Rather than posting a new topic every time, may as well just post papers and links here. Please keep it to concrete strategy ideas, the more explicit the better, and preferably those that could be implemented in Quantopian!

Profitable Mean Reversion after Large Price Drops: A Story of Day and Night in the S&P 500, 400 Mid Cap and 600 Small Cap Indices

235 responses

Good idea and nice paper.

The VIX Futures Basis: Evidence and Trading Strategies should be able to implement something similar using the VIX short-term futures vs medium-term-futures ETFs, and pulling in the VIX prices with fetcher

Quite a few papers in Turnkey's alpha DB: https://alpha.turnkeyanalyst.com/ideas

EDIT: corrected link, h/t Dennis C

I've never seen that Turnkey Alpha site before. What is it?

Yes I did, my mistake. It's a free sign-in. They are one half of the guys who wrote Quantitative Value (the other half is Empiritrage), and they regularly write little papers about the sort of exploitable opportunities people here might be interested in. I am not affiliated with them.

This blog post on Limited Attention and the Earnings Announcement looks interesting (login not required).

Hello guys,

Not sure if you've seen it...if not:

On-Line Portfolio Selection with Moving Average Reversion

If you search the Quantopian Community forum with the keyword "OLMAR" you'll find several threads with Quantopian implementations. I also have some code that I can share.

The author's website: http://www.cais.ntu.edu.sg/~libin/.

Grant

http://gekkoquant.com/

Tons of strategies here to try out.

Hello Simon,

Thanks...would you be willing to provide some specific recommendations from the list above? What are the top 3 you'd recommend reading through carefully, that could be coded in Quantopian (without a heroic effort)?

Grant

Sorry, I haven't read them all! Whatever suits your temperament I guess!

Hi Grant,
I think good old trend following is always fun. It's very practical. In case you haven't checked it out, I noticed that Claus Herther has a great starting point. I'd like to add in measurement of the slope of a trend, momentum, and williams to help add some "trend anticipation" into a standard trend following system.

Just stumbled upon this goldmine of hundreds of papers, most with pdf links, on a variety of topics:
http://www.finance.martinsewell.com/

Also, the Kaggle tutorials / free Books are well worth a look:
https://www.kaggle.com/wiki/Tutorials

http://www.priceactionlab.com/Blog/2013/09/psi-the-probability-state-indicator/

Note he doesn't actually give the formula for this indicator, so one would have to do some work to try and figure out what he's talking about...

Hi everyone, is ist possible to program the black litterman approach with Quantopian? Tips are highly welcom. Thanks in advance for your help.

It should be possible, someone wrote a minimum variance portfolio re-balancing algorithm a few months ago. You'd need to use fetcher to get your index weights for your prior, make sure to fetch them "as-of" the date you are at in the back-test. Then you "just" need to do all the bayesian matrix manipulations, along with your input market views/shades, come up with the target weights, then submit orders to move from your current portfolio to your target portfolio.

It would be an excellent demonstration and example, perhaps you can get the quantopian folks to code it up!

Thank Simon for your comment. I wrote my last thesis about BL so I have the theoretical background. But to be honest with you, I am not quite good in programming. Nevertheless I will try and let the community know.

Cheers Grant. You made my day. I would appreciate more of articles like that.

Hello Fabian,

Thomas Wiecki posted the article first on https://www.quantopian.com/posts/interesting-papers. I just copied the link here. If you have comments on the article, I suggest posting them to Thomas' thread.

Grant

Mebane Faber has a few interesting papers at Cambria Investments' website http://www.cambriainvestments.com/research/, especially one of Relative Strength strategies http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1585517.

Several of Mebane's systems were implemented on quantopian six to twelve months ago, global TAA, relative value, relative value + TAA. I also wrote some picloud+zipline brute force optimization of the TAA model. If you search for Mebane you should find them. I don't know if they still work in the backtester.

The gist: implement a market-neutral high vs low momentum strategy, but trim the shorts as the market drops. This will, of course, add a strong long-term long-biased mean reversion factor to the system.

http://www.plosone.org/article/info:doi/10.1371/journal.pone.0067721

Being on the Field When the Game Is Still Under Way. The Financial Press and Stock Markets in Times of Crisis

This paper looks at the relationship between negative news and stock markets in times of global crisis, such as the 2008/2009 period. We analysed one year of front page banner headlines of three financial newspapers, the Wall Street Journal, Financial Times, and Il Sole24ore to examine the influence of bad news both on stock market volatility and dynamic correlation. Our results show that the press and markets influenced each other in generating market volatility and in particular, that the Wall Street Journal had a crucial effect both on the volatility and correlation between the US and foreign markets. We also found significant differences between newspapers in their interpretation of the crisis, with the Financial Times being significantly pessimistic even in phases of low market volatility. Our results confirm the reflexive nature of stock markets. When the situation is uncertain and unpredictable, market behaviour may even reflect qualitative, big picture, and subjective information such as streamers in a newspaper, whose economic and informative value is questionable.

Disclaimer

The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.

In response to Simon's post of Profitable Mean Reversion after Large Price Drops: A Story of Day and Night in the S&P 500, 400 Mid Cap and 600 Small Cap Indices, has anyone coded an algo that replicates the strategy outlined in this paper that they wouldn't mind sharing? There is a clear and consistent dropoff in return as years progress from 2000 toward 2010, and I'm curious to see if this trend has continued in the three years since.

Bitcoin arbitrage.

https://github.com/maxme/bitcoin-arbitrage

I've noticed that the many cryptocurrency exchanges out there have a significant spread. The spread between Mt. Gox and BTC-e, for example, is typically \$100, and can go even higher if Mt. Gox has a surge. That's not even getting into the opportunities for arbitrage trading BTC to LTC (litecoin) and other cryptocurrencies that largely follow the BTC market trends. Personally I'm fascinated by it.

I found this overview of quant investing by Max Dama http://www.decal.org/file/2945.
At page 16 he very briefly explains a possible trading idea through the exploitation of the "first day of the month concept".
His description:

"The First Day of the Month. Its probably the most important trading day of the month, as inflows come in from 401(k) plans, IRAs, etc. and mutual fund have to go out there and put this new money into stocks."

