Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
Adriano
@rsvp
CME has raised their initial margin on BTC from 35% to 47%, quite understandably given the recent extreme volatility.
Brian
@BlackArbsCEO
I just read the notebooks on gaussian mixtures and boltzmann portfolios. I found it very informative and plan to implement some of the techniques. I have a couple questions about it:
  1. How has the boltzmann portfolio approach, performed in the wild OOS?
  2. Are the notebooks for parts 3 and 4 created yet?
  3. Who(where) should I contact(post) if I have more questions about the underlying mathematics and its interpretation?
Adriano
@rsvp
@BlackArbsCEO hi Brian, Boltzmann portfolios quantify risk more accurately than the usual Gaussian-based approach, and they are designed not to overfit the noise in the covariance matrix -- thus they perform better out-of-sample. Part 1: https://git.io/boltz1 Part 2: https://git.io/boltz2 Part 3 is not yet public: it covers the hyperparametization for intertemporal usage. Part 4 on the mathematics may point to a forthcoming paper instead of a Jupyter notebook. Post your questions here, many are happy to help.
Adriano
@rsvp

The pre-Xmas decline of over 25% in Bitcoin price in less than 24 hours induced a backwardation in futures pricing relative to spot. Easy to imagine a trading strategy somewhat like a put, given the borrowing rate against such assets.

But is it hard to imagine a call option one-year out at $50,000 strike? Some institution paid close to a million dollars in premiums last Wednesday for notional 275 Bitcoins (source: LedgerX CEO Paul Chou). The usual Black-* models should avoided to price this option, for the stochastic process is extremely non-Gaussian (Levy).

Brian
@BlackArbsCEO
@rsvp how would you/your team like to be credited/attributed for the work that you have done? I plan to build on, test, and incorporate your boltzmann portfolio approach in future projects and want to make sure I give proper credit where it is due.
Adriano
@rsvp
@BlackArbsCEO hi Brian, that's terrific -- just mention the repository at https://git.io/fecon235 which cites some acknowledgements. Your blog posts are very interesting, so we will look forward to your insights. How did your backtesting for Boltzmann portfolios turn out? Hope you figured out how to adapt the framework to statistical arbitrage... Happy Holidays!
Adriano
@rsvp
2017-12-30T02:23:14Z CME BTC spread between H8 and F8, 13970-13745 at settlement implies annualized contango of +9.82% for Bitcoin. Volume on the H8 side was 64 contracts, equivalent to 320 Bitcoins notional.
Adriano
@rsvp
2018-01-06T19:12:39Z CME BTC spread between H8 and F8, 16780-16790 implies annualized contango of -0.36% for Bitcoin, i.e. backwardation, given price increase of 22% since last week. Open interest on the H8 side was 67 contracts. F8 contract continues trading until the third Friday this month.
Adriano
@rsvp

Fed Funds forecast one-year out is 1.82% using forefunds('18h', '19h') so this implies two 25 bp rate hikes expected. The first presser by the new Fed chairman will be most interesting to see.

[Please see https://git.io/fedfunds for forecasting the Fed Funds rate using futures contracts on LIBOR.]

Adriano
@rsvp
... Jerome Powell will succeed Yellen as chair in Feb 2018. The first FOMC press conference will be held March 21. Then June 13, Sep 26, and Dec 19, 2018.
joaquind
@joaquind_twitter
@rsvp I installed fecon235 following all the instructions, but get ImportError: attempted relative import with no known parent package. Any suggestions?
Adriano
@rsvp
@joaquind_twitter hi, usually that error results from starting work from a directory which contains the __init__.py file. So, for example, start exploring in the nb notebook directory which does not include that file.
Haisam
@Haisam
Hiee every one
Adriano
@rsvp
@Haisam thanks for checking in... how's the application of Holt-Winters/EMA going?
Adriano
@rsvp
2018-01-23T20:05:13Z CME BTC spread between H8 and F8, 10425-10355 implies annualized contango of 4.06% for Bitcoin. After the recent "crash" the backwardation has disappeared. Open interest on the H8 side was 138 contracts.
Adriano
@rsvp

