Counteracting High Speed Traders

On the Planet Money show, there's been discussion about the practice of High Speed Trading. This is where traders use computers and high-speed connections to the stock market to analyze trade data and put in trades making penny or sub-penny profits on transaction trends by responding faster than other players.

So, for example, if someone is trying to make a buy of stocks, and they can't get them in one lot, their first partial buy might trigger an algorithm in a computer. Then the computer might use it's high speed trading connection to buy up available lots of the stock that it anticipates will soon go up by a small tick, reselling to a slower moving buyer trying to close out the complete order with a slower connection.

Institutional investors, such as those managing your pension mutual funds, are understandably wary of the practice because they're frequently the ones whose transactions become marginally more expensive due to high-speed trading algorithms. So how might they neutralize the effects of high-speed traders?

I see an opening where a larger traders could help to neutralize or even marginally improve transaction costs due high-speed traders via counter trades. In aircraft control theory, there is a phenomenon called pilot induced oscillation (PIO). This is where a pilot tries to damp out an oscillation, but because they're responding with the right response, but out of phase, the situation ends up inducing larger oscillations in the path of the aircraft. Usually, aircraft designers (and pilots) are looking to avoid PIO, but a counter trade would be trying for the opposite by inducing high speed trading responses.

The basic idea would be to issue a number of counter trades to trigger the high speed trading algorithms, then use the window of movement in price that the high-speed algorithm creates to run your intended overall transaction. For example, if you were selling, make a few buys first. Let the high speed algos try to drive the price up by buying up all the shares at that local price level, then you would issue your sale at a higher transaction price. If the high speed trading algorithms stay agressive, then you can continue to open windows for your intended transactions. Alternately if the high-speed algorithms are dialed down to become less aggressive then your marginal losses from the actions are drastically reduced. Either way, counter trading would be one way to bound the high-speed trading algorithms.

Update: in the same vein, here's an article in the Atlantic about how odd trading patterns have been identified in trading data, presumeably from high frequency trading bots.

Phase Change Capsules in Drywall

What a great idea. Daytime heat is absorbed into little capsules of a material that changes phase. The material is mixed into drywall plaster. Rooms stay cool. At night the heat is released. Great primary function, but the material is made from a process that coats Paraffin wax. I have to wonder how that does for fire resistance.

iPad and Interface Single Tasking

Geeks of the world went through a collective wave of anticipation and lament yesterday at the unveling of the Apple iPad. If there was a single most missed feature, it was probably the lack of multitasking. I'm going to say something contrarian here. The iPad doesn't need multitasking now, and possibly not ever.

With the iPad, Apple is taking a step down the path where separation between interface and application (or Model from the View and Controller if you prefer) is not just a convenient software architecture boundary, but is an actual physical interface boundary. As much as you might wish it, humans are, at best, serial-taskers when it comes to interfacing to a computer system. Notice I said system. The next logical step after the iPad is to seamlessly connect it to other higher horsepower devices; devices on the net, or devices in a cloud. Personally, the link net devices are fairly mature over the browser, but I'd like to start maturing the link to my personal, home computing devices, the cloud can come later.

If the iPad is just an interface device with just enough computing power to ease local controls processing, you can then multitask to your hearts desire by giving server type devices tasks to go off and perform. Basically, as long as you can push images onto the iPad and control commands back to a server, and do lightweight processing for compromises between those two extremes, you can think of the iPad as a physical manifestation of a application gui window on your desktop computer.
The only thing you need is a multi-event notification system, which iPhone already has a least laid a foundation.

There are some short-term, established ways to pursue this link, the X window system is a prime example. For more recent examples, you could look at VNC/Remote Desktops type applications to extend your interface device. Ironically, the VNC/Remote desktop applications are less rich than the X system. And of course, browsers are increasingly delivering reasonably responsive application over web technologies. This should only get easier with the rise of HTML5. However, Apple has always dabbled in interesting interface and display technologies. I'm hoping that Apple has it's sights set on a longer term view of this path and is preparing some newly rethought method of linking server side power to an exportable iPad interface that may blow the doors off of all these paradigms.

Healthcare Soda: Predictably Broken Competitive Markets

Interesting data on health care by Umair Haque at the Harvard Business Review in this article. (he actually has a lot of interesting articles in HBR.) I've been thinking that the private healthcare industry as a competitive market is pretty broken in the US. It's not really surprising to me that a competitive market works very poorly with its characteristics. For a market to work best, a purchase should have verifiable quality and needs to be repeatable or reversible, meaning you make that purchase decision again and again, or if you find a choice was bad, you can change the choice or reverse it. For example, that may mean giving back the product, or discarding it for another, or selling to someone else. Here's an example:

Suppose you buy a soda, then take a sip, and find that you hate the taste. No big deal - if its really that bad, you could stop drinking it partway through and toss it. Choose another brand next time.