"Over the past 16 years, buying the close on SPY (the S&P 500 ETF) on the last day of the month and selling one day later would result in a successful trade 63% of the time with an average return of 0.37% (as opposed
to 0.03% and a 50%-50% success rate if you buy any random day during this period)."

"Various conditions take place that improve this result significantly . For instance, one time I was visiting Victors office on the first day of a month and one of his traders showed me a system and said, If you show this to anyone we will have to kill you.
Basically, the system was: If the last half of the last day of the month was negative and the first half of the
first day of the next month was negative, buy at 11a.m. and hold for the rest of the day.
This is an ATM machine
the trader told me. I leave it to the reader to test this system.""

So e.g. if at 31th of march at 12:am the choosen equity has a negative return for the day and the day after it has a negative return until 11 a.m.
then buy and hold until close.

I tried this using excel and intraday data I got from a russian website giving away free historical prices for the 40 most traded stocks in the US, but obviously
quantopia is a much better way of trying this simple strategy.

The few stocks that actually had this pattern of negative-negative->buy-hold until close showed a small positive gain.
I didn't calculate the sharpe ratio, but my thinking is that if the sharpe ratio is high and you do this 12 mths a year and use a healthy amount of leverage
you can make a nice stat arb payoff.

I'm a novice to coding so I haven't made an attempt yet at coding this, so if any of u guys who are fast at this feel free to try it and post a backtest.

Cheers/ Patrick

Applying Deep Learning to Enhance Momentum Trading Strategies in Stocks

This one looks particularly easy to implement in Quantopian, since it's basically just technical analysis.

http://jonathankinlay.com/index.php/2014/05/implementation-of-a-scalping-strategy/

Looks promising, but probably requires a tick-level backtester/microstructure simulator

http://m.seekingalpha.com/instablog/709762-varan/2990923-naive-graham-passive-investing-according-to-the-master

There's a followup or two, and this feels like a repeat but just in case

http://www.citeulike.org/user/lehalle/

http://portal.idc.ac.il/en/main/research/caesareacenter/annualsummit/documents/conference%20may%202014/do%20etfs%20increase%20stock%20volatility.pdf

Do ETFs Increase Volatility?

We study whether exchange traded funds (ETFs)—an asset of increasing importance—impact the volatility of their underlying stocks. Using identification strategies based on the mechanical variation in ETF ownership, we present evidence that stocks owned by ETFs exhibit significantly higher intraday and daily volatility. We estimate that an increase of one standard deviation in ETF ownership is associated with an increase of 16% in daily stock volatility. The driving channel appears to be arbitrage activity between ETFs and the underlying stocks. Consistent with this view, the effects are stronger for stocks with lower bid-ask spread and lending fees. Finally, the evidence that ETF ownership increases stock turnover suggests that ETF arbitrage adds a new layer of trading to the underlying securities.

Wow excellent, I had not seen this paper. Classic!

I found this pretty interesting, seems relevant.

http://quantpapers.com/

http://www.lse.ac.uk/fmg/events/capitalMarket/pdf/CMW-ST-2014-Vissing-jorgensencycle_paper_24Jun2014.pdf

Being in/out of the market on certain weeks according to the FOMC meeting calendar. Looks promising, and simple for someone to implement!

Man is it ever hard to find this thread every time, searching doesn't work well. Anyway, not a strategy per se, but a great paper on the VIX ETPs:

http://www.diva-portal.org/smash/get/diva2:742887/FULLTEXT01.pdf

EDIT: I was wrong, there is a trading strategy in the second half!

Hello Simon,

Is there anything in this thread that would be particularly interesting to code in Quantopian and backtest?

Grant

Grant,
I just came across this, Critical Line Algorithm for Portfolio Optimization, it includes a Python implementation. I would check out quantpapers.com, there's hundreds of papers on there.

Dave

Grant, I think that's really a personal question, what sort of trading strategy does someone want to deploy, and how does it fit in with their existing trading strategies? For purely academic interest, I am not sure I would be doing quant trading :)

@Grant, Simon. Anything dealing with Vix, Vix term structure, Vix etn’s would be of much interest

Well, let me put the question another way. Have any of the ideas listed in this thread been launched as paper/live trading algos at IB? If so, what has been the result? --Grant

Darell: http://volatilitymadesimple.com/ follows a dozen or so VIX ETP strategies, and their own one of course.

Grant: sorry, I haven't done any work in Quantopian for about a year. Can't speak for others.

Hello all, can anyone point me in the direction of an end of day / swing system for the S&P or Dow or Nasdaq? Something with a good win loss ratio would be ideal.
I would appreciate it.
Thanks.
Robin

Anyone know if you can import Futures data?
"comparing first and second month VIX futures. Traders often use this simple approach to determine whether the VIX futures term-structure is in contango (favoring XIV) or backwardation (favoring VXX)"

Mainly, if we can import front and back month VIX futures to initiate positions on XIV and VXX respectively?

@Sam, I don't know about getting the data from volatility made simple, but you can use Quandl to import the data, or get it directly from CBOE.
http://cfe.cboe.com/data/historicaldata.aspx https://www.quandl.com/c/futures/cboefe-s-and-p-500-vix-futures

Update: You can also get the daily composition of front/back month holdings of the ipath ETNs on their website, that might help you refine your strategy a bit more too. I believe they have the historical holdings as well. This link is for VXX, the others are available as well though. http://www.ipathetn.com/US/16/en/details.app?instrumentId=259118

Their concept of "Dual Momemtum" is very intriguing. As well, extending it in the manner which is described here:

Disclaimer

The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.

Campbell Harvey's website is also a useful site for financial glossaries and papers on risk. http://people.duke.edu/~charvey/

http://jonathankinlay.com/index.php/2015/03/developing-etf-longshort-strategies/

It's not clear if this is a mean-reversion strategy on this cointegrated basket, or whether it's a static investment portfolio somehow optimized for low variance.

Simon - have you looked through the "premium" offerings on Quantpedia at all? Am curious whether they are worth the fee or not

I haven't, no, I was just planning on going through their free stuff to see what anomalies and papers look interesting and suitable.

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1664823

I really love Tony Cooper's papers, so clear and readable.