pandas_datareader v0.6.0 Release

Some new connectors for equities data. @bashtage released this 3 hours ago -- tremendous work!
Adriano
@rsvp
Surprisingly how little attention the media gave to SPX (S&P 500 index) all-time record highs this month of January 2018. Perhaps because the climb was like a boring clockwork. gemrat() computes the mean geometric annualized rate at +3.97% (volatility of 20.17% with kurtosis at nearly 15), data starting at 2008-01-01. The street talks of possible "melt-up" -- how would a blow-off phase and its expected duration be characterized statistically?
Mangiafuoco
@Mangiafuoco
Got some error in fresh installation: No matter, from which directory is called. File "D:\xxxxxxx\fecon235-master\fecon235.py", line 46, in <module>
from .lib import yi_0sys as system
ImportError: attempted relative import with no known parent package
Adriano
@rsvp
@Mangiafuoco hi, usually that error results from starting work from a directory which contains the __init__.py file. So, for example, start exploring in the nb notebook directory which does not include that file.
Adriano
@rsvp
Didier Sornette posted the thesis of his ETH student Jialang XI on FX hedging by genetic algo.
Adriano
@rsvp
@omartinsky wrote a post (2013) on Black-Litterman model for portfolio reverse optimization which includes a Jupyter notebook: http://www.quantandfinancial.com/2013/08/black-litterman.html
Adriano
@rsvp
Finally polished... notebook https://git.io/infl gives an in-depth analysis of inflation, including a combined forecast using three orthogonal methods. Inflation rate likely to move up to 2.33% in a year. Comments?
Adriano
@rsvp
Add foreinfl() to forecast Unified Inflation. The best documentation for this function is https://git.io/infl which shows how it was derived by interacting with data and plots. This single function distills the forecasting process derived in the notebook. It's further discussed in new Appendix 2.
Adriano
@rsvp
CME BTC spread between M8 and H8, 8260-8235 implies annualized contango of 1.21% for Bitcoin. Open interest on the M8 side was 12 contracts, so the spread market appears very shallow.
Adriano
@rsvp
All eyes on new Fed Chairman Powell this Wednesday 2018-03-21 for his FOMC presser at 14:00 ET. Many expect a rate hike delivered in plain spoken terms, in a style similar to William Martin (longest running Fed Chairman: two decades).
Steven Weaver
@stevenweaver
Should we be preparing our brokerage accounts in any particular way for the event?
Adriano
@rsvp
@stevenweaver hi Steven, Powell is faced with a more energetic economy than Yellen, requiring more active rate hikes and a keen pulse on the statistics. He is a lawyer by training, not an economist (Bernanke and Yellen were top notch professors) -- so expect the FOMC language to change considerably. Wall Street is waiting to see how Powell will perform as events unfold.
Steven Weaver
@stevenweaver
@rsvp I’m looking forward to the FOMC notes as well.
Thanks.
Adriano
@rsvp
The FOMC minutes for this Wednesday will be out in two/three weeks, see https://www.federalreserve.gov/monetarypolicy/fomccalendars.htm
Adriano
@rsvp
>>> foreinfl()  # Inflation forecast
    _1  UTC 2018-04-09 16:59:04 
[2.1506, '2018-02-01', 1.5708, 2.7909, 2.0900]
Unified Inflation forecast in one function: Up to February 2018 data points, the geometric mean rate is 1.57%, the optimized Holt-Winter method expects 2.79% over the next year, whereas the 10-y BEI break-even rate using Treasuries is 2.09%. Thus, averaging them out: 2.15% is our near-term forecast. Notebook https://git.io/infl gives an in-depth analysis of inflation.
Adriano
@rsvp
Fed Funds forecast one-year out is 1.99% using forefunds('18m', '19m') so this implies barely two more rate hikes of 25 bp. Definitely not three more. Yellen's academic interest specialized in the labor market, but Fed Chairman Powell will probably have to rely on his staff to assess current unemployment which is less than what many consider its natural rate.
Adriano
@rsvp
Completed spin-off of fecon235 source code to fecon236 -- more info at https://git.io/econ -- your review and feedback would be much appreciated!
Adriano
@rsvp

New notebook for fecon236 >= 10.7.1

Bootstrapping has two benefits: small-sample statistics and simulation from controlled population. We simulate alternate histories: to visualize sample price paths, and for estimating probabilities of events such as investment loss. Bootstrapping facilitates study of small-sample behaviour for which asymptotic statistical theory is unsuitable, or where closed-form mathematical analysis is intractable, for example, geovolatility which is the volatility of the geometric mean rate. Shortcut to notebook: https://git.io/bootspx
EconDB
@econdb
Hey there, for many of you this will be the first time that you hear of econdb.com, which is a website with economic indicators scraped regularly from national agencies. It's open to suggestions on new data and features. Please, ping me if you would like to see something done or you simply have questions about how it works. Cheers, Oriol.
Adriano
@rsvp
@econdb hi Oriol, wondering what additional data your site offers beyond that of FRED and Quandl? Is there a list of unique time-series data we can see?
EconDB
@econdb
Hey Adriano, first of all I think Quandl and FRED are great and beat our website in many aspects. The core distinction of Econdb is that we track a wider set of macroeconomic indicators from country official agencies, without having to rely on IMF or OECD, which publish the data usually with a lag of 3 to 6 months since the official release. Here's the full list of sources: econdb.com/source. We aim at incorporating agency forecasts, flash estimates, etc. in due course, and it would be great to hear feedback to help us prioritize. Cheers.
Adriano
@rsvp

@econdb hi Oriol, yes, getting economic data outside the US is frustrating due to time lags and incompatible data formats. So your service has a great niche! https://www.econdb.com/source (The site does not mention costs, so please kindly clarify.) Nice to see the databases of many central banks included for monetary data -- though BOJ is missing, definitely a priority item.

