As time went by, CNN trimmed its expert pool down to a dozen or so regulars who earned the on-air title "CNN aviation analysts": airline pilots, ex-government honchos, aviation lawyers and me. We were paid by the week, with the length of our contracts dependent on how long the story seemed likely to play out. The first couple were for seven days, the next few were for 14 days and the last one ran for a month. We'd appear solo, in pairs or in larger groups for panel discussions - whatever it took to vary the rhythm of perpetual chatter.

I soon realised the germ of every television-news segment is: "Officials say X". The validity of the story derives from the authority of the source. The expert, such as myself, is on hand to add dimension or clarity. Truth flowed one way: from the official source, through the anchor, past the expert and onward into the great sea of viewerdom.

What made MH370 challenging to cover was, first, that the event was unprecedented and technically complex and, second, that the officials were remarkably untrustworthy. For instance, the search started over the South China Sea, naturally enough, but soon after, Malaysia opened up a new search area, in the Andaman Sea, 650km away. Why? Rumours swirled that military radar had seen the plane pull a 180-degree turn. The Malaysian government explicitly denied it but, after a week of letting other countries search the South China Sea, the officials admitted they'd known about the U-turn from day one.

Of course, nothing turned up in the Andaman Sea, either. But in London, scientists at British company Inmarsat, which provides telecommunications between ships and aircraft, realised its database contained records of transmissions between MH370 and one of its satellites over the seven hours after the plane's main communications system shut down. Seven hours! Maybe it wasn't a crash after all - if it were, it would have been the slowest in history.

These electronic "handshakes" or "pings" contained no actual information but by analysing the delay between the transmission and reception of the signal - called the "burst timing offset" or BTO - Inmarsat could tell how far the plane had been from the satellite and thereby plot an arc along which the plane must have been at the moment of the final ping. That arc stretched nearly 10,000km but, if the plane was travelling at normal airliner speeds, it would most likely have wound up around the ends of the arc - either in Kazakhstan or China in the north or the Indian Ocean in the south. My money was on Central Asia but CNN quoted unnamed United States-government sources saying that the plane had probably gone south, so that became the dominant view.

Other views were circulating, too. A Canadian pilot named Chris Goodfellow went viral with his theory that MH370 suffered a fire that knocked out its communications gear and diverted from its planned route in order to attempt an emergency landing. Keith Ledgerwood, another pilot, proposed that hijackers had taken the plane and avoided detection by ducking into the radar shadow of another airliner. Amateur investigators pored over satellite images, insisting that wisps of cloud or patches of shrubbery were the lost plane. American singer Courtney Love, posting on her Facebook timeline a picture of the shimmering blue sea, wrote: "I'm no expert but up close this does look like a plane and an oil slick."

Then: breaking news! On March 24, Malaysian Prime Minister Najib Razak announced that a new kind of mathematical analysis proved the plane had, in fact, gone south. This new analysis involved another aspect of the handshakes called the burst frequency offset, or BFO, a measure of changes in the signal's wavelength, which is partly determined by the relative motion of the airplane and the satellite. That the whole southern arc lay over the Indian Ocean meant all the passengers and crew would certainly be dead by now. This was the first time in history that the families of missing passengers had been asked to accept that their loved ones were dead because a secret maths equation said so. Not all took it well. In Beijing, outraged next-of-kin marched to the Malaysian embassy, where they hurled water bottles and faced down paramilitary soldiers in riot gear.

Guided by Inmarsat's calculations, Australia, which was coordinating the investigation, moved the search area 1,100km to the northeast, to a 319,000 sq km patch of ocean west of Perth. Ships and planes found a lot of debris on the surface, provoking a frenzy of breaking news banners, but all of it turned out to be junk. Adding to the drama was a ticking clock. The plane's two black boxes had an ultrasonic sound beacon that sent out acoustic signals through the water. (Confusingly, these were also referred to as "pings", though of a completely different nature. These new pings suddenly became the important ones.) If searchers could spot plane debris, they'd be able to figure out where the plane had most likely gone down, then trawl with underwater microphones to listen for the pings. The problem was that the pingers had a battery life of only 30 days.

On April 4, with only a few days' pinger life remaining, an Australian ship lowered a special microphone called a towed pinger locator into the water. Miraculously, the ship detected four pings. Search officials were jubilant, as was the CNN green room. Everyone was ready for an end to the story.

The only problem was me. I pointed out that the pings were at the wrong frequency and too far apart to have been generated by stationary black boxes. For the next two weeks, I was the odd man out on Don Lemon's six-guest panel blocks, gleefully savaged on-air by my co-experts.

The Australians lowered an underwater robot to scan the seabed for the source of the pings. There was nothing. Of course, by the rules of TV news, the game wasn't over until an official said so. But things were stretching thin. One night, an underwater-search veteran taking part in a Don Lemon panel agreed with me that the so-called acoustic-ping detections had to be false. Backstage after the show, he and another aviation analyst nearly came to blows. "You don't know what you're talking about! I've done extensive research!" the analyst shouted. "There's nothing else those pings could be!"