Identifying small mean reverting portfolios:

Awesome resources.

A Simple Way to Estimate Bid-Ask Spreads from Daily High and Low Prices

Great piece with lots of implementable ideas on co-integration, High frequency implementation etc
Statistical Arbitrage and High-Frequency Data With An Application to Euro Stoxx 50 Equities.

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2432061

Simon,

Thanks for the article! It's a good read.

Michael

Not sure where else to put this. I won't classify this as strategy but it's good to know the bid/ask spread % of companies. Useful for HF algo development.

Detail Liquidity, Spread % of Large Cap US companies

Statistical Summary of Lap Cap US companies - Liquidity and Spread

Jonathan Kinlay writes some of the best stuff out there, thanks!

Totally agree with you Simon on Jonathan Kinlay.

3rd Annual RavenPack Research Symposium - "BIG DATA ANALYTICS FOR ALPHA, SMART BETA & RISK MANAGEMENT" Covers many aspects - Big Data, Smart Beta, News Analytics, Alpha Generation, Machine Learning-Based Trading Strategy etc.
Videos and PDFs are available.

Statistical Arbitrage and Algorithmic Trading : Overview and Applications

Amazing overview of the mathematics available to design quantitative strategies

Matthieu, that looks like a great resource indeed. The link seems to have changed, here is an updated one:
http://e-spacio.uned.es:8080/fedora/get/tesisuned:CiencEcoEmp-Mnoguer/Documento.pdf

Folks, whilst all these seem to be great resources, they need a certain amount of knowledge in Statistics. What is the base amount of statistical knowledge from where one can kick on? Any books or resources for the uninitiated?

Preliminiaries are (I think) basic single and multivariable analysis (maybe some real analysis and intermediate combinatorics), linear algebra then get into basic probability theory and after that statistical inference, stochastic processes (and simulation) and econometrics and after that look at financial mathematics and optimization theory and stochastic partial differential equations. Just Google or go to Amazon etc to find books (with solutions).

MS M,

I pretty much agree with the order of Patrick.

You can grab the basics on probabilities and statistics on www.statlect.com

Then you can follow the good introduction machine learning class from Andrew Ng on Coursera

If you want to move to more advanced understanding of learning algorithms you may want to have a look at The Elements of Statistical Learning

After that (and maybe some stochastic calculus and time series analysis) you should be able to understand most of the articles you are interest in or at least know what to Google to fill the gap

Market neutral portfolio construction with excel implementation.

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2600014

Expected skewness and momentum portfolios. Some bonkers-good results in there.

Portfolio Optimization for strategies using sort information on expected returns.
http://www.courant.nyu.edu/~almgren/papers/sort.pdf

Edit: Subbed to this twitter feed a long time ago and rediscovered it today. They post quant papers from SSRN

Check this out! Most of the papers have been mentioned by you guys above.

A couple of good tutorial style resources I found recently:
* "AHL explains", a couple of videos going over key concepts like momentum trading: https://www.man.com/DE/ahl-explains Would be cool to implement them in Quantopian (although we don't have futures yet).
* "Developing & Backtesting Systematic Trading Strategies": https://r-forge.r-project.org/scm/viewvc.php/*checkout*/pkg/quantstrat/sandbox/backtest_musings/strat_dev_process.pdf?root=blotter
* "Mean-Reversion and Optimization" http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2478345

There are lot of papers and detailed litratures in this link.
I ust came across this 130 30 stretagy thought it would be good place to
post.

Interesting article about ETF liquidity and the liquidity of underlying securities:

http://blog.acolyer.org/2015/08/26/mining-high-speed-data-streams/

Haven't read the paper yet, but seems to have promising applications to trading.

http://www.zerohedge.com/news/2015-08-26/was-mondays-etf-collapse-just-warmup

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2417974

This is a paper by Michael Gayed, CFA and Charlie Bilello, CMT that visits the idea of beta rotation.
The paper was a 2014 Dow Award Winner

A bunch of articles/papers written by Cliff Asness of AQR. Pretty interesting.

There's already some posts about end of the month stock behavior above, but here's a detailed paper about it:

From the intro: "we find that since July 1926, one could have held the US value-weighted stock index (CRSP) for only seven
days a month and pocketed the entire market excess return with nearly fifty percent lower volatility
compared to a buy and hold strategy."

The equity curve graph on page 22 of the paper is eye opening.

http://mintegration.eu/2015/09/14/risk-budgeted-portfolios-with-an-evolutionary-algorithm/

A form "risk parity" using Differential Evolution to optimize portfolio contributions to risk.

Another D'Aspremont paper.

Another pair trading algorithm using 2-stage correlation and co-integration based approach on 15 minute OHLC intra-day data on oil sector stocks. They claim monthly 2.67 Sharpe ratio and an annual 9.25 Sharpe ratio for the period between 2012-13. Will be interesting to see if this can be replicated in Quantopian.

Claims that acceleration (difference of returns) has more explanatory power than simple momentum.

Not a 'trading strategy' per se, but an interesting site with some python related code, and some clear thinking.
http://www.turingfinance.com/hacking-the-random-walk-hypothesis/

Hey all. I'm at academic finance conferences quite often these days as part of our academic outreach. I see a lot of interesting papers and would be happy to make some best-of lists the next time I'm at a conference. Would people be interested in lists like this for potential ideas?

Disclaimer

The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.

That sounds great Delaney, I'd definitely be interested.

Great, I'm at FMA in Orlando next week. I'll start up a forum thread and post live once I'm there.

I hope I'm not being presumptuous but I think everyone following this thread is interested.

Another statistical arbitrage paper but using step-wise regression and variance ratio tests to identify co-integrated baskets. Paper claims a sharpe of 7+ with 50 basis points transaction costs. Quite old paper though.

I've read it (Mean reversion after price drops) multiple times because I'm testing some Josef Rudy's research for my thesis to see if his findings hold water.

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2674086

Not implementable in Quantopian yet, but perhaps soon... ? :)

Looks interesting! Thank you, Simon.

Delaney that would be fantastic. I've been working on converting the ideas from this paper into Python code
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2600798

I'll be adding papers over the next few days.