Healthcare doesn't work that way. Now imagine if the only way to shop for soda is to buy a can every month and put it in the garage. If you open that soda you have to keep that can, finish the drink, finish what's in the garage, and keep buying it for a while. That's what happens if you get seriously ill and really need to lean on that health insurance. You might know how an insurer handles routine care, but you really don't know the quality of your health provider until you're in dire straits. At that point, you can't switch.

In these circumstances, is it any wonder that there is a fundamental disconnect between the profit of health insurance companies they quality they deliver. In California, there's been an ongoing fight about insurance companies cancelling policies when when serious care is needed (search for "California lawsuit health insurance cancellation" or click here. Some of insurance companies even paid employees bonuses to cancel policies of very sick people based on technicalities on their initial applications (sometimes years back). Settlements in some of those cases after the fact must be very bittersweet I'm sure that some of those patients received worse care, or suffered permanent injury or death. Finally, if you can't work, eventually you hit the end of what you're employer is willing and able to provide and you go bankrupt. This drives a suprisingly large portion of bankruptcies in the United States (article here).

I haven't even touched upon how things are even more broken because we're not even the ones shopping for this healthcare soda. Our employers are usually the one shopping for it.

I don't know how to fix things - I do like to idea of a publicly chartered competitor to private insurance, but really we need to find a way to connect up the healthcare market costs to delivered value to us, the consumers.

Touchable Holograms

Popular Science posted this article and video about touchable holograms at Siggraph '09, this year. It's a combination of three items: a 3d display using concave mirrors something like this 3D mirascope toy, two wii-motes strapped statically for position detection, and a home-brewed (university-brewed?) phased array of ultrasound drivers.

A Sheeva Plug Computer would make a great NAS

I've been looking at the Sheeva Plug dev kit from Marvell. It seems to be an ARM core System-on-Chip (SoC), Linux compatible, with an interesting combo of device controllers hanging off of it, including ethernet, USB, and SATA ports.

The dev kit seems to be the size of a large wall wart and part of Marvell's "Plug Computing" marketing drive. Not sure about plug computing bit, but I'd love to fit the board into a small two or three bay external drive enclosure, or maybe a drive dock. Running a minimal linux system, a looks like it would be a nice way to setup a local Network Attached Storage (NAS) backup system with enough smarts to do offsite backups to Amazon S3 or Rackspace's Mosso.

Telephones over the Internet, a Skype vs Gizmo5 comparison

Engineers are always trying to optimize value - sometime's we choose the wrong values to optimize on, but thats a discussion for another day. Recently, I've been looking through different VoIP services to see how viable they are as ways of augmenting or replacing my residential land-line phone service. VoIP is Voice over Internet Protocol. The per-minute prices for VoIP are free-to-low, with direct computer-to-computer calls typically free and computer-to-phone service typically at a low monthly or per-minute fee. A computer-to-phone call is often called PSTN or Public Switch Telephone Network termination. There are also other variations on this where you call and the service calls back but, for me, those services are more hassle than they're worth.

VoIP phoning isn't as reliable as regular landline phones. e.g. in a power outage, you have to have your own backup to run the computer or other devices needed to connect to the internet -- and even that is futile if you net connection goes down with the power. Between the power and issues about getting correct 911 service, etc, VoIP is something I see as a way to augment my current service by reducing cost for non-emergency calls.

So far, after looking through a number of different options, Skype and Gizmo5 are two services of interest. They break up their costs in different ways, and there are various technical and market aspects that might make one favor one vs another. For example, I like that Gizmo5 uses standard communication protocols to initiate and transfer calls while Skype uses more proprietary methods. They also both offer other services that I wouldn't use very often, such as video chat.

Incidentally, one thing that both work with is my Ipod Touch 2G. An Ipod application called "Fring" allows calls over WiFi using a number of VoIP services. It's a geeky feature to play with but there constraints there that make me think that this won't be the way I put the most hours on a VoIP service. More on that later maybe, but I'll just close out talk of Fring with the statement that it's use of standard protocols such as a SIP VoIP interface is what allows Gizmo5 to work with Fring. This is one reason to like standards

As of this writing, both offer service-to-service calls for free. So really I'm comparing the cost of making computer to phone calls. Gizmo5 offers bridging calls to Skype for a flat fee for time. This seems somewhat unreliable, presumably because Skype doesn't encourage this. Skype has a larger user base, but is owned by questionably consumer friendly Ebay (also owner of PayPal).

In terms of flat rate service, Gizmo5 offers 1.9 ct/min rates while Skypes are 2.1ct/min. Skype also charges a connection fee of 3.5 cents. So for flat rate service Skype loses out pricewise. Skype, however, also offers unlimited calls for $2.95/mo. So the breakpoint is ~155minutes for Gizmo5 vs skype. Meaning if you plan to consistently use more than 155 minutes a month on a service like this then Skype will be cheaper. Since this is an addon side service, I think I'm going with Gizmo5 - I liked their use of open connection protocols anyway.

Since I first wrote this post the LA Times ran an article on a study citing the average cost that consumers pay for cellular service -- over $3 per minute wow...

(photo by: asdelwood, under a Creative Commons license on Flickr)