Soon after, the story ended the way most news stories do: we just stopped talking about it. A month later, long after the caravan had moved on, a US Navy officer said publicly that the pings had not come from MH370. The saga fizzled out with as much satisfying closure as the final episode of .

Meanwhile, a core group of engineers and scientists, myself included, had split off via email. We called our-selves the Independent Group, or IG. If you found your-self wondering how a satellite with geosynchronous orbit responds to a shortage of hydrazine, all you had to do was ask. The IG's first big break came in late May, when the Malaysians finally released the raw Inmarsat data. By combining the data with other reliable information, we were able to put together a timeline of the plane's final hours: 40 minutes after taking off from Kuala Lumpur, MH370 went electronically dark. For about an hour after that, the plane was tracked on radar following a zigzag course and travelling fast. Then it disappeared from military radar. Three minutes later, the communications system logged back onto the satellite. This was a major revelation. It hadn't stayed connected, as we'd always assumed. This event corresponded with the first satellite ping. Over the course of the next six hours, the plane generated six more handshakes as it moved away from the satellite.

The final handshake wasn't completed. This led to speculation that MH370 had run out of fuel and lost power, causing the plane to lose its connection to the satellite. An emergency power system would have come on, providing enough electricity for the satcom to start reconnecting before the plane crashed. Where exactly it would have gone down was still unknown - the speed of the plane, its direction and how fast it was climbing were all sources of uncertainty.

The MH370 obsessives continued attacking the problem. Since I was the proprietor of the major web forum, it fell on me to protect the fragile cocoon of civility that nurtured the conversation. A single troll could easily derail everything. The worst offenders were the ones who seemed intelligent but soon revealed themselves as Believers. They'd seized on a few pieces of faulty data and convinced themselves that they'd discovered the truth. One was sure the plane had been hit by lightning and then floated in the South China Sea, transmitting to the satellite on battery power. When I kicked him out, he came back under aliases. I wound up banning anyone who used the word "lightning".

By October, officials from the Australian Transport Safety Bureau had begun an ambitiously scaled scan of the ocean bottom, and, in a surprising turn, it would include the area suspected by the IG. For those who'd been a part of the months-long effort, it was a thrilling denouement. The authorities, perhaps only coincidentally, had landed on the same conclusion as had a bunch of random people from the internet. Now everyone was in agreement about where to look.

While jubilation rang through the email threads, I nursed a guilty secret: I wasn't really in agreement. For one, I was bothered by the lack of plane debris. And then there was the data. To fit both the BTO and BFO data well, the plane would need to have flown slowly, likely in a curving path. But the more plausible autopilot settings and known performance constraints would have kept the plane flying faster and more nearly straight south. I began to suspect that the problem was with the BFO numbers - that they hadn't been generated in the way we believed. If that were the case, perhaps the flight had gone north after all.

For a long time, I resisted even considering the possibility that someone might have tampered with the data. That would require an almost inconceivably sophisticated hijack operation, one so complicated and technically demanding that it would almost certainly need state-level backing. This was true conspiracy-theory material.

And yet, once I started looking for evidence, I found it. One of the commenters on my blog had learned that the compartment on 777s called the electronics-and-equipment bay, or E/E bay, can be accessed via a hatch in the front of the first-class cabin. If perpetrators got in there - a long shot - they would have access to equipment that could be used to change the BFO value of its satellite transmissions. They could even take over the flight controls.

I realised I already had a clue that hijackers had been in the E/E bay. Remember the satcom system disconnected and then rebooted three minutes after the plane left military radar behind? I spent a great deal of time trying to figure out how a person could physically turn the satcom off and on. The only way, apart from turning off half the entire electrical system, would be to go into the E/E bay and pull three particular circuit breakers. It is a manoeuvre that only a sophisticated operator would know how to execute, and the only reason I could think for wanting to do this was so that Inmarsat would find the records and misinterpret them. They turned on the satcom in order to provide a false trail of breadcrumbs leading away from the plane's true route.

It's not possible to spoof the BFO data on just any plane. The plane must be of a certain make and model, equipped with a certain make and model of satellite-communications equipment, and flying a certain kind of route in a region covered by a certain kind of Inmarsat satellite. If you put all the conditions together, it seemed unlikely that any aircraft would satisfy them. Yet MH370 did.

I imagine everyone who comes up with a new theory, even a complicated one, must experience one particularly delicious moment, like a perfect chord change, when disorder gives way to order. This was that moment for me. Once I threw out the troublesome BFO data, all the inexplicable coincidences and mismatched data went away. The answer became wonderfully simple. The plane must have gone north.

Using the BTO data set alone, I was able to chart the plane's speed and general path, which happened to fall along national borders. Flying along borders, a military navigator told me, is a good way to avoid being spotted on radar. A Russian intelligence plane nearly collided with a Swedish airliner while doing it over the Baltic Sea in December. If I was right, it would have wound up in Kazakhstan, just as search officials recognised early on.