Marco Avellaneda & Stanley J. Zhang Plata startegy: Statistical arbitrage on ETF [https://www.math.nyu.edu/faculty/avellane/Lecture8Risk2011.pdf]

http://acquirersmultiple.com/2015/10/negative-enterprise-value-a-primer/

Simple idea - buy negative EV stocks and hold for a year! In the microcap segment, it allegedly has mean returns of 60% per trade over the holding period (one year). Probably a wicked drawdown though.

Stocks on Thursdays and Bonds on Fridays

A Simple Learning System

A simple learning system. Good for learning about market behavior and over-fitting.

http://henrycarstens.com/simple-learning-system/

The idea is to simulate the composition logic of any ETF/Index stock picks and invest in stcock to be added/deleted from it. Keeping it from the announcement date till 14 days later (when the actual action is done) will result positive retunrs for going long on added stocks and short on deleted ones. The idea is that once stock is announced to be added/deleted to an index , then the index must buy/sell it around 14 days after and the market reacts. buying it before, and sell it at the end of the 14 days announcement.

There are plenty of ETF's so lots of arbitrage is available.

This is the S&P composition logic as example

if someone did something or wants to work on it together.

This paper has a collection of strategies that may be helpful. Looking through the list and although some are simple there are several that look interesting...

From the Abstract:
We present explicit formulas - that are also computer code - for 101 real-life quantitative trading alphas. Their average holding period approximately ranges 0.6-6.4 days. The average pair-wise correlation of these alphas is low, 15.9%. The returns are strongly correlated with volatility, but have no significant dependence on turnover, directly confirming an earlier result by two of us based on a more indirect empirical analysis. We further find empirically that turnover has poor explanatory power for alpha correlations.

Could you please tell me what does 'alpha' mean?
For example, there is simple mean-reversion alpha −ln(today􏰑s open / yesterday􏰑s close)
Or it is just useful signal (=feature) for learning algorithm?

Hi Philipp,

Alpha is a commonly used metric of how much new information is contained in another signal. It is found by performing a linear regression between the return stream generated by the new signal, and existing factors such as the market. The equation might look like this.

R_new = alpha + beta * R_market + beta * R_oil + ...

By seeing how much of your returns are historically explained by each of the other factors, you can make an estimate for how much of your returns are coming from new information, which is what is left over in the alpha. For more info on this see lectures 4, 13, and 14 in the Quantopian Lecture Series.

Thanks,
Delaney

https://www.spcapitaliq.com/our-thinking/resources-ideas/Exploring%20Alpha%20from%20the%20Securities%20Lending%20Market.pdf

Not a new one but have been digging in to short-related data lately and found this interesting.

Not really a paper but this is an excellent quandl post on the general process to test trading ideas:

https://www.quandl.com/blog/interview-with-a-quant-part-one

Entropy theory of mind. Numerically derives the link between Entropy in physics and finance. Also builds a quantitative model framework that blends entropy, value of judgement/bias, trading decisions and volume. The only paper I've read that models market volume in a somewhat intuitive way.

Forecasting Volatility in the S&P500 Index

The link to the PDF is in the first paragraph. Written by Jonathan Kinlay, he lays out the framework for the ARFIMA-GARCH method of volatility estimation and comes to the conclusion that traditional Option Pricing by Black-Scholes is inefficient and proves it by testing a simple options strategy based on the results of his volatility forecasts.

A few studies of mine these models actually traded real money for a long time like 20 years, not hypothetically

Here is the link to Li-Xin Wang latest paper Modeling Stock Price Dynamics with Fuzzy Opinion Networks .pdf

Built to illustrate the idea of trading standard deviation, here is the link to a simple Crude Oil strategy with a z of 1.5.

Built to illustrate the ideas of trading relationships, fundamentals, yesterday and seasonals, here is the link to a second simple Crude Oil strategy. This one has a z of 2.2.

Built to illustrate the ideas of trading a seasonal, trading volatility, and trading yesterday, here is the link to a third simple Crude Oil strategy. This one has a z of 2.3.

Built to illustrate the ideas of trading the tails of a candlestick and trading volatility, here is the link to a fourth simple Crude Oil strategy. This one has a z of 3.0.

Built to illustrate the ideas of portfolios of systems and reusing systems, here is the link to the portfolio of the four previously described Crude Oil strategies. The portfolio has an annual return of 13.6%, a max drawdown of 9.2% and a Sharpe of 1.4 from years 2006 thru February 2016.

Built to show the idea of trading the tail of a candlestick instead of the body when volatility leaves a big tail after the natural gas supply report on Wednesday, here is the link to the first simple Natural Gas strategy. This strategy has a z of 2.8.

Built to illustrate the ideas of trading other traders and trading a fundamental, this Natural Gas system trades the positioning prior to the Wednesday supply report. Here is the link. This strategy has a z of 1.8.

does this site have a vocabulary section. like what is a z score

A z-score is a statistics term, it measures how many standard deviations a value is from the mean of a set of values. Z = 0 means same as the mean, Z = 1 means the value is 1 standard deviation above the mean, etc.

I promise I'm not trying to be snarky, but you can learn that yourself in about 3 seconds by searching "z score" in Google. That will probably be true for most of the finance and statistics terms you see here. Some of them will be complex (like how a GARCH process works) but most will not.

It's normally just the (innovation - mean) / standard deviation, but I think Henry has made up his own definition, I am not sure what he is referring to.

z score is the statistical significance of the test/system. Greater than 1.6 means roughly 95% chance results aren't random.

Thank you both. i conclude that the z score is a way of quantifying the quality of a back test so you can know if you do the same thing by flipping a coin (or not). Sorry i have to reduce everything to some oversimplified format.

I used to trade a a local on the NYFE and now live in Colombia S America. Medellin to be exact. I own a coffee farm called Finca Milena and will put you up if you come down here and get me caught up on quants, algos, thoery etc etc. By the way Mat I did google z score and it came up as a theory for quantifying a company´s future chances of filing for bankruptcy and no offense taken. I wonder what the z score is for that algo????

Fair point - a guy named Edward Altman didn't really do anyone any favors when he also named his bankruptcy prediction model the "z score".

