# Bet On Biden

At this point, for many tax situ­a­tions, you can­not ra­tio­nally be­lieve both ‘Nate Silver is cred­ible’ and ‘I should not be bet­ting on Bi­den’. If you have any trust in Nate you can­not think Bi­den is be­low ~80% to win. You can get ‘Bi­den’ for 65-66c on Pre­dic­tit. Notably, there are sev­eral de facto copies of the same bet (Fe­male VP, Dems win pres­i­dency). Even with fees, you are mak­ing a lot of stochas­tic money. Your cap­i­tal will prob­a­bly not be held up that long.

If you are out­side the USA you can plau­si­bly get even bet­ter odds. There are many places you can get im­plied odds of 62-63%. This bet gets worse if you are pay­ing tons of in­come tax and can­not offset losses. But this is a very good bet for lots of peo­ple.

I ed­ited in an image of the rol­ling pol­ling av­er­ages so peo­ple can judge this for them­selves.

Here is an­other good anal­y­sis of why we should be con­fi­dent in Bi­den.

In gen­eral, I think Ra­tion­al­ists should be a lot ‘louder’ about their ad­vice when they are sure. For many tax situ­a­tions, this is a com­plete slam dunk. I re­ally wish ra­tio­nal­ists had been louder about similar ad­vice in the past when I could benefit.

Con­test En­trants Think Trump is Go­ing to Lose

Of the 98 en­tries, only six said that the Demo­cratic can­di­date was un­der 50c to win the pres­i­dency. In fact, the me­dian pre­dicted prob­a­bil­ity for the Demo­cratic can­di­date (here­after Bi­den, for brevity) among con­test en­trants was 87%; no­tably, the me­dian prob­a­bil­ity given by the mod­els out there is also 87% (though Nate Silver has yet to re­lease his and based on his fights with G. Elliott Mor­ris it’s prob­a­bly not go­ing to come in very high).

This con­fi­dence in Bi­den (he is “Likely” to win the pres­i­dency, in the par­lance of ver­bal hand­i­cap­pers) isn’t shared by the mar­kets. Both BetFair and Pre­dic­tIt price Joe in the low 60s, sug­gest­ing that the pres­i­dency “Leans” Bi­den (the price ranges on PI de­pend­ing on the mar­ket you’re look­ing at). Even Silver thinks that’s too low: “I don’t think peo­ple re­al­ize how dumb and some­times even ir­ra­tional the prices are at poli­ti­cal bet­ting mar­kets as com­pared to al­most ev­ery other type of mar­ket (which is not to say other mar­kets are always ra­tio­nal, ei­ther).” and “Too low on Bi­den.”

Note: My claims are com­pletely or­thog­o­nal to whether Bi­den should win. Ra­tion­al­ists, in gen­eral, have some faith in Nate Silver and similar anal­y­sis. This post is about the log­i­cal con­clu­sion of that faith. Please do not dis­cuss the mer­its of which can­di­date is bet­ter for the coun­try.

edit: At this point if any­one is read­ing and can ac­cess in­ter­na­tional mar­kets the ‘safe’ line is to buy both ‘pop vote’ and ‘Bi­den’. Pop vote is safe so you can ‘hedge’ Bi­den bets with a safe +EV pop vote bet. This way you win money un­less Dems lose the pop vote (ex­tremely un­likely).

• I have bought $400 worth of Trump No con­tracts on Pre­dic­tIt which will pay off if Trump loses. The price as of this writ­ing is 61 cents for a con­tract that pays$1 if Trump loses.

• Notably, the 538 pre­dic­tion doesn’t in­clude a num­ber of out­side fac­tors, pri­mar­ily around mail-in bal­lots and voter sup­pres­sion. 538 has already talked about the prob­lems with mail-in bal­lots be­ing re­jected, and there are also con­cerns about not hav­ing all of the bal­lots counted be­fore the cut-off point where they have to finish count­ing. Repub­li­cans have also made it harder for Demo­crat-lean­ing bases to vote. Th­ese are fac­tors that will hurt Bi­den more than Trump. All those links are to 538, and there are other ar­ti­cles on the site about those same is­sues. If you be­lieve in 538′s model, you should prob­a­bly also be­lieve in their ar­ti­cles that in­di­cate that these out­side fac­tors will be im­por­tant. If you don’t be­lieve in the ar­ti­cles, then why do you be­lieve in their model?

Either way, this is not a clear case where the mar­ket is wrong.