There aren't a lot of places to land a plane as big as the 777, but, as luck would have it, I found one: a place just past the last handshake ring called Baikonur Cosmodrome. Baikonur is leased from Kazakhstan by Russia. A long runway there called Yubileyniy was built for a Russian version of the Space Shuttle. If the final Inmarsat ping rang at the start of MH370's descent, it would have set up nicely for an approach to Yubileyniy's runway 24.

Whether the plane went to Baikonur or elsewhere in Kazakhstan, my suspicion fell on Russia. With technically advanced satellite, avionics and aircraft-manufacturing industries, Russia was a paranoid fantasist's dream. (The Russians, or at least Russian-backed militia, were also suspected in the downing of Malaysia Airlines Flight 17 in July.) Why, exactly, would Vladimir Putin want to steal a Malaysian passenger plane? I had no idea. Maybe he wanted to demonstrate to the US, which had imposed the first punitive sanctions on Russia the day before, that he could hurt the West and its allies anywhere in the world. Maybe what he was really after were the secrets of one of the plane's passengers. Maybe there was something strategically crucial in the hold. Or maybe he wanted the plane to show up unexpectedly somewhere someday, packed with explosives. There's no way to know. That's the thing about MH370 theory-making: it's hard to come up with a plausible motive for an act that has no apparent beneficiaries.

As it happened, there were three ethnically Russian men aboard MH370, two of them Ukrainian-passport holders from Odessa. Could any of these men, I wondered, be special forces or covert operatives? As I looked at the few pictures available on the internet, they definitely struck me as the sort who might battle Liam Neeson in mid-air.

About the two Ukrainians, almost nothing was avail-able online. I was able to find out a great deal about the Russian, who was sitting in first class about 15 feet from the E/E-bay hatch. He ran a lumber company in Irkutsk, and his hobby was technical diving under the ice of Lake Baikal. I hired Russian speakers from Columbia University, in New York, to make calls to Odessa and Irkutsk, then hired researchers on the ground.

The more I discovered, the more coherent the story seemed to me. I found a peculiar euphoria in thinking about my theory, which I thought about all the time. One of the diagnostic questions used to determine if you're an alcoholic is whether your drinking has interfered with your work. By that measure, I definitely had a problem. Once the CNN cheques stopped coming, I entered a long period of intense activity that earned me not a cent. Instead, I was forking out my own money for translators and researchers and satellite photos. And yet I was happy.

Still, it occurred to me that, for all the passion I had for my theory, I might be the only person in the world who felt this way. Neurobiologist Robert A. Burton points out in his book that the sensation of being sure about one's beliefs is an emotional response separate from the processing of those beliefs. It's something the brain does subconsciously to protect itself from wasting unnecessary processing power on problems for which you've already found a solution that's good enough.

"'That's right' is a feeling you get so that you can move on," Burton told me. It's a kind of subconscious laziness. Just as it's harder to go for a run than to plop onto the sofa, it's harder to re-examine one's assumptions than it is to embrace certainty. At one end of the spectrum of sceptics are scientists, who by disposition or training resist the easy path; at the other end are conspiracy theorists, who'll leap effortlessly into the sweet bosom of certainty. So where did that put me?

Propounding some new detail of my scenario to my wife over dinner one night, I noticed a certain glassiness in her expression. "You don't seem entirely convinced," I suggested.

She shrugged.

"OK," I said. "What do you think is the percentage chance that I'm right?"

"I don't know," she said. "Five per cent?"

Springtime came to the southern ocean and search vessels began their methodical cruise along the area jointly identified by the IG and the ATSB, dragging behind it a sonar rig that imaged the seabed in photographic detail. Within the IG, spirits were high. The discovery of the plane would be the triumphant final act of a remarkable underdog story.

By December, when the ships had still not found a thing, I felt it was finally time to go public. In six sequentially linked pages that readers could only get to by clicking through - to avoid anyone reading the part where I suggest Putin masterminded the hijack without first hearing how I got there - I laid out my argument. I called it "The Spoof".

I got a respectful hearing but no converts among the IG. A few sites wrote summaries of my post. The online International Business Times headlined its story "MH370: Russia's Grand Plan to Provoke World War III, Says Independent Investigator" and linked directly to the Putin part. Somehow, the airing of my theory helped quell my obsession. My gut still tells me I'm right, but my brain knows better than to trust my gut.

Last month, the Malaysian government declared that the aircraft is considered to have crashed and all those aboard are presumed dead. Malaysia's transport minister told a local television station that a key factor in the decision was the fact that the search mission for the aircraft failed to achieve its objective. Meanwhile, new theories are still being hatched. One, by French writer Marc Dugain, states that the plane was shot down by the US because it was headed toward the military bases on the islands of Diego Garcia as a flying bomb.

The search failed to deliver the plane, but it has accomplished some other things: it occupied several thousand hours of worldwide airtime; it filled my wallet and then drained it; it torpedoed the idea that the application of rationality to plane disasters would inevitably yield ever-safer air travel. And it left behind a faint, lingering itch in the back of my mind, which, I believe, will quite likely never go away.