@William, Here is a simple example of zscore of an asset, others will comment if its wrong in any way

9
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
 Returns 1 Month 3 Month 6 Month 12 Month
 Alpha 1 Month 3 Month 6 Month 12 Month
 Beta 1 Month 3 Month 6 Month 12 Month
 Sharpe 1 Month 3 Month 6 Month 12 Month
 Sortino 1 Month 3 Month 6 Month 12 Month
 Volatility 1 Month 3 Month 6 Month 12 Month
 Max Drawdown 1 Month 3 Month 6 Month 12 Month
import talib as ta
import pandas as pd

def initialize(context):
set_benchmark(symbol('SPY'))
context.spy = symbol('SPY')

def zscore(series, value):
mean = ta.SMA(series, value)
std = pd.rolling_std(series, value)
zscore = (series - std) / mean
return zscore.iloc[-1]

def handle_data(context, data):
hclose = history(200, "1d", "close_price")

spy_zscore = zscore(hclose[context.spy], 64)

record(
SpyZscore = spy_zscore
)


There was a runtime error.

i assume from this that you would want to see a z score of 1 or better to conclude that the system is better than just any random approach. i.e. coin toss

Darell,

You should change

zscore = (series - std) / mean


to
 zscore = (series - mean) / std 

Backtest of Darrell's z-score algo w/ z = (series-mean)/std dev

16
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
 Returns 1 Month 3 Month 6 Month 12 Month
 Alpha 1 Month 3 Month 6 Month 12 Month
 Beta 1 Month 3 Month 6 Month 12 Month
 Sharpe 1 Month 3 Month 6 Month 12 Month
 Sortino 1 Month 3 Month 6 Month 12 Month
 Volatility 1 Month 3 Month 6 Month 12 Month
 Max Drawdown 1 Month 3 Month 6 Month 12 Month
import talib as ta
import pandas as pd

def initialize(context):
set_benchmark(symbol('SPY'))
context.spy = symbol('SPY')

def zscore(series, value):
mean = ta.SMA(series, value)
std = pd.rolling_std(series, value)
zscore = (series - mean) / std
return zscore.iloc[-1]

def handle_data(context, data):
hclose = history(200, "1d", "close_price")

spy_zscore = zscore(hclose[context.spy], 64)

record(
SpyZscore = spy_zscore
)


There was a runtime error.

This third Natural Gas system illustrates the ideas of trading relationships, trading change and trading rate of change between Natural Gas and Crude Oil. This strategy has a z of 2.2. Rules and results are Here.

http://alta5.com/blog/bittman/

It's not clear how dependent this strategy is on the recent regime.

We've actually already done a bunch of work implementing the paper you posted, Pravin. Figured linking to it might be useful to some folks.

We've actually already done a bunch of work implementing the paper you posted, Pravin. Figured linking to it might be useful to some folks.

http://blog.factorwave.com/options-and-earnings

Can't be traded with Quantopian, but looks legit.

Looking forward to the actual talk, to find out what the method is! :) (Marcos Lopez de Prado of Guggenheim Partners at Global Derivatives 2016)

Thanks for sharing, Simon.

Dr. Lopez del Prado's website is here.

Knowing de Prado's stuff, which is very good, he'll be making the point that mean variance analysis doesn't work in practice any more. It's easy to overfit it to some historical period by naively optimizing, but will have little correlation to out of sample performance. This is similar to Thomas Wiecki's recent paper on how sharpe ratio also has no correlation between in and out of sample performance.

I am curious what his suggestion/replacement is. Bootstrapping works great for avoiding overfitting, but you end up with pretty average portfolios.

I suspect just not using mean-variance and using other more sophisticated portfolio selection techniques. Correlation reduction filters, sector neutrality filters, etc.

The book, Systematic Trading, by Robert Carver, was recommended to me by Simon, and I just finished an entire chapter dedicated to over-fitting. There is a quantitative discussion of relevant backtest time scales to distinguish one approach from another. And approaches to avoiding fitting to a single historical period. Etc. Flipping ahead in the book, bootstrapping is covered, as well. The author seems to be very sober and realistic and is not promoting particular strategies, per se (althoug he does distinguish styles of trading). The focus is on the process and the pitfalls. It is very approachable from a technical standpoint. No fancy math/statistics. It might be a good starting point for many Quantopian users who are aspiring quants.

Agree with you on that book Grant. Must say there are parts that I have difficulty getting my head around. A practitioner's book. His blog is excellent as well.

@Vladimir, I was trying to understand how the z-score can be applied to the simple XLP+TLT portfolio algo you posted elsewhere. Would you be able to add the z-score code to it and repost here?

Also, if we are looking for a z-score of >1.6, what are we looking for? That the z-score curve stays above 1.6 most of the time? Or something different? Thanks in advance..

@rb rb, z-score is really just a measurement of how "rare" an event is in terms of it's distribution. So if you have a z-score >1.6 it would mean that it has a roughly 5% chance of occurring, so a relatively "rare" event indeed (for those who are not old enough to have used this in math class #throwback, the z-table is a great way to illustrate a z-score for normal distribution. In this case, this is a positive table so one would do 0.50 - p(z = 1.6) = 0.50 - 0.4452 = 0.0548 http://access-excel.tips/wp-content/uploads/2015/09/z-score-02.png).

Applied to any trading strategy, z-scores are a common way to assign a statistical probability value of something occurring, which can act as a "confidence" interval. Using Henry Casten's quick z-score example from above, the attached is an algorithm that shorts SPY when the z-score > 1.6 and long when z-score < -1.6, and closes out positions when -1.4<zscore<1.4, based on the assumption that it is "rare" event and SPY will revert to it's mean price over time.