• Only the first ar­ti­cle in the com­ment is by Silver, on whose ex­per­tise the origi­nal poster is bas­ing his recom­men­da­tion. That ar­ti­cle doesn’t dis­cuss mail-in bal­lots or voter sup­pres­sion, and in fact his main point is that the time re­main­ing un­til elec­tion day (al­most three months when the ar­ti­cle was writ­ten) com­bined with un­cer­tain­ties due to Covid-19 meant that the race was still open back then. Those con­sid­er­a­tions have much more limited force at pre­sent, when only 16 days re­main, and Bi­den’s lead has widened con­sid­er­ably.

If you’ve been at all listen­ing to Silver re­cently, you’ll know that he thinks his model prob­a­bly un­der­es­ti­mates Bi­den’s chances. This shouldn’t be sur­pris­ing, since as Silver ac­knowl­edges, in this new ver­sion of the model he has made a spe­cial effort to build con­ser­va­tive as­sump­tions into it.

In any case, I would en­courage peo­ple hes­i­tant to bet for Bi­den to re­sist the temp­ta­tion of “throw­ing in a bunch of con­sid­er­a­tions” for why the mod­els may be wrong, and in­stead try to calcu­late what the cor­rect fore­cast should be in light of those con­sid­er­a­tions. For ex­am­ple, if you think mail-in bal­lots will be a big fac­tor, try to es­ti­mate the mag­ni­tude of this effect.

Fol­low­ing my own ad­vice, I just built a sim­ple Guessti­mate model of the im­pact of mail vot­ing on the pop­u­lar vote. I cre­ated the model very quickly, so if any­one spots any er­rors, please men­tion them be­low. And if you think some of the pa­ram­e­ters should be differ­ent, sim­ply copy the model and ad­just those pa­ram­e­ters to your satis­fac­tion. Note that the effect of “re­jected” in-per­son bal­lots is not mod­eled. This effect fa­vors Bi­den, since a greater pro­por­tion of Trump votes will be in per­son, and hence sus­cep­ti­ble to be­ing “re­jected” (i.e., not cast due to failure to bring an ID, long lines, in­abil­ity to find a pol­ling sta­tion, etc).

ETA: The up­shot of the model is that mail vot­ing shrinks the ex­pect pop­u­lar vote gap be­tween Bi­den and Trump by about 2%. If we as­sume that the elec­toral col­lege gives Trump a ~2% pop­u­lar vote ad­van­tage, the model im­plies a drop in Bi­den’s chances of win­ning the elec­tion from 87% to about 79%. [I mod­ified the model and im­proved some of the es­ti­mates, and now the effect is less than 1%.]

(Dis­clo­sure: I have bet a to­tal of USD 12k on Bi­den, mostly back when his odds where roughly equal with Trump’s.)

• This model seems rea­son­able, but I think bet­tors should mostly ig­nore the pos­si­bil­ity of re­jected mail in bal­lots, be­cause the effect is ex­tremely un­cer­tain and around the same mag­ni­tude as many other idiosyn­cratic fac­tors that should mostly wash out. For ex­am­ple, if there’s se­vere weather or an out­break of COVID on elec­tion day in a cru­cial swing state, that will hurt Trump much more than Bi­den be­cause a much greater pro­por­tion of his vot­ers are vot­ing by mail (es­sen­tially the “re­jected in per­son bal­lots” effect you men­tion).

• The 538 dis­tri­bu­tion cur­rently has Bi­den fal­ling be­tween… <squints> ~255 − 440 elec­toral votes 80% of the time (47% − 82%). Up­dat­ing your guessti­mate sheet with those ranges give a mean pro­por­tion to Trump of .38 with a range of .17 − 0.54

• This is to­tal mis­in­for­ma­tion; mail-in re­jec­tion will prob­a­bly be can­cel­led out by in­creased turnout from low-propen­sity vot­ers (it’s much eas­ier to vote by mail than to vote in per­son). In any case, the effect of a 1-3% re­jec­tion rate is min­i­mal and in­dis­t­in­guish­able from gen­eral noise/​un­cer­tainty.

Most of the claimed “GOP voter sup­pre­sion” ei­ther has a min­i­mal effect or hurts the GOP, be­cause the nuts and bolts of elec­tion ad­minis­tra­tion are im­ple­mented on a county level. In most states, ur­ban coun­ties have the re­sources and in­cli­na­tion to counter the voter sup­pres­sion, while ru­ral (pre­dom­i­nantly GOP) coun­ties do not.

• I was merely not­ing that 538, the mak­ers of the pre­dic­tion model that the post is dis­cussing, be­lieves in the voter sup­pres­sion. If you think they’re wrong about the voter sup­pres­sion, then you prob­a­bly also shouldn’t be­lieve in their pre­dic­tion model. On the other hand, if you think they’re right about the pre­dic­tion model, then why are you doubt­ing their voter sup­pres­sion re­search?

