Monday, May 31, 2010

The VOR Visualised (as a lighthouse)

Last night, I watched the recorded video session of TWiT.tv network podcast, "Maxwell's House" (which is currently undergoing a slow redevelopment along strictly aviation lines).

The topic for the episode (74) was aviation navigation. Ray Maxwell tried to explain how a VOR beacon worked using a visual metaphor. I found his explanation a bit confusing in the moment, but after re-listening today, I now have his visual idea burned into my head, and it's so elegant that I now just have to write it out!

Ray kept expanding the acronym as visual omnirange, but VOR actually stands for VHF Omnirange. However, a search of the interwebs revealed quite a few aviation authorities who use the word, visual, as the first word encoded by the acronym, strangely, as the signal is entirely radio.

I suppose they're meaning "visual" in the sense that interpreting the signal in the cockpit for navigation is normally done by visual reference to an instrument (In the way way back, there were aural navaids that depended on your listening to a certain formatted sound for changes to determine your location in relation to the signal).

A VOR is an improved radio beacon which, by the way of it's construction, permits the user receiving its signals to determine with great precision their exact bearing from the beacon's perspective (called a radial in the parlance).

Before VORs, most radio navaids were NDBs (non-directional beacons). So called because they provide no directional cues. They are simple AM stations. An ADF (automatic direction finder) receiver in your aircraft (or boat/ship, they were also commonly used for coastal navigation at sea) used a directional antenna to localize the NDB, causing a pointer on your aircraft instrument to literally point to the direction from which its signal was strongest (presumably co-incident with the station's transmitting antenna...but this is radio).

Because an NDB's signal can tell you nothing about what direction you happen to be from the station, it's use for fixing your position is less accurate. Its direction from your perspective must be estimated with respect to a compass reading and the ability of your ADF equipment to precisely pinpoint the azimuth of the strongest signal.

To get a sense of how this might not be the most accurate thing, take a handheld AM radio (its internal bar antenna is directional) and swivel it around while listening to a station. Notice how the signal fades? Notice how the strongest signal occurs over a rather wide span of arc?

With a compass, you can compute a compass bearing to the NDB. And with such a reading from two NDBs, you can plot a position fix, and thereby know where you are!

Unfortunately, accuracy is limited by your compass and ADF equipment. Magnetic variation is also a factor, possibly a major one, because it varies with location. When you measure your compass bearing to two NDBs, it's with respect to the magnetic variation at your location, which cannot be known precisely because you don't yet know your location.

This is its own interesting problem to think about, but suffice it to say that if you're following a constant compass track toward or away from an NDB, the varying magnetic variation along your route will cause this line of constant bearing to be curvy, perhaps significantly so in areas on sectional charts called out with special warnings about magnetic disturbances.

With VORs, the variation programmed into the station will be the same or often similar to the actual variation at the station's location. It actually doesn't matter because the programmed variation is depicted on your chart, evidenced by the differing cant of the north-arrow for different VOR stations. Since the unique signal of a VOR lets you determine in what direction you lie with respect to that station, you can use those values to much more precisely plot your location when you take readings from two or more stations. You don't need to take magnetic variation into account (in fact you don't even need a compass) because the bearing being determined is referenced from the station, where the variation is known, to you. Not the reverse, where the variation is not known.

The way a VOR does this is where the novelty of Ray Maxwell's description comes in, re-imagined now by me into hopefully clearer terms. This is so cool, and so simple, it blew my mind!

Imagine a VOR station, not as a radio beacon, but as a lighthouse. It has a bright white rotating beacon, which emits a powerful beam in a single direction and rotates at a constant speed, say one RPM, sweeping clockwise when viewed from above. Now, imagine that as the beam rotates through north, a red beacon on top of the lighthouse flashes at the instant the rotating beam hit 360 degrees.

If you were standing due north from this lighthouse, staring south back at it, you would see the white beam rotate past you giving you a flash of white as it hit 360 degrees. You would also see the red beacon above flash at the same instant. Now, the steady white beacon is still rotating, one RPM, or in other words, six degrees per second.

