fit your Z view (if it is too excessive).
    * If you like your positioning, consider a 3 month (or 6mo) fly (or double fly) trade that will counterbalance your Z positioning, so your net Z positioning will be reduced.
    * You should also do a review of “weighted” Z positioning. The year-end turn for various years could vary from 2 to 4 days. Weight the number of contracts you have in each bucket by the number of turn days, to make sure your positioning is not “excessive.”

    This is currently not something you need to regularly, because the Z turns have been reasonably stable the past few years. However, it is possible rates could move a lot going forward, and it is possible that the value of Z turns could change by 0.5bps or more. You should be prepared.

  • Laurent
    Participant
    Post count: 7

    Thank you for this, however I think I am still struggling to turn adjust the Z contracts and would appreciate some help if possible.
    Would you be able to share the current 6 months fly curve by any chance? I have attached a copy of mine with the 3 months curve. They both look very jagged…
    I have tried to adjust them by slightly changing the turn but that doesn’t really help.
    I am still building my spreadsheet based on your recommendations in the “Basics: Basic Eurodollar Analytics” thread and still have a lot to do but I would like to fix this issue first.

    Any hints on what could be wrong?

    Thank you!

    PS: The red line is formed using outrights, the blue line is formed using spreads/flies exchange quotes (all of them are weighted). For the weighting I use the following formula: (AskSize*BidPrice+BidSize*AskPrice)/(AskSize+BidSize).

    Attachments:
    You must be logged in to view attached files.
  • Curve Advisor
    Keymaster
    Post count: 612

    The bar charts are pretty sexy. 😉

    I have to ask for some more information:
    * what is “ws” and “wo”? “wo” seems to be the “standard” calculation, but I’m not sure what additional adjustment you are making in “ws”. Is this is calculation based on the spread (or fly) market, that I assume is on another sheet?
    * what is “FoF”?
    * Do you use the “91” or “98” daycount in any of your calculations?

    Here are some observations:

    * The 3 mo fly curve is jagged because of the scale – your major horizontal lines are only a half bp apart. The jaggedness is a function of what trades people want to do (people don’t look at the flies when they place bids and offers). It looks like the spikes are generally within 0.25bps of the smoothed curve. This basically means the center contract of the fly may only be off by 0.125bps (since the center contract is calculated 2x), which is not a lot, if you consider the bid/ask is 0.5bps wide. AND the adjoining flies will also be “off” in the opposite direction by 0.125bps, because its negative is used in the calculation. So it only takes a little move in the bid/ask sizes to get magnified in 3 consecutive flies. The 6mo curve is smoother because the scale is different and the noise gets distributed over more months.

    * You need an additional methodology for when the bid/ask is wider than 0.5bps. This is what is causing the noise further out the curve (where there is less liquidity). Because if it is that illiquid, the bid/ask sizes may not be as relevant. You can just use the middle point, nudged based on the other pricing information available to you.

    * Most people just “back out” each year-end turn. I use a 10 day running average, but I have not done any thorough statistical analysis on it. So my methodology may not me any more correct than just using “0.7”. As of Monday, I am using 0.7, 0.7, 0.7, 0.5 and 0.3 for the first five turns. This is more art than science.

  • Curve Advisor
    Keymaster
    Post count: 612

    I’m not sure what your background or interest in looking at curvature (academic, day trading, job-related, etc) is, but if you like visual representations, here are some thoughts on things you can add to the charts:

    * some measure of historicals may help if you want to see where we’ve been the past (period high, period low, 3 months ago, etc). This may be more interesting on the “bigger” structures like the 1 year flies.

    * some measure of the current market (daily high, daily low, current bid, current ask, etc). This may be more useful if you like jobbing the markets, as it this “may” help you identify legs you may want to lean on. This will be more interesting on the “smaller” structures like the 3 and 6mo flies.

    Also, the single contract, spread and fly markets feed into each other, so you probably won’t see a huge amount of divergence. However, the prices can only feed in if there are the necessary legs to lean on, so there is information that can be gleaned from all three markets.

  • Laurent
    Participant
    Post count: 7

    Thank you for your answer!

    To answer your questions:
    * Regarding s or wo, you are correct, it is the standard weighted calculation based on either the outrights or the spread (fly) market (the data is hidden in a group).
    * FoF is a fly of fly, I think you call this an arb, am I correct? I don’t really use it but I wanted to keep an eye on them.
    * I do not use any day count in my calculations. I tried but it didn’t really work and then I decided to leave it on the sheet even though I do not use it.

    I’ll update my sheet to get the midpoint if the bid/ask is wider than 0.5bps.
    I am still trying to understand how to be able to come up with a systematic way of calculating the turn as I’d like to be able to use historical data (and without spending too much time on it…). From what I understood by reading this thread, I should try to smooth the 3mo fly curve and calculate the turn values that minimize the differences (and taking a 10 day average as you suggest), does this sound sensible?

    Thanks for the suggestions, I’ll review them and see what can help me and will update my spreadsheet. Regarding my background, I’m very new to curvature trading but I find this very interesting and want to learn more (trading and job related).

  • Curve Advisor
    Keymaster
    Post count: 612

    That method of estimating the turn is crude but very simple. I’ve sat with guys who did anything from use a constant for all turns to smoothing each turn daily by eye. Even our swaps dealer at the time (and JPM had a very large swaps desk) seemed to just wave a finger in the air (or at least that’s what he claimed). The thing is, you have a lot of error as it is because there’s some uncertainty about whether any contract is going to settle on the bid or offer at the closing snapshot. If you look at the 3 month fly settles on any given day, it’s pretty jagged. And the jaggedness can vary from day to day. So your inputs are going to have quite a bit of noise to begin with.

    As a general rule, I think it’s good to just get the “meat” of the spreadsheet going, and then you can decide if you want to spend more time on a particular area honing it.

  • Curve Advisor
    Keymaster
    Post count: 612

    Viewer Mail: Have noticed that open interest on both edz8 and edz9 are elevated in relation to rest of the curve. Appears has been that way since September. We dropped some after dec midcurve option expiration but still high. Either someone has large edz8/edz9 position. Or maybe related to the Turn??

    Answer: Actually, the EDzs have had elevated OI for well over a year: https://www.curveadvisor.com/2016/05/algos-among-us/

    Interestingly, they seem to lighten up on their positioning for year-end. The OIs in the Z contracts have dropped noticeably in the past week+. The OI decreased noticeably last year as well. https://www.curveadvisor.com/2016/12/rip-z-algo/

    I don’t think this is related to the turn, as the turn has been fairly stable for a while. But I think there is upside… https://seekingalpha.com/article/4134111-return-year-end-eurodollar-futures-turn

You must be logged in to reply to this topic.