Of course, it’s perfectly con­sis­tent to think that they are wrong about both the model and the voter sup­pres­sion, but the post was as­sum­ing that you be­lieved in the pre­dic­tion model.

• I’m de­bat­ing to what ex­tent to write a de­tailed ex­pla­na­tion, but

(1) yes the odds are in­sane at this point, and I felt the need to note this in my weekly Covid post. Trump is not 0% but he sure isn’t 38%.

(2) the rel­a­tive odds be­tween the odds now, and the odds ear­lier in the year, are how you know the mar­ket is in­sane here. Nate’s up­date from 71% to 87% re­flects that Trump’s po­si­tion is ob­vi­ously vastly worse than it was be­fore, with less time to go, fewer op­por­tu­ni­ties for things to change, many votes in the bank and Bi­den sev­eral points farther ahead when the baseline was tight­en­ing. Yet the odds on Trump re­main stub­born.

When I wrote my re­cent re­view of Pre­dic­tIt prices, I thought Trump’s odds were very gen­er­ous but per­haps plau­si­ble. That’s no longer the case.

• Hey OP -

Just want to say thank you for this post. This along with the failure of the ra­tio­nal­ist com­mu­nity to buy bit­coin led to me bet­ting $150 AUD on Bi­den, and Sports­bet Aus­tralia called the elec­tion for him already (no, I don’t un­der­stand why ei­ther—I guess the pub­lic­ity), so I got$235 in my bank ac­count now (win: $85 AUD). I would not have done this with­out your post. https://​​www.9news.com.au/​​na­tional/​​us-elec­tion-2020-sports­bet-pays-out-early-on-joe-biden-win/​​a722588b-85ec-4a87-a524-a1e592123691 • If you have any trust in Nate you can­not think Bi­den is be­low ~80% to win. That seems too strong. I’m as­sign­ing a 50% chance to Nate be­ing the best au­thor­ity on this, and a 50% chance to mar­kets be­ing the best au­thor­ity. I still agree that bet­ting on Bi­den has pos­i­tive ex­pected value. My main rea­son for doubt­ing Nate is the like­li­hood that the pan­demic will have strange effects on turnout. I’m un­will­ing to bet on which di­rec­tion that will sur­prise peo­ple. • For what it’s worth, 538′s fi­nal pre­dic­tions have (nar­rowly) out­performed bet­ting mar­kets in 2008, 2012, 2016, and 2018 (I haven’t looked at 2010 or 2014). Also, al­most ev­ery con­sis­tently prof­itable poli­tics bet­tor is putting mas­sive bets on Bi­den this cy­cle, and my im­pres­sion is that they ei­ther be­lieve 538 is ~ac­cu­rate or bi­ased in fa­vor of Trump. As an ex­am­ple, here’s the twit­ter ac­count of one of the best pro­fes­sional poli­tics bet­tors around; he be­lieves Bi­den has a ~96% chance of win­ning. • There’s a lot of par­ti­san sneer­ing against Trump sup­port­ers on that Twit­ter ac­count which IMO calls his ob­jec­tivity on the odds is­sue into ques­tion. • He’s been a pro poli­ti­cal bet­tor for years, so if he was bet­ting based on bi­ased odds he’d have gone broke a long time ago. It’s just a fact that Trump is ex­tremely over­priced, and his share price is be­ing propped up by ex­tremely -EV bet­tors; him us­ing that as an ex­cuse to at­tack Trump sup­port­ers doesn’t un­der­mine the qual­ity of his anal­y­sis. • In gen­eral I’m more in­clined the trust the mar­ket value than any par­tic­u­lar prog­nos­ti­ca­tor. Why shouldn’t I be? edit: The ar­gu­ments here have con­vinced me to bet some money on Bi­den (on Bet­fair), but to­day the odds are still mov­ing against him (down to just un­der 60% on elec­tion­bet­tin­godds.com). This does leave me rather con­fused—surely the “dumb money” can’t out­weigh the “smart money” so heav­ily? Espe­cially not on Bet­fair which doesn’t have Pre­dic­tIt’s limi­ta­tions. • Yeah, I think the OP’s ar­gu­ment re­quires a good ex­pla­na­tion of why the mar­kets have failed to ad­e­quately price in Silver’s pre­dic­tion. I’m not say­ing mar­kets are always right and im­pos­si­ble to beat no mat­ter what, but the EMH is a pretty good de­fault start­ing point, re­quiring strong ev­i­dence to over­come in any spe­cific case. ‘I trust this guy’s pub­lic pre­dic­tion, based on equally pub­lic in­for­ma­tion about his meth­ods and track record’ doesn’t seem nearly enough. • The effi­cient mar­ket hy­poth­e­sis is an over­rated and du­bi­ous hy­poth­e­sis when ap­plied to the stock mar­ket (a mar­ket which could plau­si­bly have the nec­es­sary con­di­tions for EMH to be rea­son­able), bet­ting mar­ket’s like pre­dic­tit though are noth­ing like the stock mar­ket and sig­nifi­cantly more in­effi­cient. The nec­es­sary con­di­tions for a mar­ket to be effi­cient re­quire the mar­ket to be heav­ily/​ma­jor­ity used by profit max­i­miz­ing in­vestors and this just isn’t the case for most gam­bling mar­kets. In gen­eral bet­ting on a gam­bling mar­ket is a much worse idea than bet­ting on stocks, since the trans­ac­tion costs are a lot higher and you are bet­ting on a zero-sum game in­stead of a his­tor­i­cally pos­i­tive-sum one. That be­ing said there are some no­table ex­am­ples of the mar­ket com­pletely failing. Take the May­weather vs McGre­gor fight where the best boxer in the world for a decade only had a −400 edge against some­one who had never had a pro­fes­sional box­ing match (the line started at a much more rea­son­able −2500). If I had more money back in 2017 I would have bet that line heav­ily. Just be­cause you are bet­ting on a usu­ally sub­par mar­ket doesn’t mean there aren’t still good deals to be had on oc­ca­sion. • My ar­gu­ment doesn’t re­quire the EMH to be true in any strong sense, just true enough that you can’t beat a big, liquid mar­ket by a large mar­gin via easy anal­y­sis of very pub­lic in­for­ma­tion. Pre­dic­tit’s price is similar to Bet­fair’s—in fact Bet­fair is cur­rently slightly less favourable to Bi­den—and pre­sum­ably won’t di­verge too far in ei­ther di­rec­tion due to ar­bi­trage. Bet­fair has matched about$180 mil­lion on its ‘next US Pres­i­dent’ mar­ket. Of course a lot of its users are wild gam­blers, but it is big enough to at­tract se­ri­ous pro­fes­sion­als (and smart op­por­tunists with plenty of cap­i­tal) too.