We'll trial your Python API https://github.com/inquirim/inquisitor over at https://github.com/MathSci/fecon236 when we redo the notebooks on the Europe region. Thank you very much!

Adriano
@rsvp

FOMC presser will be held Sept 26. The current Fed Funds range is 1.75 to 2.00%.

Our Fed Funds forecast one-year out is 2.42% using fe.forefunds('18z', '19z') so this implies two rate hikes of 25 bp each during that period [see https://git.io/fedfunds to forecast the Fed Funds rate using futures contracts on LIBOR]. Updated fecon236 function

Adriano
@rsvp
Kaggle competition: Can we use the content of news analytics to predict stock price performance?, sponsored by Two Sigma, has $100,000 prize! NLP is one of the core areas which brought success to RenTech. Keep us posted on your new developments.
Adriano
@rsvp
... further re Kaggle and the issue of patents on algos: https://www.kaggle.com/c/two-sigma-financial-news/discussion/66828
Adriano
@rsvp

Visualizing possible EU crisis due to Italy's budget deficits

They are currently in excess of EU limits, and thus the yield spread between the 10-year government bonds of Italy and Germany, BTP-Bunds, has been increasing. One way to monitor the situation is through the shorts on BTP futures: 5BTS chart, last quoted at 55.25.

5BTS is a fully collateralised, UCITS eligible Exchange-Traded Product. The ETP provides five times the inverse daily performance of the Long Term BTP Rolling Future Index, which tracks front-month Long-Term Euro-BTP futures, plus the interest revenue earned on the collateralised amount. Long-Term Euro-BTP futures are traded on EUREX and deliver Italian government bonds with 8.5-11 years to maturity. For example, if the index rises by 1% over a day, then the ETP will fall by 5%, and if the index falls by 1% over a day, then the ETP will rise by 5%, excluding fees and interest revenue.

Adriano
@rsvp

Machine Learning in Investing, interview with Jeremiah Lowin: http://investorfieldguide.com/machinelearning

For related topics and details, follow sub-Quoras: https://optimal.quora.com for the financial aspects, and https://compute.quora.com for the ML computational aspects.

Adriano
@rsvp
Elevated default risk across bond markets is indicative of a weak economy. But how can a policy maker calibrate credit spreads to assess changes in interest rates? We consider mortgage and corporate
credit spreads to construct a robust Unified Credit Profile (a tutorial on MAD, Median Absolute Deviation, in rescaling non-Gaussian time-series), see https://git.io/creditprof
Adriano
@rsvp
It seems that Nasdaq take-over of Quandl has shut-off access to the freely available data: see MathSci/fecon236#7
EconDB
@econdb
@rsvp CME might be a problem on its own. CME and NASDAQ are competitors, so the conditions by which the CME was licensed to Quandl might not apply anymore. Feel free to send any data requests we might add on econdb.com. Unfortunately I haven't been granted permission to distribute CME.
Adriano
@rsvp
@econdb Updated MathSci/fecon236#7 with statement from Quandl CEO, and I also narrowed the disruption dates of specific CME data. Not all freely available data has disappeared, but the loss of futures data would be substantial. I tend to agree with your suspected cause, and will get definitive answers this week. Thanks!
Adriano
@rsvp

Is credit risk signaling a weak economy? Nah.

>>> cprof = fe.creditprof()
>>> fe.tail(cprof)
2018-12-04  0.706186
2018-12-05  0.685567
2018-12-06  0.806503
2018-12-07  0.805115
2018-12-10  0.732950
2018-12-11  0.509615
2018-12-12  0.366673

The Fed can start worrying when that measure zooms past +2.0. Their next presser is Dec 19, by the way, expecting a 25 bp rate hike.

ambituous
@ambituous

Hey Gang! I came across an interesting debate going lately on the overall general consensus issues and felt the devs here would be some of the best to shed some light on.

Kyle Samani from Multicoin and Will Martino from Kadena are hosting an AMA - Town Hall style on the next wednesday (20th) over at Discord - inviting specifically blockchain devs and high level players in the space to drop questions on the issue of:

Proof of Work vs Proof of Stake. Pivoting around New Findings, Perspectives & Tradeoffs

Proof: https://twitter.com/kadena_io/status/1104102013355335680
Discord: https://discord.io/kadena

Guillaume Chevalier
@guillaume-chevalier