Free Money at PredictIt?

Link post

Pre­vi­ously on Pre­dic­tion Mar­kets (among oth­ers): Pre­dic­tion Mar­kets: When Do They Work?, Sub­si­diz­ing Pre­dic­tion Markets

Epistemic Sta­tus: No huge new in­sights, but a lit­tle fun, a lit­tle free money, also Happy Petrov Day?

Yes­ter­day, with ev­ery­thing hap­pen­ing re­gard­ing im­peach­ment, I de­cided to check Pre­dic­tIt to find out how im­pact­ful things were. When I checked, I no­ticed some ob­vi­ous in­con­sis­ten­cies. They’re slightly less bad to­day, but still not gone.

I figured it would be fun and po­ten­tially en­light­en­ing to break this down. Be­fore I be­gin, I will state that un­less I messed up this post ex­presses zero poli­ti­cal opinions what­so­ever on what elec­tion or other out­comes would be good or bad, and does its best to only make what I con­sider very safe ob­ser­va­tions on prob­a­bil­ities of events. All com­ments ad­vo­cat­ing poli­ti­cal po­si­tions or can­di­dates will be deleted in reign-of-ter­ror style. No ex­cep­tions.

Mar­ket Analysis

Odds are rep­re­sented as cost in cents for each con­tract that pays $1, so they dou­ble as prob­a­bil­ities out of 100.

Let us look at the demo­cratic nom­i­na­tion odds, us­ing last. All are 1 cent wide:

Eliz­a­beth War­ren 50

Joe Bi­den 21

An­drew Yang 10

Bernie San­ders 8

Pete But­tigieg 6

Hillary Clin­ton 5

Ka­mala Har­ris 4

Tulsi Gab­bard 3

Amy Klobuchar 2

Can be sold for 1: Corey Booker, Tom Sny­der, Beto ’o Rourke

All other can­di­dates can be bought for 1 and can­not be sold.

Ad­ding that up we get 112. We could buy all the no sides for a to­tal of 111 – you can get these prices on no ex­cept for War­ren, where you’d sell at 49.

That’s cer­tainly some free money. If you sell all of them, you don’t tie up any money, al­though you do have to de­posit, so it’s a pretty great trade, albeit with an $850 limit.

I’ve already done many of the legs of that trade. Some are bet­ter than oth­ers. Hillary Clin­ton at 5 is com­plete in­san­ity. An­drew Yang is trad­ing at 10 be­cause in­ter­net. That likely cov­ers most of the rea­son you can sell the field for 111. Lower them to sane num­bers (let’s be su­per gen­er­ous and say Hillary Clin­ton 1, and say An­drew Yang 5) then the field would add to 102. Com­plet­ing the trade is mostly about free­ing up your cap­i­tal. You also get some value for it be­ing some­one not on the above list, as the ‘bro­kered con­ven­tion causes weird­ness’ sce­nario is definitely not im­pos­si­ble. The weird thing is ex­pect­ing that to some­how nom­i­nate Hillary Clin­ton.

The big not-au­to­mat­i­cally-in­sane opinion is mak­ing War­ren 50% to win the nom­i­na­tion. That is rather bold at this stage of things, but we’re think­ing about ar­bi­trage and out­right mis­takes.

Let’s now look at the Pres­i­den­tial odds. For any Demo­crat, this is al­most iden­ti­cal to a two-part bet, where that per­son wins the nom­i­na­tion and then wins the gen­eral elec­tion.

Don­ald Trump 41

Eliz­a­beth War­ren 35

Joe Bi­den 13

An­drew Yang 6

Bernie San­ders 6

Pete But­tigieg 3

Nikki Haley 2

Ka­mala Har­ris 2

Mike Pence 2

Tulsi Gab­bard 2

Corey Booker 1

Amy Klobuchar 1

That adds up to 114. If you look at ac­tu­ally available prices, you could sell the field for 110. Again, pretty good idea. I’d get on that, and I mostly did.

One could also point out the im­plied gen­eral elec­tion win per­centages of democrats where round­ing isn’t a big deal.

War­ren 70%

Bi­den 62%

Yang 60%

San­ders 75%

Sum of All Repub­li­can odds is 45% (Trump, Haley and Pence) out of 114%, for odds of 39.4%. Thus, Demo­cratic vic­tory should be about 60%. War­ren is 50% to win the nom­i­na­tion, so that 70% num­ber is re­ally weird. This does not add up, and makes me re­luc­tant to sell War­ren at 50% odds in the pri­mary.

In both these cases, the free money seems real enough. You get to use your cap­i­tal in both mar­kets if you sell the whole field and then have it free for a third mar­ket as well, and you can’t re­ally lose. Doesn’t mean it’s worth the effort, but it’s a nice thing to no­tice.