• I’ve been told that the elec­tion is con­sid­ered a mas­sive and rare op­por­tu­nity among pro­fes­sional gam­blers, which that per­son thought was the ‘smart money’.

• Any­one who thinks that bet­ting mar­ket prices are ra­tio­nal will not bet on them. Why would you tie up your cap­i­tal (even for a few weeks) and pay fees if the prices are fair?

• Can you elab­o­rate? I’m not sure what the point of dis­agree­ment is here.

For a big mar­ket like this, on an ex­change like Bet­fair, my rough men­tal model is that there is a com­bi­na­tion of recre­ational/​com­pul­sive gam­blers, more se­ri­ous pro­fes­sional/​wannabe pro­fes­sional gam­blers, and (given that the US elec­tion cap­tures the at­ten­tion of a great por­tion of the world) peo­ple who don’t bet for a liv­ing but will keep an eye on the mar­ket and take any ob­vi­ously prof­itable bets on offer.

So a lot of the bets that get matched will be be­tween un­se­ri­ous play­ers (who ei­ther don’t care or don’t think care­fully about whether the price is ra­tio­nal) and se­ri­ous play­ers (who do). Others will be matched within the group of se­ri­ous play­ers, who of course can dis­agree with each other about the ‘cor­rect price’. But if there’s a mis­pric­ing ob­vi­ous enough that you or I can de­tect it just by read­ing 538 or Less Wrong, and large enough to be hugely +EV af­ter trans­ac­tion and op­por­tu­nity costs, my de­fault as­sump­tion is that it will prob­a­bly self-cor­rect be­fore I get a chance to take ad­van­tage. Not all the money in the mar­ket is ra­tio­nal, but there are ra­tio­nal bet­tors with enough cap­i­tal to keep the prices roughly where they should be.

edit to clar­ify why I’m fo­cus­ing on Bet­fair: I don’t know a lot about Pre­dic­tit, but I be­lieve it heav­ily limits the amount any in­di­vi­d­ual user can bet. In iso­la­tion, that could be enough to al­low it to stay sur­pris­ingly ir­ra­tional for sur­pris­ingly long. But I as­sume its prices for any sig­nifi­cant mar­ket will stay rel­a­tively close to those of a big ex­change like Bet­fair, be­cause of the easy ar­bi­trage op­por­tu­ni­ties when they di­verge. And the 65-66c you quote is very close to the Bet­fair price of about a day ago (though the lat­ter has since tight­ened to the equiv­a­lent of about 62c).

• At least a cou­ple of peo­ple have down­voted this with­out re­ply­ing. One of the things I usu­ally like about LW and similar spaces is that good-faith com­ments get a re­ply, not just a silent dis­mis­sal.

I get that the com­ment is not ex­actly punchy or brilli­antly writ­ten, but I was try­ing to do the right thing and ex­plain some of my own as­sump­tions, rather than write the equiv­a­lent of ‘wat?’ and leave deluks917 to do all the work of bridg­ing our com­mu­ni­ca­tion gap.

• I have a much higher con­fi­dence in this be­ing a good bet than in any par­tic­u­lar the­ory of why peo­ple are bet­ting on Trump.

• The ar­gu­ments here have con­vinced me to bet some money on Biden

Which ones con­vinced you? I still haven’t seen a se­ri­ous ex­pla­na­tion for why ev­ery­one is ap­par­ently leav­ing so much ex­pected value on the table. It seems to me that the mar­ket is big enough to be worth the at­ten­tion of se­ri­ous peo­ple, but not so big that it would take a huge num­ber of them to snap up all the ‘dumb money’ and push the odds some­where rea­son­able.

• The main fac­tors were Nate Silver’s record of good cal­ibra­tion, and (alleged) “smart money” suc­cess­ful bet­tors (in­clud­ing lo­cal Zvi) be­ing onto it. I’m still un­cer­tain why the mar­ket could be so in­effi­cient in this case, so I’m not bet­ting a lot of money, but it’s pos­si­ble that in­sti­tu­tional fac­tors pre­vent large amounts of smart cap­i­tal com­ing in—per­haps di­rectly bet­ting on the elec­tion is still too weird for a large in­vest­ment firm.

• I just made my ac­count but I want to re­mind ev­ery­one that you can­not make in­fer­ences on how good your pre­dic­tion (or how good of a bet this was) based on one data point (how this elec­tion turned out). If you want to dig deep into the odds that ev­ery state was given, you can start to make a case, but any­one with the gut re­ac­tion that since the elec­tion was close, this was a bad bet, are wrong.

• Agreed. The best pre­dic­tion would have been to as­sign Bi­den 100% cre­dence. A defence of their po­si­tion might say that “Given the in­for­ma­tion 538 had at the time, the best pos­si­ble pre­dic­tion ma­chine would have less cer­tainty than Nate”, but this is too hard to prove, and it’s in­con­sis­tent with the fact that Nate is a cal­ibrated fore­caster who (as far as I know) con­sis­tently beats the mar­ket.

• For whom is bet­ting for Bi­den ac­tu­ally a good idea? Can non-US cit­i­zens effec­tively par­ti­ci­pate? What when you fac­tor in trans­ac­tion costs? What are my ex­pected gains (are they high enough to reg­ister an ac­count and go through the has­sle)?