If you were now at some random location nearby the lighthouse, knowing its configuration as we do, once you sighted the beacon on the horizon, you would be able to quickly determine your location relative to the beacon by timing the appearance of the red and white flashes. Start your stopwatch when you see the red beacon flash. Stop timing when you see the steady white rotating beacon sweep past you. Then, your location relative to the lighthouse is:







timefrom the lighthouse
simultaneous red+white flashdue north
white 15 seconds laterdue east
white 30 seconds laterdue south
white 37.5 seconds laterdue southwest


For each additional second between the red and white flashes, you are located someplace along a bearing from the lighthouse, 6 additional degrees from the north reference. These bearings radiating away from the station are then known as radials. If you know the timing between the red and white pulses, you know which radial you are on.

This can be plotted accurately on your chart because the depiction includes the station's north reference. You can draw a line away from the station corresponding to the radial you observed. Repeat this with a second station and you have an accurate position that's immune to the effects of compass accuracy and magnetic variation!

As you travel along, by keeping the radial indicated by a given station constant, you are assured to be travelling along that radial, either toward or away from the station.

For the real VORs, the signals are radio, not light. The signal analogous to our flashing red timing beacon is broadcast omnidirectionally, while the signal analogous to our rotating white beacon is a tight directional radio beam, rotating through a full circle, just as our light beacon was. On old VORs, there is a directional antenna inside the VOR housing that physically rotates. Newer VOR stations have a circular array of antennas that are electrically modulated to create a tight, directionally focused radio beam which rotates without moving parts.

Simplifying slightly, these two VOR radio signals are modulated differently so the VOR receiver in your airplane can tell them apart. The receiver handles comparing the two signals to determine their timing with respect to each other, as we did by looking at the lighthouse in our example. And in this way the receiver knows which radial it is located on from the VOR it is receiving. Because of this fact, no special directionally tunable reception antenna is required for a VOR receiver, unlike ADF. This makes VOR navaids more accurate, and the airborne equipment is simpler and cheaper.

A technical discussion is available on Wikipedia, but this simple and intuitive visual analogy just made so much sense to me that I decided I had to share it here.

On doppler VORs, or D-VORs, I think both radio signals are omnidirectional, but the phase of the modulation of the "rotating" signal varies around the circle of antenna elements, creating a unique phase relationship between the two signals as one travels around the circumference of the station. There is no directional beam, but the VOR receiver computes the radial you're on by referring to this unique phase relationship between the signals that couldn't be received along any other radial from the station. I think... The Wikipedia article started to get math heavy at that point.

To my knowledge, our visual lighthouse analogy for VOR operation has never actually been used for real navigation by nighttime visual reference, but it could be. An additional light signal would be required to flash out an identifier signal, so you could determine which lighthouse-VOR you were looking at on your chart.

An audio morse-code identifier is part of a radio VOR's signal, to which you listen after tuning the station to confirm the signals you are receiving are from the station you wanted (and these days, more advanced avionics monitor this morse ID code for you and present the three or four letter code on your navigation display, saving you the trouble).

The lighthouse example did remind me of my coastal navigation days, where different signal lights and lighthouses do indeed flash various ID codes at you, to which you can reference on your nautical charts to determine which light you're looking at. But in this case, the signal lights are the visual equivalents of NDBs. The flash pattern contains no embedded directional signals. Taking a bearing to the light from the pelorus, and along with the ship's heading you can draw lines of position on your chart radiating from lights sighted, a process essentially the same as using an ADF with NDBs in the air, with the same accuracy pitfalls from compass and magnetic variation variances.

Later in life I observed that the visual airport beacons for different airports in my local area tended to rotate at significantly different speeds. From my earlier nautical light knowledge, I assumed that aeronautical charts might have timing data for each airport's beacons so that one might confirm the identity of the field they have in sight at night. Alas, it's just a coincidence. The timing is crudely specified, a slower range of flash rates for airports, a faster range for heliports. There are also color codes in the flashes which serve to differentiate civil, military, land, water, and emergency services aerodromes, but specific aerodromes are not positively identified by this signal.