Let’s look at Repub­li­can nom­i­na­tion odds:

Trump 78

Haley 7

Pence 7

Ka­sich 2

Rom­ney 2

Weld 2

San­ford 2

That only adds up to 98%, which makes sense, since if Trump is ac­tu­ally gone then any­thing could hap­pen. This mar­ket seems sane on that level, per­haps even rich. What’s most in­ter­est­ing is that Trump is highly un­likely to not win the Repub­li­can nom­i­na­tion and win the pres­i­dency, so if he’s 41% to be re­elected but 78% to be nom­i­nated, then Trump has a gen­eral elec­tion win rate of 52% (47% if we knock off 10% for the mar­ket be­ing in­flated by adding up to 114%). But per­haps this is rea­son­able? If Trump is gone it’s be­cause some­thing brought him down so it’s go­ing to be su­per hard for any­one else to win? It’s not like much of that prob­a­bil­ity is that Trump’s health fails, given the time frame.

Also note­wor­thy is Trump is only 20% to win the pop­u­lar vote, al­though the available vol­ume here is very low. That im­plies a stun­ning 21% chance that Trump loses the pop­u­lar vote but wins the elec­tion. Put an­other way, given Trump is re­elected, he’s still an un­der­dog to have won the pop­u­lar vote. The elec­toral col­lege seems to fa­vor Trump, but that’s a huge prob­a­bil­ity to put in such a nar­row space, even if you as­sume the states all look iden­ti­cal to 2016. I be­lieve that pre-elec­tion, 538 had Trump at 10% to do this, with the polls only a few per­cent away from that re­sult. How do you get to 20%?

You can sell “Hillary Clin­ton runs for pres­i­dent in 2020” at 12% odds. Is that a worth­while re­turn on cap­i­tal? You could also sell Michele Obama at 8%, Cuomo at 6%, and Oprah or Mark Cuban at 5%.

They have Trump at 88% to be Pres­i­dent at the end of 2019 and 73% to com­plete his first term. They think he is 41% to be im­peached this year and 63% to be im­peached at all. Congress is ex­pected to work fast. Have they met congress?

There are a num­ber of other similar good bets available. One gets the idea. The catch is that those all tie up cap­i­tal. Also, if you take risk and win, you have to pay 10% of your net win­nings and po­ten­tially taxes. Again, three cheers for ar­bi­trage.

Look­ing at such sys­tems of prices, and look­ing for op­por­tu­nity, is of­ten good train­ing as not only a trader or gam­bler but also for cal­ibra­tion and prob­a­bil­ity es­ti­ma­tion in gen­eral, which are ex­cel­lent skills for any­one to de­velop.

What Does This Say About Pre­dic­tion Mar­kets in Gen­eral?

Not much we didn’t already know. Pre­dic­tIt has an $850 limit on any one mar­ket, for any one can­di­date or other po­ten­tial out­come. This does not in­crease if you do ar­bi­trage. This is why pure ar­bi­trage that frees up cap­i­tal can con­tinue. I am liter­ally at risk for $44 in the gen­eral elec­tion mar­ket, but that does not al­low me to con­tinue to trade.

Other mar­kets in the past such as InTrade have not had this re­stric­tion. This re­sults in less egre­gious ver­sions of the same prob­lems, as you can use big­ger size to trade against the mis­takes. How­ever, there is no point in fully cor­rect­ing a mis­take, as do­ing so would offer min­i­mal or no prof­its. If you have a mar­ket that is in­effi­cient, and a chance to trade to make it more effi­cient, that’s a good trade, but at some point it isn’t worth the time and trou­ble and cap­i­tal in­vest­ment, so you stop. That point is nec­es­sar­ily be­fore full effi­ciency, but in places like the stock mar­ket you can po­ten­tially get (in ex­pec­ta­tion) very close.

In pre­dic­tion mar­kets, cost of cap­i­tal to do trades is a ma­jor dis­tort­ing fac­tor, as are fees and taxes and other phys­i­cal costs, and par­ti­ci­pants are much less cer­tain of cor­rect prices and much more wor­ried about im­pact and how many oth­ers are in the same trade. Most ev­ery­one who is look­ing to cor­rect in­effi­cien­cies will only fade very large and very ob­vi­ous in­effi­cien­cies, given all the costs.

Thus, we see the same in­effi­cien­cies pop up over and over again and not be cor­rected. The most well-known and uni­ver­sal one is that if the prob­a­bil­ity is un­der about 40%, the odds will likely be too high. The lower the odds be­low that, the more (as a per­centage of the chance listed) the price will be too high. For low per­centages, the peo­ple sel­l­ing the con­tract are treat­ing it as if it is a bond that pays in­ter­est over time, with a tiny de­fault risk, rather than say­ing that the 7% chance is too high and should have been 5%. One also has to be wary of model er­ror.

In poli­tics, it is also in­evitable that any­thing that sounds su­perfi­cially good to peo­ple on the in­ter­net but is un­likely to ac­tu­ally hap­pen is al­most always go­ing to trade rich.

If any­thing, it is re­mark­able how lit­tle differ­ence it made to limit ac­counts to $850 in trad­ing, be­yond there be­ing free cash ly­ing around.

Any­way, thought that would be fun to write up for­mally given I had been tricked into ac­tu­ally trad­ing the mar­kets, and maybe some of you would get to do some good trades, so I figured why not. Have fun, ev­ery­one.