• Do you have an es­ti­mate of ex­pected profit per $100 bet (for a few of the most plau­si­ble sce­nar­ios)? My im­pres­sion is that Pre­dic­tIt is +EV is you make lots of not-too-cor­re­lated bets so that your losses can offset your wins (though maybe not by enough to be worth the time & effort), but it’s gen­er­ally -EV (or at best barely +EV) if you de­posit to make a one-off bet where you have to pay fees & taxes on your win­nings (and don’t get any tax benefit from your losses). • I would say Bi­den is con­ser­va­tively 80% to win from here. Polls are in­cred­ibly good. The elec­tion is in six­teen days. When I made the thread you could get Bi­den for 0.65. You need to pay 3.5 cents tax on win­nings if you buy at 0.65. There is also a 5% with­drawal tax. So if you win you get (100/​0.65)(0.965)(0.95)(0.8) = 112.8 . 12.8% re­turns in 16 days is noth­ing to laugh at. You are still get­ting al­most 6% ex­pected re­turns if you think Bi­den is only 75% to win. This all as­sumes you im­me­di­ately with­draw your win­nings from Pre­dic­tit which is ob­vi­ously bad for your ex­pected value. If you are not in the US you can both get bet­ter odds than 0.65 and you don’t have to pay the in­sane rake. • It looks like this in­cludes the fees you pay to Pre­dic­tIt, but not the taxes you pay to the gov­ern­ment. • Yes. I was clear in the OP Pre­dic­tit is a worse deal de­pend­ing on your tax situ­a­tion. If you can­not deduct losses things look much worse. But many peo­ple ei­ther have low in­comes or already deduct. Bet­ting on in­ter­na­tional sites should be prof­itable re­gard­less of tax sta­tus since the re­turns are much bet­ter. • Is there any­thing a per­son liv­ing in Ger­many can do other than place a bet via some­one else? • Bet 100% of your up­com­ing char­i­ta­ble dona­tions. Even if you are loss averse the char­i­ties are not. Donate the win­nings if the bet pays off. • You could try us­ing Augur us­ing a VPN I think. Though you need Cryp­tocur­rency (Etherum). Which is the rea­son I didn’t use it so far (too much effort to set up). • Ok so I am through the pro­cess of mak­ing an acount and adding funds to it. You have to trans­fer your money into the Cryp­tocur­rency DAI. The bet with the low­est fees and the longest ex­pira­tion date (to al­low for a longer time un­til the re­sults are known), has Trump at 62%. The cheap­est NO-share that you can buy im­me­di­ately is at .69. There is a 0.77% mar­ket fee. But there is a rel­a­tively high fee for mak­ing a trans­ac­tion which is payed in Etherum. it cur­rently costs the equiv­a­lent of around 10$. But this trans­ac­tion doesn’t scale with your bet, so if you already plan to make a big bet (<500$) this shouldn’t af­fect your de­ci­sion much. • I went the crypto route a bit, but then de­cided to fun­nel it through an ac­quain­tance and use pre­dic­tit in­stead. If I un­der­stand ev­ery­thing cor­rectly and calcu­lated all the fees prop­erly (prob­a­bly haven’t), then the over­all deal from the point where money leaves my pay­pal to the point where hope­fully more comes back, is analo­gous to bet­ting on Bi­den as if he had a 73,6% on the mar­kets. Which I guess is still pretty good, but man do I wish that there was just one large mar­ket with min­i­mal fees, in­stead of fees on trans­fer­ring, cur­rency con­ver­sion, profit on pre­dic­tit, and with­drawal from pre­dic­tit. So many fees :x Also, it takes a few more days. I hope the mar­kets don’t shift in Bi­den’s fa­vor in the mean­time (then the deal ob­vi­ously gets worse). It’s amus­ing how I now want them to move into the op­po­site di­rec­tion. And all that is as­sum­ing a 0% risk that the per­son just screws me over and keeps the money for them­selves, which is prob­a­bly roughly ac­cu­rate. • man do I wish that there was just one large mar­ket with min­i­mal fees Such a mar­ket ex­ists, though un­for­tu­nately it is re­stricted to coun­tries where most LW users are not cit­i­zens of. • Ok you don’t need Etherum. You can just buy Augur’s cur­rency us­ing a credit card. I tried us­ing coin­base, but it seems to have pretty high ex­change fees. • Bet­fair.de should work. Be­sides sports, there is a poli­tics sec­tion. • Sadly, while you can place sport bets from Ger­many via Bet­fair, you can’t ac­cess poli­ti­cal bets. • I con­cur with this. In my case, I set aside some amount of money I was com­fortable los­ing, calcu­lated the Kelly bet based on the ex­pected win, and bet some amount on Bi­den. I used Bet­fair, which is available to Euro­peans. Some com­menters men­tion the EMH. As coun­terev­i­dence, I pre­sent that Bet­fair is offer­ing even odds that Bi­den will win at least one state Trump won last time. (The 538 model gives 96% to this) (This was wrong) • From the blog of An­drew Gel­man, one of the au­thors of the Economist’s model: As noted above, I think that with wider cen­tral in­ter­vals and wider tails we could lower that Bi­den win prob­a­bil­ity from 96% to 90% or