I should note that terrestrial radionavigation aids are effectively obsolete now by the extremely accurate, increasingly cheap, and highly available satellite navigation systems. It's expected that these satellite constellations will become more numerous and feature rich, increasing the likelihood of guaranteed signal access for civil navigation uses. It's practically this way already for lower accuracy positioning, but for the most precise uses (i.e. for instrument landing systems and automated landing) there are periods of unsuitability due to the configuration of the orbits of the satellites, or maintenance of the error correction mechanisms (check RAIM and WAAS). Most of the time though, you can get accuracies of a few feet, anywhere with a decent view of the sky. That's better than any terrestrial system.

The FAA desires to shut down most VOR stations as soon as the GPS system is deemed available and precise enough, as they are expensive to maintain. But if the decommissioning of LORAN is any guide, we'll probably have access to VOR signals for some years to come.

In the event of a major space storm disabling a good deal of orbiting satellites (plausible, but not very likely), access to such terrestrial systems would become highly desired again.

Most VORs and NDBs are used for navigating from one station to another, or along specific VOR radials. The better GPS based systems allow one to specify any GPS waypoint for use as a "virtual VOR", whereby traditional VOR-style course-deviation-indication (CDI) and omnibearing-selector (OBS) controls allow you to precisely fly a given course to or from the waypoint. Such functions can be helpful when given random holding fixes to fly, for example. Today, however, much of the emphasis is on database navigation. Where any and all possible waypoints and procedures for flying between them are encoded in a database stored inside the GPS navigator, and you fly by querying the database for the route between desired waypoints and procedures for instrument approach. The navigator, if coupled to an autopilot, can then navigate the aircraft all the way through the approach with little for you to do but monitor and assure yourself it's going where you expected and intended.

With the newest "highway in the sky" features available as part of the synthetic vision systems of avionics like Garmin's G1000, flying complex procedures with curving approach and departure paths and tight glidepath and obstacle restrictions is as easy and ubiquitous as using the flight controls to guide the little computer airplane through the series of hoops on the video display. It's becoming literally as easy as a video game. Remember Pilotwings for the SuperNES? Or Independence War on the PC? Or the dropship sequence from Aliens on the C64? Like that.

You just have to be sure you programmed your navigator correctly before you start out!

This is one area where general aviation has advanced (temporarily) ahead of the airlines. Most airliners do not have avionics as advanced as a G1000 system with synthetic vision. A rare instance were the smaller guys were able to be more nimble, thanks to ever cheapening computer technology. The airlines are catching up, however, and G1000-type systems with synthetic vision are going now into business jets, with similarly featured systems planned for the latest iterations out of Boeing and Airbus.

That's all, thanks for reading! Hope you had fun!

Friday, May 7, 2010

Busted Trades

Fun times were had last Thursday (6MAY10) as the stock markets saw some really wild volatility occur in the mid-afternoon. At one point, the Dow was down around 1,000 points, the biggest intraday decline in the history of the index.

In the aftermath, there's a lot of blather going on about what exactly happened and who's responsible, as nobody believed the declines were real or that markets could really move that fast. Plenty of finger pointing as the heads of the highly competitive exchanges blame each other for the event.

The trading behavior and price movement of the markets are a collective emergent phenomenon, a large-scale effect of individual entities (large and small) behaving according to their own local rules (personal tradeplans, automated trading platform configuration, values of parameters, etc.). In this sense, the price action and movement of the indexes is more like the weather. Back away a bit, and you can get a general gist of what's happened and what may be likely to happen in the short term, but you just can't predict where bolts of lightening will strike, or when and where a tornado may form.

I don't think in any way that Thursday's price action was the result of any one thing, or even a small collection of things. Rather, I'm inclined toward the notion that it was probably a phenomenon that emerged when great masses of various individual conditions happened to line up just so, and trigger a storm. To me, that's what this felt like, a market storm.

By the end of the day, the market recovered much of the staggering afternoon meltdown, but was still down sharply. The NYSE Composite started that day at 7259, and took most of the day to trade down to 7000, then within 15 minutes it moved down, then back up, over 300 more points, touching 6667 at its lowest.

Apparently, the action in individual equities was quite something to see. Jim Cramer, on with Erin Burnett in his usual afternoon CNBC segment, reacted in a manner that was probably consistent with quite a few traders out there.