19
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
 Returns 1 Month 3 Month 6 Month 12 Month
 Alpha 1 Month 3 Month 6 Month 12 Month
 Beta 1 Month 3 Month 6 Month 12 Month
 Sharpe 1 Month 3 Month 6 Month 12 Month
 Sortino 1 Month 3 Month 6 Month 12 Month
 Volatility 1 Month 3 Month 6 Month 12 Month
 Max Drawdown 1 Month 3 Month 6 Month 12 Month
import talib as ta
import pandas as pd

def initialize(context):
set_benchmark(symbol('SPY'))
context.spy = symbol('SPY')

def zscore(series, value):
mean = ta.SMA(series, value)
std = pd.rolling_std(series, value)
zscore = (series - mean) / std
return zscore.iloc[-1]

def handle_data(context, data):
hclose = history(200, "1d", "close_price")

spy_zscore = zscore(hclose[context.spy], 64)

print(symbol('SPY') in context.portfolio.positions)
if symbol('SPY') in context.portfolio.positions:
print(context.portfolio.positions[symbol('SPY')].amount)

if spy_zscore > 1.6 and not(symbol('SPY') in context.portfolio.positions):
order_target_percent(symbol('SPY'),-1.0,None)

elif spy_zscore < -1.6 and not(symbol('SPY') in context.portfolio.positions):
order_target_percent(symbol('SPY'),1.0,None)

elif spy_zscore > -1.4 and spy_zscore < 1.4 and (symbol('SPY') in context.portfolio.positions):
order_target_percent(symbol('SPY'),0.0,None)
print('sell positions')

record(
SpyZscore = spy_zscore
)


There was a runtime error.

Z scores can only be interpreted as a measure of event rarity when the underlying distribution of data is known. In most cases distributions in finance are not normally behaved, so assuming normality will not be a good estimator of the rarity of an event. It is better often to think of a z score as a measure of extremity, and only convert to actual rarity when you know more about the data generating process.

Yes thats right. Pardon my oversimplification :)

No problem, it's a super common and easy to miss mistake that shows up a lot in professional finance practice. Can lead to nasty surprises when you get hit with way more extreme events than you expect.

One of my favourite papers that had a huge impact on my FX trading, unfortunately quantopian doesn't have FX (or FX futures) yet as this doesn't apply to equities.

Stop-Loss Orders and Price Cascades in Currency Markets
https://www.newyorkfed.org/research/staff_reports/sr150.html

I also found this paper quite interesting

I'm also not sure if this has been posted here

Grant, nice paper - no surprise that downside returns are followed by positive returns - buy and hold an its simplest and best (if not buy and hold, then long bias "algos" are affected by the general market to such extent that they end up resembling buy and hold, less transaction costs)! The more subtle issue is that upside returns contain no information about future returns, which means that they 1. are not skilled at taking profit, or 2. taking profits results in subsequent poor decision making...both of which make sense.

Here's a strategy idea/exploration called Ebb and Flow. It trades ES and Bonds when both are at extremes and is Interesting because it goes long stocks and bonds.

The idea, rules and results are here (henrycarstens.com): http://wp.me/p6O8fA-aT

--h
Henry Carstens

I am thinking about implementing a macro trading strategy that will produce trading signals based on changes in measures such as: risk premium, interest rates, margin requirements and haircuts of pledged collateral.

At the moment for the universe of stocks to trade that I have in mind is (can be expanded): shadow banking ETFs, safe asset bond ETFs, clearing houses, financial institutions in the repo business, derivatives trading hedge funds and other heavily OTC involved companies.

I am not sure where to find data on haircuts and margin requirements, but I've seen an announcement from IB that they will be offering OTC data:
http://www.prnewswire.com/news-releases/otc-markets-group-expands-relationship-with-interactive-brokers-to-display-real-time-level-2-market-data-300286610.html

The idea comes from my master thesis which is titled: "The Decline of Safe Assets and Shortage of Collateral". I've been heavily engaged with this topic for years now and I think that it explains the modern macro world pretty well, so a trading strategy based on it should be profitable.

I am looking for comments, suggestions or questions from other Quantopian traders. This is still just an idea, there are some questions still to be answered like: whats going to be the universe of stocks, where will I find data, how will signals be interpreted etc. but I think that there's a lot of potential and I haven't seen many macro strategies on Quantopian.

Here's a strategy idea called Silver and Gold and trades Gold based on momentum, pullbacks and Silver. It might be really interesting to adapt to silver and gold equities.

The idea, rules and results are here (henrycarstens.com): [http://wp.me/p6O8fA-b5][1]

--h
Henry Carstens

Here's a strategy idea called Corn Predator-Prey that trades corn based on the agriculture ecosystem viewed as a predator-prey model. Wheat and soybeans are the prey and the dollar is the predator.

The idea, rules and results are here (henrycarstens.com): [http://wp.me/p6O8fA-b8][1]

--h
Henry Carstens

Here's a strategy idea called Effectiveness that trades the dollar based on its relative ease of movement vs bonds.

The idea, fully disclosed rules and results are here (henrycarstens.com): http://wp.me/p6O8fA-bh

--h
Henry Carstens

Here's another dollar strategy that tries to find the beginning of a trend in the dollar.

The idea, fully disclosed rules and results are here (henrycarstens.com): http://wp.me/p6O8fA-bn

--h
Henry Carstens

Here's a strategy idea called Econ101 based on the Krebs Cycle idea from 101 Trading Ideas. Econ101 uses the employment report and the dollar to trade bonds. Strategy idea with rules.

henrycarstens.com: http://wp.me/p6O8fA-bs

--h
Henry Carstens

Here's a strategy idea based on camoflage: How does the market camouflage it's moves? When crude oil and natural gas move in opposite directions is it a signal or camouflage?

Idea, rules and notes are here (henrycarstens.com): http://wp.me/p6O8fA-bv

--h
Henry Carstens

Here's a strategy idea based on trading tomorrow: How does gold react when bonds go the opposite direction?

Idea, rules and notes are here (henrycarstens.com): http://wp.me/p6O8fA-c5

--h
Henry Carstens

Here's a strategy idea for gold based on fear: How does gold react to fear?

Idea, rules, and notes are here (henrycarstens.com): http://wp.me/p6O8fA-cn

--h
Henry Carstens

How to measure when you need new trading ideas,
Ways to measure the effectiveness of trading idea creation,
Ways to measure the effectiveness of trading ideas

henrycarstens.com: http://wp.me/p6O8fA-d9

--h
Henry Carstens

http://arxiv.org/pdf/1212.2129v2.pdf Mostly posting this so I don't forget about it lol

I had posted this in the public forum, but here might be more beneficial

I just been introduced to Robinhood and caught wind of the Quantopian intergration.

I do not know Python at all, but I am an options trader that uses the MACD using the values of 9, 20, 6 for my entries and an 11 MA as my exit position.

