Blair: Buy

Current price:
bid 0, ask 3, last 1
Due date:
27 November 2008
Recommendation:
Sell.
Comment:
Today, Tony Blair announced that «the next party conference in a couple of weeks will be my last party conference as party leader.» Since the next party conference after that will be held about a year from now, this means that he will quit before September 2007. That date is more than a year before the date set in this claim for it to true, and as a consequence, it may be expected that this claim will be judged False early within one year from now.
My current holdings in this claim: 103

BgEU: Buy

Current price:
bid 91, ask 92, last 92
Due date:
1 January 2011
Recommendation:
Buy.
Comment:
Last week, the European Commission gave a «qualified yes» to Bulgaria and Romania to join the European Union on 1 January 2007. The final decision is postponed until October later this year, but if the two countries don't join the European Union in 2007, they will probably join in 2008, still long before the deadline of 1 January 2011 used in this caim.
My current holdings in this claim: 109

The famous 50 cent bias

The graph on the right shows the historic trading prices for Poisson 1 claims that had 0 as the default price. This type of claims is to be judged false, unless something happens that makes it true. Examples for this type of claims range from earthquakes and meteorites to diseases or the outbreak of wars. Claims on election results do not belong to this category, because in the set of claims dealing with one election, one claim has to become true because one candidate eventually will be elected. The 153 claims used to produce the graph are listed at the end of this article.

The green line in the graph plots for every price what the chances were a claim was judged true if at least one transaction at that price was made. The violet line does the same thing, but weights the frequencies of a claim coming true by the number of transactions made at each price. The blue line is the most usefull, because it weights the frequencies by the volume at which a claim was traded at each price. Finally, the red line shows what the results would have been if FX had been perfect, that is, if the price of a claim reflected its probability.

The graph shows that the probabilities of this type of claims in general is being overestimated, that is, a claim has a lesser probability of coming true than its price would suggest, right up to 70 cents. Actually, there is one exception, 47 cents, were the frequency weighted by volume was measured to be 52,82%. Above 70 cents, the prices seem to be about right according to the probabilities, the exceptions being the prices from 87 to 90 cents.

However, one should keep in mind that the number of transactions and the volume of trades decreases for increasing prices, and that they are spread over fewer claims. This is a natural consequence of the simple fact that by nature, these claims tend to become false: only 27 out of the 153 selected claims became true. The graph on the left shows for each price the number of claims that had a transaction at the given price (green), the number of transactions at the given price (violet), and the volume. As can be clearly seen, the green line starts above 100 at the low price end, and goes down to around 20-30 at the high price end. As a result, the results in the first graph are not so reliable for the higher prices as for the lower prices. Also notice the spikes at the 5 cent marks.

So is there a 50 cent bias? I don't think so. If there would have been a 50 cent bias for this type of claims, prices should underestimate the probabilities above 50 cents. The graph however shows clearly that FX keeps overestimating the probabilities until 70 cent, and is rather realistic above that price. Actually, the graphs for Poisson 1 (100) claims, which are claims that are default true unless something happens, show more or less the dual situation: prices are rather realistic until 20 cents, and above that price an underestimation of the probabilities of the claims. Underestimating the probabilities of a Poisson 1 (100) claim is the same as overestimating the probabilities of the dual Poisson 1 (0) claim. My conclusion is that in general, FX overestimates the chances of «events happening» not until a 50% probability, but rather a 75% probability.