Jim watched P&G start to trade far away down from it's opening price, and failing to understand why such a price was justified, exclaimed that if he were still at his hedge fund, "Forty-nine and a quarter bid for fifty-thousand Procter" would be his order, and "Just buy it, who cares?"

Jim's a believer in the fundamental picture of P&G, that it's a good long-term type of stock to have in a retirement portfolio, and that it's also attractive due to it's relatively good dividend yield, make unnaturally juicy by Thursday's downward action. So on those merits, Jim's in there thinking, 'I don't know why it's down here, but it's a great company and WOW, THAT'S A LOW PRICE!'

As he talked over the course of the next couple of minutes, P&G traded through 49 to 42, and back to 59. By the end of his segment, he was saying, "I now flip it at 59 and I've just made 500Gs."

You might have been able to do that trade, if you were watching the market all day and right in front of a trading screen. I'm sure plenty of people did do precisely this sort of trade.

So, now comes the crux of why I decided to write this post. After-hours, word came down from the NASDAQ and also from NYSE that trades which executed 60% or more away from the last market print between 14:40 - 15:00 eastern time, would be canceled.

In the vernacular, this is known as "busting trades" and typically occurs when exchanges can point to obvious system breakdowns, corruption of price data, or intentional fraudulent market manipulation leading to faulty reporting of trade data.

Well, I suppose it may turn out to be the case that the first snowflake to start the avalanche was the result of intentional manipulation, but the way things are playing out, I highly doubt it.

Duncan Niederauer, CEO of the NYSE, speaking before the opening bell on CNBC Friday gave me some great food for thought about the implications of busting trades.

Picture Jim Cramer's hypothetical trading in PG the previous afternoon. Now, presumably he's watching the prices print on the "tape" and reacting in his own interest. He's playing completely open and by the book, barking a limit order for 50k PG at $49.25. Let's say he got filled, saw the tape print prices around $59-$61, as it did a few minutes later, and he flips with a new limit order, selling at $59 we'll say. Let's assume he got filled there too, and made his $500,000 profit.

It's certainly possible, and this very type of thing probably did happen to many traders Thursday night, that the first trade in that pair was found to conform to the "busting" criteria and reversed. However, the second trade in the pair, perhaps due to other executions at nearby prices happening within milliseconds of Cramer's hypothetical order, was deemed fair and not busted. What is the result Friday morning as you log onto your trading platform and await the opening bell?

Well, those 50k shares you sold did "sell", but the preceding "buy" never happened. Through no fault of your own, that $500,000 you thought you'd made Thursday afternoon has turned into a liability. You sold shares you never owned. You're short PG to the tune of $2.95 million worth of stock, and the market's about to reopen!

Talk about a shock! What do you do now? Do you buy back to cover the short right away, or wait a while to see if the stock resumes the prior day's downtrend? No time to think about it...here's the bell and PG's about to print is first trade, and it is: $60.61 (crap!)

You decide to sit on your hands a moment and call your broker and try to get some answers...you sit on the phone awhile, your broker is fielding hundreds of calls like yours perhaps. By the time you're off the phone at 10:20, you've learned that you in fact ARE short PG for 50k shares, and the last trade now is a throat-grabbing $61 (maybe other traders who awoke to find themselves unintentionally short are scrambling to cover...how many of them could there be? This might be bad if I don't act now, a real serious short-squeeze). You squelch further risk and buy-to-cover at $61. Rather than having a great Thursday $500,000 wealthier, you now have a disastrous Friday, having closed out your position (for real this time) for a $100,000 loss!

Who was at fault in this scenario? You played by the rules, placed your orders, and got filled according to exchange rules. How is it that for playing fair, you have to part with $100Gs? Now perhaps the decision by the exchanges to bust some of these trades save some people serious bacon. But these folks would've placed those orders eyes-open too, expecting them to execute if the conditions for their triggering were satisfied.

It seems to me, since someone is on the other side of every trade, that for every potential $500Gs-winning hypothetical Jim Cramer out there, there was a $500k loser (or collection of tiny losers, whose losses summed to $500k as they became part of the other side of Cramer's trade). Going in and intervening has the effect of converting some of the losers to winners, and some winners to losers.

The rule then ought to be, were the counter-parties acting in good faith, relying on the exchange to match their orders and execute their trades, was the data they were relying upon accurate, and did the system perform as designed? If so, the trade should stand. PERIOD.

From the evidence I've come into contact with in the aftermath of Thursday's action, there doesn't seem to be any malfeasance behind the price movements. It does in fact appear that all the price action was the direct result of the various exchanges acting precisely correctly (that is: precisely in keeping with SEC regulations, and their own published exchange rules for how trades and exchange data are to be handled and what procedures will be implemented when unusual price action develops). In short, the system behaved as designed. It did EXACTLY what we asked it to do, matching orders from buyers and sellers.

Exchanges and the market-makers and specialists at them, compete by trying to keep their assigned stocks liquid and trading. But this service comes with risk, and when the market starts to move at the same time that participants start to leave the venue, as a market-maker you have to be thinking about saving yourself at some point and cutting off your exposure to being stuck with stock at prices at which you cannot afford to trade.

On the human-side of the NYSE, this is exactly what happened. Volume began to trail off. The fastest price moment often happens when there are significantly fewer market participants, because within order books of bids and offers, the different price levels awaiting execution tend to get more spaced out and thinner. As this is happening, if someone comes in with a market order, particularly a large enough one, he'll start sweeping through those thin layers of supply at each booked price level very quickly. The exchange gets an order to sell 50k of XYZ at the market, and it's sending stock away to the first buyers it can see, starting at the highest bid on the order book. When that demand is used up, if supply in the original order still exists, the system sweeps down to the next highest bid, and so on until all the requested shares have been sold, or the buy side of the order book is used up completely.

This, in some sense, happened Thursday. The order books got thin, and consequently prices started to move about faster. (I think it's a common enough practice for many traders out there to put out speculative bids or offers far away from the last trade, "just in case" this exact sort of situation develops. If I have a new position in some stock, I'll frequently place a GTC sell limit order on at my target price, so I can be assured of booking the profit I'd hoped to realize, if the market trades up to me for ANY reason.)

On the NYSE there is a feature to help control fast markets known as liquidity replenishment points (LRPs). LRPs are trigger prices set into the order books for stocks at specific distances away from the last-traded price. They're there to control volatility such as we had on Thursday. The NYSE's designers recognized that high-speed data networks and electronic trading made it possible to execute trades so fast, no human could follow the price action. And "quant" systems are designed to exploit this fact, analyzing trade data and submitting orders on behalf of an individual or institution as fast as possible according to pre-programmed rules for behavior. These systems are human designed to try to make profits by trading quickly or according to complex dynamic criteria, and work because they can see and adapt to market conditions faster than humans can type in individual orders. It's a perfectly legitimate way to put technology to work for you.

The NYSE recognizes that one side-effect of this type of assistive technology is that unanticipated feedback mechanisms could become set-up as automated trading systems start to react to patterns generated by other automated systems. Market liquidity could suddenly dry up or flood in, and prices could start moving faster than humans could react to really understand the situation.

So when an LRP is triggered at the NYSE, the electronic execution temporarily flips into a slow, manual or auction-type mode to allow the human beings on the exchange floor, and at home or the office at their computers, time to see what's just happened and react as they think appropriate. It's a kind of "circuit-breaker" to ensure that markets don't move into absolutely absurd price territory for reasons nobody can make sense of, or have time to act upon. It's a good idea! This IS about the humans after all, and our technology is there to try and serve our interest.

And here's where I get to put my anti-regulation libertarian hat on: Thursday's market volatility was magnified by our very own SEC, and a recently new bit of regulation they imposed called "Reg. NMS" (for National Market System).

A few years ago, the SEC (in their porn-surfing-all-day kinda infinite wisdom), decided that investors needed to be "protected" from "unfair" price activity that might be happening on the exchange they're dealing with, when prices might be better outside that exchange, on another exchange or wholly electronic order matching system. The rule was hotly debated during its gestation, but the SEC planted its feet and dragged the exchanges along, who eventually acquiesced to the new order of things.

Reg NMS was supposed to protect market participants by seeking to ensure that the best displayed bid and offer (the top of the book) were protected and unable to be ignored if they came from an electronic system. Manual quotes and hidden orders are exempt. What that means is that if an order comes into the NYSE, but on the other side of the trade, another exchange is displaying a better price at the top of its displayed order book than exists on the NYSE order book, the incoming NYSE order must first "trade around" the NYSE to match with the order on the away exchange or system. If the incoming order wasn't fully satisfied by that away quote, then it can continue executing back on the NYSE, where it had been sent to begin with.

This was supposedly to protect smaller traders using brokers. Such traders don't interact directly with an exchange to execute their orders, and rely on market data typically presented in a consolidated stream representing the best bid and offer available within the entire market system, not just one particular exchange.

Before this rule, you might see an offer quoted on your trading platform, and this might prompt you to send out a bid to your chosen exchange (this is from a broker's perspective, from your perspective, your order goes into your broker's system). The exchange might have been able to execute your order completely from their own order book, and in many cases this would happen, ignoring (or trading-through) the quote you saw on your screen, as it was from another exchange. On the other side of things, the entity on the other exchange with the superior offer might wonder why he saw a trade print which executed at a price inferior to his offer. He's just been traded-through on a foreign exchange. Reg. NMS was designed to prevent this situation, at least for the displayed quotes at the tops of each market centers' order books.

So as the markets in certain stocks started to get volatile, at the NYSE, LRP points started to be hit. At NYSE, these stocks then flipped into manual or auction mode for a set time. Electronic execution stopped, and quotes from there were designated as slow. Specialists now have time to examine the market and decided if they're going to add or change orders on behalf of themselves or the institutions they're representing. Since automatic execution stopped, if pricing becomes superior at the NYSE compared to other market centers, these superior quotes will still get disseminated, but other market centers are permitted under Reg. NMS to "trade through" them if they wish. So if a new order comes into a different market center, that order might get traded through the NYSE's superior (but slow) quotes and execute on that market center fully. How this is controlled must depend on the order type that was sent to the market center. I believe it is the case that, using the proper order type, if you sent an order to a given market center away from the NYSE, and the NYSE was showing a superior price, you could control whether the order was forwarded over to the NYSE for slow execution, or executed directly at the center you'd sent it to, using the quotes on that center's order book.

Meanwhile, back at the NYSE, electronic orders coming in can start participating in an auction system which determines at which price the first execution will occur once the LRP halt time limit expires. In addition, floor specialists can manually execute trades at any time during the LRP halt period. Automatic electronic execution resumes immediately after such a manual execution.

So as trading volume diminished and liquidity bled away from certain stocks, the increase in their price volatility started to trigger LRPs at the NYSE, causing this market to slow down. This had the effect of isolating the other market centers' automatic electronic execution from the liquidity which might otherwise have been available at the NYSE. These market centers had some degree of choice of either continuing to match trades from their own order books, or, when the NYSE began showing superior quotes, route orders to it with the knowledge that execution will be slow (nothing will happen until a manual floor trade takes place, or failing any manual trades, until the auction price is determined by the end of the LRP halt period).

As I interpret Duncan Niederauer, it seems that most of the time, the other market centers chose to continue automatic electronic execution from their own books, eschewing the liquidity building up at the NYSE during the LRP periods and trading through NYSE's superior quotes. In one sense, these other market centers were preferring speed of execution and transaction rate to execution quality.

Choosing this path meant trades were executing off of thinner books with less available at each price point, and the result seems to have been very very rapid sweeps through those order books down to price points that, to a slow-acting human trader, would be completely baseless and silly. Because the all-electronic market was so fast and efficient, prices moved totally outside of a rational range before any human could recognize an opportunity was available and step in. Now, with last trades printing well beyond normal, such human traders on the sidelines are thinking..."is something wrong? Was that a real price? Perhaps I should just sit on my hands a moment."

Now, you can see why the NYSE's system of LRPs is such a good idea. It applies a brake when all-electronic algorithmically-driven trading causes the market to move more quickly than seems rational, giving the humans time to react, notice the prices, and start placing "bets" in a manner they think appropriate.

Duncan's thrust during his CNBC interview before the bell, was that pricing almost certainly would not have become so irrational if other market centers had policies similar to the NYSE's concept of LRPs, and that in any case, once LRPs were hit at the NYSE, movement would have been more rational if other market centers chose instead not to trade through superior NYSE quotes.