maybe 80%. But, given what the polls say now, to get it much lower than that you’d need a di­rec­tional shift, some­thing asym­met­ri­cal, whether it comes from the pos­si­bil­ity of vote sup­pres­sion, or turnout, or prob­lems with sur­vey re­sponses, or differ­en­tial non­re­sponse not cap­tured by par­ti­san­ship ad­just­ment, or some­thing else I’m for­get­ting right now. But I don’t think it would be enough just to say that any­thing can hap­pen. “Any­thing can hap­pen” start­ing with Bi­den at 54% will lead to a high Bi­den win prob­a­bil­ity now mat­ter how you slice it. For ex­am­ple, sup­pose you start with a Bi­den fore­cast at 54% and give a stan­dard er­ror of 3 per­centage points, which has gotta be too much—it yields a 95% in­ter­val of [0.48, 0.60] for his two-party vote share, and no­body thinks he’s gonna get 48% or 60%. Any­way, start with that and Bi­den still has a 78% chance of win­ning (or 75% us­ing the t_3 dis­tri­bu­tion). To get that prob­a­bil­ity down be­low 80%, you’re gonna need to shift the mean es­ti­mate, which im­plies some di­rec­tional in­for­ma­tion. • If you have BTC ly­ing around and trust my­bookie you can get 1K free if you wa­ger enough. Get­ting 1K free on top of 10K is a lot of value. The site is a lot worse with­out BTC. • again, watch out for rol­lover re­quire­ments be­fore tak­ing signup bonuses! Un­less you’re a pro­lific sports bet­tor it’s al­most cer­tainly a bad idea to take the bonus. • Is there a sim­ple guide on how to bet on Bi­den if I already have Ethereum and I don’t live in the US? It seems I can do it on Augur and FTX, but both plat­forms seem very com­pli­cated. • Prac­ti­cal ques­tion: Say I were will­ing to break the law. Is there an easy way to bet on BetFair from the US? • I do not know of any safe to do this. If you are found out you risk los­ing money if they freeze your ac­count. This bet looks a lot worse if you are risk­ing your ac­count when you get found out. If you have out­side friends who trust you maybe they will ‘bet for you’. There are some offshore sites tak­ing Amer­i­can bets. You can also place bets on­line legally if you are in NJ (I think). • Any­one with available cap­i­tal in­ter­ested in mak­ing money from this elec­tion should spend ~1hr re­search­ing TX, come to the ac­cu­rate con­clu­sion that Bi­den is fa­vored there, and bet ac­cord­ingly (buy Bi­den in TX, Dems to win TX sen­ate race, Bi­den elec­toral col­lege mar­gin >210 EC votes, etc). The ex­pected re­turns are much higher than bet­ting on Bi­den to win over­all, be­cause Bi­den TX is cur­rently trad­ing at around 30% to win TX, when fair value is ~65%. Th­ese twit­ter feeds from (semi)pro poli­tics bet­tors are a good place to start: 1, 2. The key ar­gu­ment is that pol­ling sys­tem­at­i­cally un­der­es­ti­mates Democrats with Lat­ino vot­ers in Texas, and that the elec­torate will be much more fa­vor­able to Democrats than in 2018, when Dems lost the Se­nate race by ~2.5 points in a worse na­tional en­vi­ron­ment. • I spent some time. I do not see the smok­ing gun. Can you post more de­tails or links? • Pol­ling shows Bi­den and Trump very close in the state • pol­ling in Texas sys­tem­at­i­cally un­der­es­ti­mates Dems with Latinos • Turnout in TX seems to be much higher than past years, which is great for Dems be­cause peo­ple who didn’t vote in 2016/​2018 are dis­pro­por­tionately young/​nonwhite • Dems nearly won in 2018 de­spite de­pressed Lat­ino turnout • Do you think FiveThir­tyEight and the Economist haven’t ap­pro­pri­ately ac­counted for these con­sid­er­a­tions in their mod­els? I don’t think the dis­crep­ancy with the mar­kets are so large. Where did ~65% come from? • Nei­ther model takes into ac­count any of these fac­tors, and in­deed their dis­crep­ancy with the mar­kets are much smaller. 65% just comes from my es­ti­mate of the fair value of a TX biden con­tract, which comes from...what­ever in­ter­nal pro­cess gen­er­ates prob­a­bil­is­tic cre­dences. • Yeah the +EV of this bet is marginal at best. • Are there any im­por­tant con­sid­er­a­tions in the op­po­site di­rec­tion? • There’s plau­si­ble ar­gu­ments for Trump be­ing fa­vored (he is cur­rently ahead in the polls, af­ter all), but I can’t think of any good ar­gu­ments for him be­ing at ~75c. EDIT: Trump and Bi­den are now tied in the 538 TX pol­ling average • The Texas bet (TX) seems EV neu­tral to me, and clearly far worse than the na­tion­wide elec­toral col­lege (EC) bet. Bi­den for EC: 95% likely (The Economist model), priced at 62% Bi­den for TX: 26% likely (The Economist), priced at 29% The two Twit­ter feeds are full of a lot of shit­post­ing, and don’t up­date me much. • I mean yes, clearly if you be­lieve the Economist and 538 mod­els are rea­son­able, then the TX bet isn’t +EV. My point is that the mod­els are clearly un­rea­son­able for the rea­sons listed above, and the bet is ac­tu­ally ex­tremely +EV. • There are three ar­gu­ments (1) polls un­der­es­ti­mat­ing