I would like to take this strategy and turn it into an algorithm and have it running in Robinhood.

The strategy would work like this:

A entry uses 20% of available buying power (if a robinhood instant account, PDT counter should be no greater than 1 for safety purposes)
A buy order is triggered when MACD has a crossover and stock price is above 11MA.
And when stock price falls below 11 MA, liquidates position
If MACD signals buy, but stock price is below 11MA it's ignored

I have attached a photo, for a visual description - http://imgur.com/a/nI6X6

So stocks that are high liquidity, high momentum like FB, AAPL, NFLX, GOOG/GOOGL, BABA, PCLN, AMZN, TSLA, etc, waits to meet criteria, rinse and repeat.
The reason for the 9, 20, 6 is this triggers on the first candle, and the 11 MA minimizes the potential loss incurred.

Any help would be greatly appreciated. Thank you

This thread has gotten a bit off-topic; can we please keep it to simple links to actual papers detailing a trading strategy, rather than links to personal/promotional websites, requests for help, or other clutter.

EDIT: not to be rude, but there is an entire forum wherein one can post such things. I created this thread to be a focused place to find academic & practitioner research.

Sorry, I thought this would fall under a strategy idea

Abstract:

If managers use non-public information or misvaluation to time a
firm’s corporate actions, it is likely that equity issues will precede
Consistent with this expectation, we find evidence of earnings
predictability: the market reaction to earnings following buyback
announcements is higher by 4.56% than the reaction to earnings
following equity issuance over a 25-trading day window (-10, 15). The
difference in market reactions to earnings is smaller at 1.85% when a
5-day window (0, 5) is considered. Short-term stock returns reported
in this paper are more meaningful and sidestep the sensitivity of
long-term returns to benchmarking concerns documented in the
literature.

Disclaimer

The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian. In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.

An implementation of an idea triggered by the Clustering Illusion from List of Biases using crude oil etf's.

--h
Henry Carstens

34
Backtest from to with initial capital
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
 Returns 1 Month 3 Month 6 Month 12 Month
 Alpha 1 Month 3 Month 6 Month 12 Month
 Beta 1 Month 3 Month 6 Month 12 Month
 Sharpe 1 Month 3 Month 6 Month 12 Month
 Sortino 1 Month 3 Month 6 Month 12 Month
 Volatility 1 Month 3 Month 6 Month 12 Month
 Max Drawdown 1 Month 3 Month 6 Month 12 Month
# Clustering Illusion
#
# 081116 hac v1.0
#
# buys and sells crude oil based on Clustering Illusion bias from List of Biases and
# Scan,Pause,Idea process from 101 Trading Ideas
#
# Strategy works better on futures than etf's
# Futures implementation and more on List of Biases here:
#

def initialize(context):
# Reference to oil (USO or OIL) and heating oil (UHN) ETF's
context.oil                = sid(28320)
context.heatingoil         = sid(36034)
context.long_entry_price   = -1
context.short_entry_price  = -1

#context.message = 'initializing oil and heating oil'
#print context.message

#called every thursday
schedule_function(open_positions,
date_rules.week_start(days_offset=3),
time_rules.market_close(minutes=30))

#close positions every wednesday
schedule_function(close_positions,
date_rules.week_start(days_offset=2),
time_rules.market_close(minutes=30))

def open_positions(context, data):

# price history
# get the 20-day trailing price history of oil
hist20          = data.history(context.oil, 'price', 20, '1d')
context.message = hist20
#print context.message

hist_heatingoil = data.history(context.heatingoil, 'price', 45, '1d')

hist5             = data.history(context.oil, 'price', 5, '1d')
context.message   = hist5
#print context.message

#check the price of crude
curr_price         = data.current(context.oil, 'price')
context.message    = curr_price
#print context.message
#print hist20[1]

#average price of last 5 days
dx = 0
for index in range(15,20):
dx = dx + hist20[index]
mean_price = dx/5
#print mean_price

#count number of times heating oil was up over 5 days
ho_up_counter = 0
for index in range(40,1, -1):
if hist_heatingoil[index] > hist_heatingoil[index-5]:
ho_up_counter = ho_up_counter + 1
#print ho_up_counter

# long rules
# curr_price is higher than it was 20 days ago
# curr_price is less than 5 day average
# heating oil has been up over 5 days more than it has been down
if curr_price > hist20[1] and curr_price < mean_price and ho_up_counter > 20:
long_rule = True
else:
long_rule = False

# short rules
# opposit long rules
if curr_price < hist20[1] and curr_price > mean_price and ho_up_counter < 20:
short_rule = True
else:
short_rule = False
#s = repr(curr_price < hist20[1]) + ' ' + repr(mean_price)
#print s

# Position 10% of our portfolio to be long or short crude oil etf
if long_rule:
order_target_percent(context.oil, .075)
context.long_entry_price = curr_price
print context.message
else:
context.long_entry_price = -1

if short_rule:
order_target_percent(context.oil, -.075)
context.message = 'selling oil'
context.short_entry_price = curr_price
print context.message
else:
context.short_entry_price = -1

# close positions on Wednesday
def close_positions(context, data):

order_target_percent(context.oil, 0)
context.message = 'closing positions'
#print context.message

def handle_data(context, data):

#check stops and profit targets, pct of entry price
stop_size = .05
profit_target_size = .075

curr_price = data.current(context.oil, 'price')

#long profit targets
if context.long_entry_price <> -1:
if curr_price > context.long_entry_price * (1 + profit_target_size):
order_target_percent(context.oil,0)
context.long_entry_price = -1
print 'profit target long'

#short profit target
if context.short_entry_price <> -1:
if curr_price < context.short_entry_price * (1 - profit_target_size):
order_target_percent(context.oil,0)
context.short_entry_price = -1
print 'profit target short'

#long stops
if context.long_entry_price <> -1:
if curr_price < context.long_entry_price * (1 - stop_size):
order_target_percent(context.oil, 0)
context.long_entry_price = -1
print 'stopped long'

#short stops
if context.short_entry_price <> -1:
if curr_price > context.short_entry_price * (1 + stop_size):
order_target_percent(context.oil, 0)
context.short_entry_price = -1
print 'stopped short'

pass
#print context.message
#print context.oil

There was a runtime error.

jf le bas,
Do you have a PDF source for this paper? I can't find it via my usual sources? It looks interesting and I may implement it, but like to keep original sources around for reference.