Reg. NMS worsened the situation once the NYSE isolated itself, by forcing the other electronic market centers to respect the superior top-book quotes of other electronic venues, which by this point were certainly all algorithmically generated orders, not based on human judgment or rationality.

Now, it's all algorithmic, all electronic, but it's also all still REAL. These are REAL trades, valid trades, placed by and crossed by computer programs at the request of their operators. Regardless whether the operators might have understood what their programs were up to exactly, these operators chose to run these programs to post these trades.

Finally, humans start reacting to the absurd last-trade prints coming in. The Jim Cramers out there are seeing the absurd prices and sensing opportunity. They step to their platforms, and place their trades, which also execute properly according to the rules of the system. Maybe they profit at the expense of the silly people who commanded their "quant" systems to trade on their behalf. Good! This is how a free market works! If algorithmic traders come by later and discover their beloved "quant" program did a bad bad thing to them, then they ought to take the pain and rethink their trading strategy a bit. It was a FAIR TRADE.

So who wins? Well from here, it appears the industry organizations which determine if a trade ought to be busted clearly stepped in to bail out the users of algorithmic systems...perhaps at the very very dear expense of later-reacting human traders who, in good faith, and in accordance with published rules, placed their own trades seeking to profit from what they rationally saw were stocks priced for opportunity. It IS the case that ANY TIME a trade is placed, the traders are placing orders because they think the price is wrong and will move a certain way. They place the trade with an aim to make the move toward the price they think is more correct yield a profit for them. If this were not so, NO TRADES would ever happen.

So just because the price seems by some vague unquantifiable measure more wrong than usual, the regulating agency is going to bust trades made honestly and intentionally?

If it comes to light that there was real fraud intentionally committed by some entity or group out there, then I will feel less indignant. The Jim Cramer hypothetical still would get no satisfaction, as he would've been acting in good faith to data presented as valid and trustworthy. If you busted his buy, the only proper solution would then be to also bust his later sell, and follow those shares and bust every other single trade involving them from that point forward to the close (a daunting data mining challenge).

But I am thinking that it will be the case that there will be no boogeyman found on this, no actual fraud committed, and no actual corruption of data discovered from some sort of system fault. In the end, my money is on the notion that the system operated precisely the way it was set up to operate, only that in this instance, the system's configuration led to a period of unexpected behavior. That might be a necessary condition to start going in and busting trades, but I argue it is not a sufficient one.

To any traders from Thursday who acted openly, seeking as always to profit from prices they think are wrong, and wound up in a position similar to my hypothetical Cramer scenario, losing serious money solely because a regulator thought the price movement was somehow too wrong to be real, I feel for you.

It's been suggested that perhaps a large institutional trader entered into a large trade or started a significant algorithmic process in error (the fat fingers excuse) and never intended to enter that trade or start that process, and the price avalanche was the direct result of a cascade of effects occurring as a result of that unintentional action. If that scenario turns out to be the case for this mess, then tough bounce for that guy! If I setup an order with my broker, mistype a number on the computer form, and before realizing my mistake, I push the send button, is the broker going to make me whole? NO WAY! It was my bad, my error, and I'll have to live with it. Perhaps it's not too late and I can get in quick and hit the cancel button on that order. Maybe that works, but brokers are always careful to point out that canceled trades are subject to prior execution, the news of which might still have been making its way toward your terminal while you were pushing the cancel button. If so, too bad, they say. And we accept this before choosing to use the platform. Why should Mr. Fat Fingers, in this instance, get a pass and see trades busted in his favor?

It was interesting to me to see, Thursday night, notice come down from the exchanges about some industry-wide consensus to go on this trade busting plan, followed quickly with the statement that there would be no appeals. How could you be so sure so soon that no one would have valid grounds to appeal unless you know the cause of the problem? You can't. A statement on Friday from the SEC included a closing comment I found to be quite telling:

"Market clearance and settlement processes functioned well and without incident."

Translation: the system behaved as it had been designed to behave. Now that to me suggests we can rule out a system fault or data corruption as causal.

These trades are what they are. If no explicit fraudulent actions or direct data corruption due to system fault is found, then shame on you, regulator. Your meddling might be what we really need protection from.