The list of claims used to produce the graph, with the result for each claim in superscript:
$bill⁰ $vIF⁰ 000D⁰ 1e7Pri⁰ 2e6Pri¹⁰⁰ 2KIRA⁰ 3UPP⁰ 6000⁰ abdt⁰ ACrash⁰ ADED⁰ alin⁰ Am30⁰ Amig⁰ As97⁰ ATrB⁰ AusR⁰ AV96⁰ B5S4¹⁰⁰ B5S5¹⁰⁰ Bilary⁰ BlnS¹⁰⁰ BlOGl2¹⁰⁰ BlOGlb⁰ Bnkrpt¹⁰⁰ btur⁰ Bust⁰ CADE⁰ Cash⁰ ccar¹⁰⁰ ChBr⁰ Ches¹⁰⁰ chnspc⁰ ChUS⁰ Clark3⁰ CLUE⁰ CNET⁰ CNTX⁰ Condit⁰ CPfP⁰ CPPr⁰ Cr56⁰ CRAK2⁰ CrpX⁰ Cryo⁰ Defect¹⁰⁰ DENG⁰ DKfg⁰ DMIA⁰ DNAT⁰ DOCX⁰ dow10k¹⁰⁰ DOW4⁰ dow7k⁰ DRAB⁰ DRD2⁰ DSpace⁰ Ellen¹⁰⁰ EMVR⁰ Epdm⁰ eUmp⁰ euro¹⁰⁰ Fact⁰ Ferm¹⁰⁰ FLT2⁰ FSTX⁰ GAPE⁰ GPSSA⁰ H1B⁰ He25⁰ HIVV⁰ HMOS⁰ Host97⁰ IDEA⁰ Infr⁰ InJo4⁰ iNSP⁰ io96¹⁰⁰ IReg⁰ ISDN⁰ JKIA¹⁰⁰ JRyan9¹⁰⁰ KorWar⁰ Kryp97⁰ LunH2O⁰ LUpriv⁰ MacC⁰ MdCd⁰ MelX⁰ Mini⁰ MLBM⁰ MMMI⁰ Mov$⁰ MSsetl⁰ N14k⁰ NASA⁰ ncryp⁰ NDSen¹⁰⁰ NIPC⁰ Nipl¹⁰⁰ NoJobs⁰ NSF$⁰ ojia⁰ OJvn⁰ OLAT⁰ os2a⁰ OS2X⁰ PClone⁰ PkNuke¹⁰⁰ PlsCom⁰ Polio⁰ Prim¹⁰⁰ RC564⁰ RoarX¹⁰⁰ rocket⁰ roton⁰ Rsgn⁰ RTON⁰ RUnew¹⁰⁰ ryldth⁰ SGAJ⁰ Smoke⁰ Smoke2⁰ SolC23⁰ solc23¹⁰⁰ SPat⁰ SPIN⁰ SShX⁰ sslk⁰ SSma⁰ SSTO⁰ Stew⁰ SWIV⁰ T2007X¹⁰⁰ Tach⁰ TAXX⁰ TBOT⁰ Terr⁰ Terr02⁰ tinpot⁰ TREU⁰ UFOx⁰ UKGE⁰ UNIX⁰ USIraq¹⁰⁰ USTerr¹⁰⁰ Veri⁰ X620¹⁰⁰ Y2Kcsh⁰ Y2Kdrf⁰ Y2Kdth⁰ Y2Kstk⁰ Y2Kwar⁰

Marbrg: Sell

Current price:
bid 10, ask 25, last 27
Due date:
15 April 2006
Recommendation:
Sell.
Comment:
On 24 August 2005, WHO reported that there had been 329 deaths, with the last confirmed death on 21 July. It seems therefore that the outbreak has come to a halt, and that it has become very unlikely that the number of 1000 deaths will be reached by April 2006.
My current holdings in this claim: 0

EUCYes: Sell

Current price:
bid 8, ask 10, last 8
Due date:
1 January 2010
Recommendation:
Sell.
Comment:
Since the previous post about EUCYes, not a single opinion poll in France has shown a majority for Yes. In addition, recent polls in the Netherlands are indicating that the result of the Dutch referendum three days later may result in a No too.
My current holdings in this claim: -110.

Arinze: Sell

Current price:
bid 6, ask 7, last 6
Due date:
1 March 2010
Recommendation:
Sell.
Comment:
As reported earlier, this claim will be resolved in the coming days. Paddy Power is running bets on who will be the next pope, and they are currently taking bets at 1:8 for Arinze. If you divide 1:8 by the sum of all odds (2.645), you get a probability of 4.7%, which is below the current bidding price.
My current holdings in this claim: -107

Arinze: Sell

Current price:
bid 10, ask 13, last 10
Due date:
1 March 2010
Recommendation:
Sell.
Comment:
Pope John Paul II died on 2 April 2005, and a conclave to elect a new pope will gather in less than 20 days. Arinze is often mentioned as a good candidate, but his chances are usually estimated to be less than 10%, often even lower. Many observers of the Vatican have suggested that a transition figure may be elected, and that would rather be an older Italian cardinal.
My current holdings in this claim: -114