@ Steven Shack sorry i don't have

i was wondering how to implement the futures based ideas. is it possible in quantopian? i know theyve been talking about futures for a while. are there other resources similar to quantopian that have some sort of backtesting like quantopian, that allows for algo-trading futures? or options?

thank you,
tyler

OPTIMAL EXECUTION HORIZON
by Easley, López de Prado, and O'Hara

This approach may be a strong complement to any short-term trading strategy.

The authors do a good job of laying out their intent:

"Execution traders know that market impact greatly depends on whether
their orders lean with or against the market. We introduce the OEH
model, which incorporates this fact when determining the optimal
trading horizon for an order, an input required by many sophisticated
execution strategies."

and apparent result:

"Our empirical study shows that OEH allows traders to achieve greater
profits on their information, as compared to VWAP. If the trader’s
information is right, OEH will allow her to capture greater profits on
that trade. If her information is inaccurate, OEH will deliver smaller
losses than VWAP. OEH is not an investment strategy on its own,but
delivers substantial “execution alpha” by boosting the performance of
“investment alpha”.

Authors: Eric C. So of MIT and Sean Wang of UNC

Abstract:

This study documents a six-fold increase in short-term return
reversals during earnings announcements relative to non-announcement
periods. Following prior research, we use reversals as a proxy for
expected returns market makers demand for providing liquidity. Our
findings highlight significant time-series variation in the magnitude
of short-term return reversals and suggest that market makers demand
higher expected returns prior to earnings announcements because of
increased inventory risks that stem from holding net positions through
the release of anticipated earnings news. Collectively, our findings
suggest that uncertainty regarding anticipated information events
elicits predictable increases in expected returns to liquidity
provision and that these increases significantly affect the dynamics
and information content of market prices.

Hey folks,

Just wanted to let you know that we've been putting together a curated list of trading strategy and research ideas from the community. At the moment, it's research that folks from Quantopian have published, but we're hoping to feature some from you. Send suggestions to [email protected]

Seong

NONLINEAR MARKET DYNAMICS BETWEEN STOCK RETURNS AND TRADING VOLUME: EMPIRICAL EVIDENCES FROM ASIAN STOCK MARKETS
http://anale.feaa.uaic.ro/anale/resurse/48_S08_WuJenChuang_sa.pdf

Deviations from Put-Call Parity and Stock Return Predictability

"Using the difference in implied volatility between pairs of call and put options to measure these deviations we find that stocks with relatively expensive calls outperform stocks with relatively expensive puts by 51 basis points per week"

Upon first-glance, appears particularly germane to the Q program of long-short algos:

"Extending Rules-Based Factor Portfolios to a Long-Short Framework"

https://www.caia.org/sites/default/files/AIAR_Q4_2015-03_BenderWang_LongShort.pdf

Note the section "The Costs and the Risks of Shorting" which is not captured yet (as I understand) in the Q backtester.

They claim very high sharpe. 4 factor model of overnight returns

A relatively recent paper on statistical arbitrage using log prices

Has anyone tried a long/short using estimize's new weekly top10 long/shorts?

Not necessarily a strategy but a paper on decomposition of risk into various factors that can be used for hedging. Anyone volunteers to port this octave code to Python?

decomposition of risk

@Aqua interesting paper on decomposition of risk. The code is copyrighted; it has a disclaimer but does not state the protections. Can it really be ported to Python AND shared? Not a lawyer here ...

Does anyone know any new (or alternative) trading strategy for forex currency market ?

The 7 Reasons Most Machine Learning Funds Fail (Presentation Slides)

45 Pages Posted: 6 Sep 2017

Guggenheim Partners, LLC; Lawrence Berkeley National Laboratory; Harvard University - RCC

Date Written: September 2, 2017
Abstract

The rate of failure in quantitative finance is high, and particularly so in financial machine learning. The few managers who succeed amass a large amount of assets, and deliver consistently exceptional performance to their investors. However, that is a rare outcome, for reasons that will become apparent in this presentation. Over the past two decades, I have seen many faces come and go, firms started and shut down. In my experience, there are 7 critical mistakes underlying most of those failures.

Nice/devastating article Grant. Has anyone here used familiar fractional differentiation when looking at price changes?

Ahead of Print: 2 October 2017
Estimating Time-Varying Factor Exposures by Andrew Ang, Ananth Madhavan, and Aleksander Sobczyk.

Does anyone try to backtest candle engulfing pattern on forex (or crude oil future) ? I tested using engulfing pattern by pulling historical data from IB but the result is not that good. I am just wondering how to make a better guess on engulfing pattern.

Not a trading paper, but would seem to be relevant in pairs searching and perhaps factor analyses:

Fast search local extremum for maximal information coefficient (MIC)

Abstract

Maximal information coefficient (MIC) is an indicator to explore the correlation between pairwise variables in large data sets, and the accuracy of MIC has an impact on the measure of dependence for each pair. To improve the equitability in an acceptable run-time, in this paper, an intelligent MIC (iMIC) is proposed for optimizing the partition on the y-axis to approximate the MIC with good accuracy. It is an iterative algorithm on quadratic optimization to generate a better characteristic matrix. During the process, the iMIC can quickly find out the local optimal value while using a lower number of iterations. It produces results that are close to the true MIC values by searching just
times, rather than computations required for the previous method. In the compared experiments of 169 indexes about 202 countries from World Health Organization (WHO) data set, the proposed algorithm offers a better solution coupled with a reasonable run-time for MIC, and good performance search for the extreme values in fewer iterations. The iMIC develops the equitability keeping the satisfied accuracy with fast computational speed, potentially benefitting the relationship exploration in big data.

Any good strategy database for crypto trading? or any link where I can study a bit more about it. Thank you very much

I was looking for gap strategies as many guys play these at the open, found this one on futures which was interesting https://quantsavvy.com/blog/second-day-gap-daytrading-strategy. This thread is a goldmine!

Articles by Ed Thorp: