Have you ever found yourself struggling to cram too many results and ideas into that 6-page conference paper limit? Maybe you should take the hint and split it into smaller pieces. Here are a few good reasons to split papers in two:

1. Clarity: Papers are just better if they are about one idea instead of eight. Readers get confused if you’re simultaneously trying to argue about several things at once.

2. Ease: It’s easier to write a paper that’s about one thing. It’s more difficult to explain all the connections necessary to support a paper about many things.

3. Depth and Completeness: You should give your great ideas the respect they deserve by providing them full coverage. If you reserve an entire paper for one idea, you will have the room you need.

4. Compartmentalization can work against you: Readers typically assume one paper has one idea.  If you cram several ideas into one paper, only one of them is going to “stand out” as the idea that the paper is about.  The other good ideas may be forever lost to your reading public.

5. Marketing: More papers with distinct ideas makes it look like, well, you’ve got a lot of ideas! Help people know the real you.

6. Search Engine Optimization: More distinct paper titles means people are more likely to find your work.

Questions?

After my recent posts about titles and abstracts, some folks have asked about the larger task of writing the entire paper.  I haven’t had time to complete that fully, but thought I might share an outline, and some material by others on this topic. First a few links to other’s coverage of this topic:

• How to write a control paper by  Joa ̃o P. Hespanha  at UCSB, recommended to me by Magnus Egerstedt. I like this one overall, especially he agrees with my view on searchable titles.  But I don’t agree with Hespanha’s assertion that the abstract “should not attempt to justify why the problem is important/timely.”

• Tips for writing technical papers by Jennifer Widom at Stanford.

• Writing a technical paper by Michael Ernst.

Send me an email if you think I should add to the above list. Also, here’s a brief outline of how I structure my papers:

Title: See blog post. Bottom line: Search Engine Optimize!

Abstract: See blog post. I believe this is the most important paragraph.

1. Background and Related Work: For conference papers I like to combine these two topics into one section (in lieu of “Introduction”).  Include: What is the problem you are trying to solve? Why is it important? Why is it hard? How have others solved it, including your own previous solutions. When you mention other approaches, suggest reasons your approach is different or better.

2. Approach: What is the key insight or innovation of your approach? If at all possible describe the method simply in words first (not equations!). Then follow up with equations or algorithmic details as necessary. Avoid nitty gritty details of implementation for now.

3. Implementation: This section is not always necessary.  The point is to explain those details not covered above that might vary from implementation to implementation but that others need to know in order to reproduce your results. Also information about the specific implementation that might explain extremely good (or bad) quantitative results: Things like machines you ran it on, language you wrote it in, other details one might need to know.

4. Methodology: Describe the experimental setup, the various conditions of the experiment, how the results were measured and recorded.

5. Experimental Results: Describe in a factual manner, the quantitative results of the experiments. Hold off on rendering an opinion or conclusion here.

6. Discussion and Conclusion: This is were you draw conclusions from the results. You might speculate on the causes of the results or opine on what future work might entail or reveal. Recap what you did, restate the main points from your abstract. Acknowledge those who helped or funded.

As you craft a name for your paper, you should consider these potential objectives for the title you choose. A title should:

1. Describe the content of the paper;
2. Distinguish the paper from others on a similar topic;
3. Catch the reader’s attention and interest;
4. Match search queries so people will find your paper (and cite it).

#4 sticks out doesn’t it?  It seems obvious that #1 would be the most important objective for a title. But as search and electronic publishing becomes more and more important in research I’ve come to believe that the last point (search) may indeed be the most important. Think of it as SEO (search engine optimization) for your research vision.

The importance of research vision SEO

I’m sorry to say that I’ve attended more than one talk at an important international conference where someone else presented a paper exactly describing an idea I had published earlier. Whose fault is that? Well it was my fault. When I dug up my old paper I realized that the particular idea wasn’t reflected in the paper at all. It was “buried” in the content.

Most people begin a research project with a comprehensive review of the literature. It is crucially important that they find your paper when they’re at the beginning of their research/writing cycle.  Assuming your work is good, if they find your paper early on, you’ll help them avoid reproducing your work, and perhaps encourage them to make use of your ideas as they go forward. And of course you’d like them to cite your paper as well.

A well crafted title is the best way to accomplish this goal.

How does one accomplish this with a title? The most important thing to realize is that long titles are OK. Don’t prioritize a short, snappy title at the expense of a full description. Make sure that your important claims are reflected in the title. Also, be sure to include the keywords that define the niche your paper occupies. Otherwise folks won’t find it.

Example: Suppose you’re a robotics researcher, and you’ve discovered that probabilistic pathfinding is far superior to earlier methods that depended on deterministic methods. You might choose a title like

Probabilistic Pathfinding: Beyond Deterministic Methods for Navigation in Rough Terrain

Note that we include both the old method ‘s name and the new idea in the title.  So folks searching for work using the old approach (deterministic) will also discover your new idea (probabilistic) when they might not have otherwise.

Making it catchy

It’s also useful to create a title that sticks with people. You might consider a few devices to help you there. How about a snappy name or acronym for your approach?

GRAMMPS: A generalized mission planner for multiple mobile robots in unstructured environments

Did you know that people actually prefer titles that contain a colon? Don’t force it, but if a two-part description of your paper fits well into an “idea:description” template, go for it:

Household robotics: autonomous devices for vacuuming and lawn mowing

What not to do

It’s probably a bad idea to choose a provocative title alone. For instance

What are the ants doing?

would be bad, but

What are the ants doing? vision-based tracking and reconstruction of control programs

is great because it is both catchy/provocative and also explanatory.

A few thoughts by others

• A research paper about the effectiveness of colons in paper titles.
• Someone’s opinion on the topic of colons.

Many writers save the abstract for last. I don’t. I write the abstract first. If the abstract isn’t interesting, compelling, downright exciting, people aren’t going to read the paper. Use the abstract to set the intellectual goals for your paper. But the paper has to live up to the promise you set.  It is a contract with the reader.

I set optimistic goals for myself with the abstract. But as I write, I consult it to make sure I’m on track. If I find I can’t deliver on a part of that promise, I revise the abstract accordingly. I rarely extend the claims made in the abstract.

So, crafting an abstract is a first step in paper writing (and sometimes a first step in research too). But the abstract isn’t static, it must evolve as the paper evolves. It’s a process. Give my process a try. Let me know how it works for you.

Overview: Step through each of the elements below and create one or two sentences per element. Don’t worry about connecting these sentences together for now. Save that for later. Also, all these elements don’t make sense for all papers: It is OK if you don’t have a sentence for some elements.

1. Problem: What is the problem your paper addresses? Avoid describing the solution here.

We address the problem of robot navigation across unmapped rough terrain.

2. Importance: How would a solution to this problem change the world? Remember that it’s not obvious to everyone else how important this problem is.

A solution to this problem will enable more rapid validation of the efficacy of sleep medication.

3. New capability: What can we do now that we couldn’t do before? Quantify if possible.

Our approach enables robots to climb trees five times more quickly than was possible before.

4. Challenge: Why is this problem hard? and/or What difficulty do other solutions face?

In order to solve this problem a robot must know its location to within 1mm.

Traditional approaches to this problem are exponential in the number of obstacles in the environment.

5. Background: A sentence or two about other approaches; yours or others.

Smith and Jones address this by building a polygonal model of the obstacle field, creating a visibility graph, then applying Dijkstra’s algorithm.

6. Insight: What did you discover? or How did you approach the problem differently?

By framing the problem as an optimization task, we are able to leverage a linear time algorithm.

7. Solution: Provide some specific detail about the solution.

Our algorithm measures the standard deviation of the quantitative features in each dimension, then sorts the result, providing an optimal solution.

8. Evidence: Summarize the evidence you have for your approach: A proof, an implementation, or quantiative results.

The algorithm was implemented on a monkey-based robot and shown to perform 17% faster than a dog of similar mass.

Now that you have some text for each element above (or most of them), try piecing them together. My order above is a typical way to order things, but it isn’t necessary that you keep the order I have. Once you have a good order, read it over and over.  Smooth the rough edges.

You will find that you need to add some “glue” verbiage. It’s usually necessary, for instance, to provide a little more background information to help the elements hang together.

Read through the abstract and revise it, polish it. Avoid repeated words. Have other people read it. This is the most important paragraph of your paper.

Good luck!

Phillip Koopman at CMU has a similar view.

As you craft a name for your paper, you should consider these potential objectives for the title you choose. A title should:

1. Describe the content of the paper;
2. Distinguish the paper from others on a similar topic;
3. Catch the reader’s attention and interest;
4. Match search queries so people will find your paper (and cite it).

#4 sticks out doesn’t it?  It seems obvious that #1 would be the most important objective for a title. But as search and electronic publishing becomes more and more important in research I’ve come to believe that the last point (search) may indeed be the most important. Think of it as SEO (search engine optimization) for your research vision.

The importance of research vision SEO

I’m sorry to say that I’ve attended more than one talk at an important international conference where someone else presented a paper exactly describing an idea I had published earlier. Whose fault is that? Well it was my fault. When I dug up my old paper I realized that the particular idea wasn’t reflected in the paper at all. It was “buried” in the content.

Most people begin a research project with a comprehensive review of the literature. It is crucially important that they find your paper when they’re at the beginning of their research/writing cycle.  Assuming your work is good, if they find your paper early on, you’ll help them avoid reproducing your work, and perhaps encourage them to make use of your ideas as they go forward. And of course you’d like them to cite your paper as well.

A well crafted title is the best way to accomplish this goal.

How does one accomplish this with a title? The most important thing to realize is that long titles are OK. Don’t prioritize a short, snappy title at the expense of a full description. Make sure that your important claims are reflected in the title. Also, be sure to include the keywords that define the niche your paper occupies. Otherwise folks won’t find it.

Example: Suppose you’re a robotics researcher, and you’ve discovered that probabilistic pathfinding is far superior to earlier methods that depended on deterministic methods. You might choose a title like

Probabilistic Pathfinding: Beyond Deterministic Methods for Navigation in Rough Terrain

Note that we include both the old method ‘s name and the new idea in the title.  So folks searching for work using the old approach (deterministic) will also discover your new idea (probabilistic) when they might not have otherwise.

Making it catchy

It’s also useful to create a title that sticks with people. You might consider a few devices to help you there. How about a snappy name or acronym for your approach?

GRAMMPS: A generalized mission planner for multiple mobile robots in unstructured environments

Did you know that people actually prefer titles that contain a colon? Don’t force it, but if a two-part description of your paper fits well into an “idea:description” template, go for it:

Household robotics: autonomous devices for vacuuming and lawn mowing

What not to do

It’s probably a bad idea to choose a provocative title alone. For instance

What are the ants doing?

would be bad, but

What are the ants doing? vision-based tracking and reconstruction of control programs

is great because it is both catchy/provocative and also explanatory.

A few thoughts by others

• A research paper about the effectiveness of colons in paper titles.
• Someone’s opinion on the topic of colons.

I recently helped my elementary school-aged daughter publish a book.  A real book with with a shiny cover, its own ISBN, and an online “store” where you can buy it for $3.99 (she gets $1).  Check it out.

It was a wonderful experience for her.  She’s really fired up now about writing, reading and creating. We’ve read her book dozens of times (me to her, her to me, me to our cat, etc.).

This is also fun for adults (James Altucher says it’s better than a business card). I’m planning to publish a cookbook soon. Here’s how you can do it too.

What you’ll have when you’re done

When you’re done with this project you’ll have a real paperback book your hands. You can have 1 to 10 copies printed for you at $2.15 each plus $5.00 for shipping (for a 24 to 100 page book). The print quality is very good, and the binding and cover are similar to the quality you’ll find in a bookstore.

You can associate an ISBN with the book at no cost if you like.  You can also sell it online, including at amazon.com.  This post is about how I published a book with my daughter, but it is also simple to publish something more complex.

What you need to get this done:

Time: However long it takes you and your child to write and illustrate + 2 hours.

Computing: A scanner and Microsoft Word.

Step 1: Choose a publisher (I recommend  createspace.com).

There are a lot of other options for doing this besides createspace.  I looked around a bit and decided that they seemed to provide the best combination of simplicity, quality and low cost. I’ve been happy with the result. Thanks to Max Dama for the pointer. If you choose another publisher (lulu.com is a good alternative) the steps are similar.

Step 2: Download a template

I chose to create a 5.5″ by 8.5″ book because this is the same size as a US Letter sheet folded in half.  It’s helpful to be able to fold a sheet like this for your child to get a sense of how big the final book will be.  Here’s the 5.5 by 8.5 Word template I used for my daughter’s book: internal-template.doc.  That template is based on the one provided by createspace for this size book.

Download it and take a look at the internal pages.  You’ll see that the main content pages have a large square on them.  That’s where the illustrations go.

If you’re interested in trying a differently sized book, or you want to start with one of create space’s templates, go to this page and scroll down a bit to view the templates for differently sized books.

Step 3: Write the book (this is the fun & easy part)

Print out multiple copies of one of the pages with the squares on them.  Have your child create a proto-book by drawing an illustration for each page within the square.  They should write text for each page outside the square.

I suggest, for a first book, that you stick with black and white, but you can have the illustrations in color if you like.

Step 4: Cut out and scan the illustrations

Cut out the illustrations.  You may choose to keep the border around the edge, or to cut it off.  If you cut it off, the illustration will not have the border in the final book.

You may find it helpful to mark the illustrations and text pairs with a small numeral somewhere so you can keep them in order.

Now use your scanner to scan in all the images.

Step 5: Assemble the book contents

Insert each of the scans into one page each (use Word’s insert>picture>from file option).

Then type the text below each image (don’t worry too much about exact formatting at this stage). My daughter and I enjoyed sitting down together and looking at each picture with me typing the the text as she read from the pages. You might find some other fun way to do this, perhaps your child wants to type in the text himself.

Once you have all the text and images in the Word file you can take time now to spruce it up.

Note that the odd-numbered pages will appear on the right.  I arranged things for my daughter’s book so that there were images and text only on the right.  This made the book a little bit more readable for kids, and it also helped get the book up to 24 pages (the minimum) more easily.

Step 6: The cover

For the cover, you’ll need to assemble a few things, as follows:

1. A photo of your child for the “author’s photo” on the back.
2. About 200 words of “about the author.”
3. A title and subtitle (subtitle optional).
4. A quote about the book or author (optional).
5. An image for the front cover (optional).

Once you have all this, you’re ready…

Step 7: Submit electronically

Go to createspace.com and click on the start a title for free link.  Just follow the instructions.  The system will ask you for the information in an orderly fashion.

Step 8: Wait

You’ll have to wait a day or so for the publisher to review the book to make sure it will print properly.  They also scan it for illegalities (not sure exactly which ones to be honest). They’ll send an email when it is ready.

Step 9: Review proof and order

Eventually you’ll get an email informing you that the files are ready for you to proof.  If you agree it looks fine, you just click “approve” and you can order your proofs.  You’ll receive them by UPS. These proofs might be all you want to order if you just want a few copies for you and your child.  They are full quality prints of the book.

Once you approve the proof, your book is immediately available for sale online.  You can set a price higher than cost so that you make “a profit” on each copy printed.

Enjoy…

I’m glad I was in the market today

You’ve probably heard the advice that market timing is a bad idea because if you exit the market you’ll miss the dramatic up days (like today).  Many, well respected investment advisors tell us that market timing is a bad idea.  Here’s an article at CBS News from December 2009 advising against market timing:

The road to investment hell is paved with market-timing efforts because so much of the long-term returns provided by the market come in short and unpredictable bursts. For example, one study, covering 107 years ending in 2006, found that the best 100 days (out of more than 29,000) accounted for virtually all (99.7 percent) of returns.

But wait!

This article in Forbes tells us:

Proof of the fallacy of buy-and-hold as a strategy is easy enough to find. In the 1929 crash and its aftermath the market lost 86% of its value and did not get back to its 1929 level until 1955, twenty-six years later. That was 26 years of waiting for buy and hold investors (if there still were any) for the market to ‘come back’.

Here’s another article making the same point.

Who’s right?

There’s really two different issues here that should be treated separately.  First, yes, blind buy and hold is dumb.  You need to actively manage the equities you hold.  Market cycles move more quickly now, especially if you’re investing in tech stocks or commodities like gold or oil.

But timing the market is impossible, especially the whole market. My view is that timing the entire market is tough because it is driven by truly unknown factors such as 9/11, bank failures, secret negotiations between Germany, France and Greece, and so on.

What is one to do?

The first thing to observe is that it really is possible to separate good companies from bad.  Buy good companies (of course).  But that’s not the complete answer.  Even if you held the 20 best components of the S&P 500 in 2009 you would have lost 40% in the down turn.

The rest of the answer is that you’ve got to have hedged positions too.  That doesn’t necessarily mean you have to short stocks, but you do need to consider equities that zig when the rest of your portfolio zags.  Gold, oil, bearish funds and short ETFs can accomplish this.  The trick is to find the right mix.

I’ll say more about that in coming posts.

Disclosure: Tucker Balch was up too late.

A recent story on marketwatch.com talks about why a 24 hour market hasn’t taken off, even though it seemed inevitable a few years ago:

The idea of a 24-hour market started as an exciting realm of possibility in the 1990s, when stock exchanges enabled trading outside of the 9:30 a.m.-4:00 p.m. Eastern session, as part of efforts to fend off competition from other trading venues, such as electronic communication networks. But it never took off, even as multiple boom-and-bust cycles came and went in the following years.

The article focuses on the lack of traders and market makers (thus liquidity) after hours.  I agree this is part of the reason.  But another perhaps more important reason is that people are just set in their ways.

Folks who are experienced in the market like the rhythm we have:  The Asian markets open and close, we note their progress as predictors for Europe.  Then the European markets open and close as we march closer to the US open.  Traders on the East coast wake up at 4AM to track this activity as well as the the news in Europe. The S&P 500 and DOW Futures begin trading and serve as bellwethers for the NYSE open. Retail traders put orders in with their brokers. 9:30AM ET is sometimes spectacular as pent up energy (represented as trading orders) explodes.  We enter the trading day on the major exchanges with auctions that set the price for every traded equity.

After the fireworks of the open, prices and volume usually settle down and we enter a phase where the market follows news throughout the day.  Finally as we approach the close the day traders and HFT houses cause a ramp up in volume and an occasional sharp turn.

The SEC and most companies wait until after market close to make big announcements: Earnings reports, CEO health problems and so on.  The pause in trading prevents monster moves in the market and it gives people some time to mull over the real meaning of the news.  In fact for really big news they’ll make the announcement on Friday so we have all weekend to calm down.

Disclosure: Tucker Balch likes the 9:30 to 4PM market!

Reason 1: Economic outlook

Oil goes up when the economy is good because energy literally fuels our economy.  There are a number of signs things are looking up.  We’ve been plagued lately by Greece and Italy.  Greece has responded by bringing on a new MIT grad prime minister, and Italy is moving Berlusconi out of the way.  These are both good signs.  In the US we’re seeing a reduction in unemployment, and positive consumer sentiment.

Reason 2: Israel may strike Iran

“Israel will launch military action to prevent Iran developing a nuclear weapon as soon as Christmas, intelligence chiefs have warned.” — Daily Mail

“The possibility of a military attack against Iran is now closer to being applied than the application of a diplomatic option” — Shimon Peres

This is strongly speculative play, so don’t invest in oil only on this basis.  However, if Israel were to strike Iran, I believe oil would would surge.  Iran is the world’s 5th largest supplier (after Saudia Arabia, Russia, the USA, and China).  I don’t think this affect would last long though, because Israel would likely not strike Iran’s oil producing infrastructure.

Would Israel really attack Iran?  Probably not.  There are indications that these noises from Israel are saber rattling aimed at convincing the rest of the world to step up economic pressures on Iran.

How to make this play

There are a few ways to approach it. A story at seekingalpha.com recommends going long on a 3x energy ETF and short the SP500 via a 3x short ETF.  I agree with many points in that article, but I think the approach is risky and extreme.

Here’s what I’m going to do: First, buy oil.  One way to do this is via the ETF OIL. OIL is presently priced at about $25.00.  Next, look for a way to hedge this. Ordinarily I would buy a good airline like Delta (DAL).  I’m not going to buy Delta today though, because it’s up nearly 4%.  I’m going to look for a dip in the next few days to enter this hedged position.

Another possibility, once you hold OIL, is to sell covered calls.  The November 25 Calls are priced at $.55, which provides 1.4% return in only one week.

Exit…

OIL closed above the strike price on expiration so this trade ended profitably.

Disclosure: Tucker Balch is long OIL and DAL.

A predictable pattern

The path of Apple’s stock price (AAPL) has followed a similar pattern each of the 3 previous times a problem with Jobs’ health has been announced.  I’m going to repeat the plan I outlined previously, when Apple announced a health problem for Jobs, but with a slight change that I hope will improve return (see “the setup” below for info on the change).

The pattern is a gap down at market open, then a strong recovery, sometimes over several days.  The gap down in price is an over reaction by investors to the bad news. But the selling drives the price down to a level that others see a buying opportunity. It’s likely that the stock will follow this pattern again on Thursday morning (October 6, 2011).

I think the pattern will be repeated, but perhaps not as strongly as in the past. This bad news has been expected since Jobs stepped down as CEO. And Jobs’ influence on the company has been reduced.  So the price presumably includes anticipation of this event.

Here is a low-risk trade that will make money if the pattern repeats, but expose us to minimal risk if it does not.

The setup

First, we should wait for some evidence to confirm the hypothesis of market over reaction: Observe the market open.  If AAPL does not gap down, our assumption is contradicted so it may not be wise to go forward.  However, with this volatile market, there is a possibility that the overall market may be sharply up (or down) and mask the situation.

The thing to do is consider AAPL’s opening in comparison with QQQ. QQQ tracks the NASDAQ 100 index — a proxy for high-tech.  We’re going to leverage the difference between the high-tech market as a whole versus Apple. If AAPL is proportionally down in relation to QQQ, we’re OK to proceed.

I will wait to confirm the gap down. And then wait for an upward tick or two.  The last time we saw this pattern I waited too long to enter and missed the opportunity for 3% worth of return — the trade was still profitable, but would have been more profitable if I had entered earlier.

Then…

The trade

The idea is to buy an approximately equal dollar value of PSQ (the QQQ short ETF) and AAPL. PSQ and APPL will hedge each other, and thus provide a low-risk pair. We’re betting though, that AAPL will climb (abnormally) more significantly than PSQ will drop.

When to exit

We should know by noon whether the trade is working.  We should see AAPL and PSQ prices converging.  Follow this link to view a chart view of the situation.  The time to exit is when they’ve stopped converging.  I’ll be out by the end of the day at the latest.

Risks

First observe that it’s important not to enter the trade if AAPL does not gap down (in relation to QQQ). Waiting for this trigger allows us to enter the trade with some confirmation of the assumption that an over reaction to the news is in play.

The key risk is that AAPL will then continue down and not recover.

OK, I’m in!

At the open AAPL was down 1.25% and QQQ was down .13%.  Just a little more than 100 basis points.  Not as aggressive a difference as we’ve seen in the past.  I think we’ll make between 0% and .5% on this today.

If nothing else it will stand as a test for my “low risk” claim.

The trades

I entered these positions at 9:36AM.  BTW: If you’re not in already, I think it’s too late.

13 AAPL @ $376.39 = $4893.07

146 PSQ @ $33.27 = $4857.42

As of 9:52AM we’re up $30 (or 30 bps).  So there was a little juice there.  We’ll revisit at noon.

Getting ready to exit

Seeing 60 bps. Out at 10:50AM. Details later.

Disclosure: Tucker Balch holds AAPL.