This morning, a single buyer swamped the robust Tradesports Hillary-nomination market from 54.5% to 68.5% in one massive buy.

I found this to be incredibly interesting on two levels. Firstly, the notion that someone who has presumably accumulated decent buying power believes any piece of information, at this point in the nominations game, would justify that kind of movement. Secondly, the staying power of speculative attacks on event market prices.

Beyond planned announcements of all major non-Hillary contenders (Obama, Edwards, etc) dropping out of the race (all but impossible), there is nothing that would, on its face, justify a 14-point jump for Hillary. The only major piece of information I’ve retrieved today is Gore’s “no plans, and no more new ways to deny that I’m running” refusal to run in 2008. To the extent that Gore would have had any influence whatsoever, he would have splintered the anti-Hillary field on a base of liberal loyalty to Gore and disenchantment with Hillary. His latest announced refusal to run only bolsters the remaining anti-Hillarys. So then, one wonders, what private information could the anonymous bettor have that would lead him to believe the above move to be a profitable one?

Therein lies an issue that prediction markets seem to have some trouble dealing with, in many cases (although apparently not this one so much). A speculative attack, as I see it, is a major move by a major player who senses that other major players are very long or short on a position, and deeply uneasy about it. Making a big push is a decisive challenge to half of the market to either add to a losing position, or start “puking” its positions and create a snowball effect. It creates an equal, opposite snowball effect for the other half of the market to pile onto its winning positions–if, that is, the move is convincing enough to cause others to join the groundswell, on both the gaining side and the losing side. However, if the market doesn’t follow into lockstep, the attacker ends up with a significant loss and a huge risk of much greater loss… It’s like a many-handed poker game that’s going normally, and is suddenly upset by a normally passive big stack (the attacker)’s huge raise. Suddenly, the ball is in the court of the smaller players (who are often now in the red, and unwilling to add to a losing position) to judge the credibility of the sudden raise.

The reason this attack was so interesting, and anomalous, was that it seemed so inherently un-credible. Considering how far away the nomination is, the enormous demand in the Democrats’ political economy for an anti-Hillary, and the deepness of their field, nobody’s announcement was going to significantly change the current dynamics of the race over the medium-long run. What did the attacker think he was getting out of this?

I launched, watched, and recorded the results of a small-scale speculative attack on the SENATE.GOP.2006 market myself, here and here. But I timed mine to coincide with the release of public data that, while it didn’t completely validate the market swing, showed some movement in my direction–making the possible existence of explosive, unreleased, private information that much more plausible. And it seemed to work–for about 24 hours after my attack, the price of the security traded at about eight points (fell from average of 72 to 64, with a low of about 58) lower than it had previously, and at several times the volume I’d risked to push the market down. Private data also have a much higher “fear factor” three weeks out from the event, rather than a year. Anything “explosive” now could just as easily be overturned by some equally inordinary event or mood shift three or six months down the road.

Event contracts, at first glance, seem more vulnerable to this kind of attack, because they are aggregations of so much diffuse information, and thus the most difficult to value; when a big player pushes, stating that his information aggregates in a certain contrarian way, it forces other players to re-aggregate, but with a huge question mark about the private data that the attacker presumably has, and precludes their new aggregation being decisive enough to cause them to take a position against the attacker. But while a big player’s move does cause large position losses and unease among smaller traders, risk aversion (wanting to dump the position) comes up against another powerful human instinct–the reluctance to admit a mistake. And the larger amount of data is required to build an aggregate probability, the easier it is to take comfort in the data’s ambiguities and not have to admit a mistake.

Which one outweighs which under a speculative attack is ultimately determined by how long the smaller players need to stare at how big of a loss on their ledger, in my experience. If a player is big enough to push a contract hugely one way or another, and create an impression of stabilization at the new level by being an informal market-maker at that level (creating a new bid-ask spread at the new, lower level), he can cause a series of recent trades to occur at the new lower valuation–thus fostering the impression the market will have arrived at a new consensus.

Does a security market become more prone to economies of scale (via attacks), the more data are required to effectively aggregate the likelihoods of the respective outcomes?

(Cross-posted at the TS Maven)

–Alex