Dems in South­ern states, and (2) bench­mark­ing against 2018 sen­ate, and (3) some low-qual­ity Tweets. It’s weird to hold a lot of stock in (2), given noise from can­di­date se­lec­tion and other vari­ables. If you place a lot of weight on (1), the ac­tu­ally sane bet would be Bi­den in AZ. It’s rated 2nd and 4th most likely to go dem by Cohn and Wasser­man re­spec­tively. Bi­den for AZ: 77% likely (Economist), priced at 54% on Elec­tion Bet­ting Odds. • I’m not a gam­bler so for­give me if these ques­tions are stupid: How are these odds be­ing calcu­lated in the first place? I think we all re­mem­ber the odds of 2016 ver­sus the out­come. What were the var­i­ous pun­dits be­ing touted this time say­ing back then? Why would book­ies be offer­ing odds that they know they’re likely to lose out with? What are their profit mar­gins? I know that cas­inos are a busi­ness that de­pends on ex­tremely high vol­umes of bets, are book­ies the same? Is it pos­si­ble to bet on the re­sults of vot­ing by state? That seems safer to me (well, as safe as any pre­dic­tion of the fu­ture can be). Can I bet on how long it takes be­fore Bi­den is re­tired by Har­ris/​the Dems? I know that it is pos­si­ble to hedge your bets and place bets on both out­comes (that take ad­van­tage of differ­ent odds). Is any­one sug­gest­ing a strat­egy like that? I know the elec­tion is flashy and top­i­cal but if we are talk­ing gam­bling for profit then are there bet­ter bets in gen­eral for other things? It seems silly to bet on some­thing just be­cause it’s there. The elec­tion pre­sum­ably will have an effect on the stock mar­ket. Given that you can already ‘bet’ on stocks, di­rectly, in fu­tures, and in hedg­ing, then is gam­bling bet­ter or worse than that? • I would en­courage you to bet on other things too if you think you have a big edge. It is not ob­vi­ous to me there are any bet­ter bets available at this mo­ment. 538 had rea­son­able odds in 2016 imo. 27% chances hap­pen. Book­ies/​Sites offer odds based on how much money is on both sides. • I got heav­ily down­voted for sug­gest­ing that Nate Silver’s cred­i­bil­ity was not strong enough to make this a good bet. The most you could say at this point (late in the night on elec­tion day) is that it looks like the elec­tion is very close. This sug­gests to me that those peo­ple say­ing that bet­ting on it at 65% odds for Bi­den was a huge steal, were over­con­fi­dent. Peo­ple were quoted who thought Bi­den’s P(win) was 96%. I am in­ter­ested in any sug­ges­tions about what went wrong. I can’t think of a lot of ed­ify­ing rea­sons. 1. Ra­tion­al­ists don’t win 2. Dun­ning-Kruger effect 3. Gen­eral over­con­fi­dence 4. Say­ing Trump had a chance runs the risk of sig­nal­ing that you are a “Trump­tard”. 5. Stat­ing high con­fi­dence is one way to sig­nal high in­tel­li­gence. “Uh I’m not sure” doesn’t sound too smart. • Aren’t most of Bi­den votes on mail votes, which are counted later? From what I’m see­ing this still doesn’t tell us that we aren’t in a “Bi­den wins with­out a huge mar­gin”, which was the largest chunk of those 80%/​90% es­ti­mates. I ex­pected wak­ing up at this time and see­ing that the votes were mo­men­tar­ily favour­ing Trump (in fact I had to or­der my brain to ex­pect and pre­pare for that). “Bi­den wins” hy­poth­e­sis has nar­rowed with­out those “Bi­den wins by a large chunk” sce­nar­ios be­ing likely any­more, but in the “Bi­den wins” hy­poth­e­sis I wouldn’t have called see­ing that un­likely. So far there’s not so much “Trump wins” ev­i­dence I’d as­sume some­thing was wrong with the prior odds, I’d think that if the re­main­ing votes would be ex­pected to be evenly dis­tributed, but they are not. My es­ti­mate 2 days ago was Bi­den at 85% to win, be­cause a lot of things had to go in a very spe­cific way for it to not hap­pen or there had to be an er­ror with­out any prece­dent in the polls. I felt like I was sup­posed to go at least 90%, but couldn’t bring my­self to do that be­cause I felt over­con­fi­dent in do­ing so, even if I was pretty sure my es­ti­mate was erring in the op­po­site side. I think that a lot of ra­tio­nal­ists as­sumed that the most com­mon odds they were hear­ing (80% or so) had been too cau­tious be­cause pol­lsters were afraid to make the same wrong call they made last time, so they aug­mented the odds a bit. What con­vinced me to go from 80% to 85% was see­ing that Bi­den lead was three time the lead Hillary had, so that kind of fluke was re­ally un­likely. If ra­tio­nal­ists had been over­con­fi­dent I’d ex­pect this cor­rec­tion from pol­lsters ex­ces­sive un­der con­fi­dence played a large part on any­one who wasn’t just calcu­lat­ing odds on sheer polls data. If Trump some­how wins… I think we’d have ei­ther wit­nessed some­thing pre­vi­ously un­known that made the polls a lot less re­li­able (this seems to be a very par­tic­u­lar elec­tion for lots of fac­tors), which usu­ally doesn’t hap­pens so bet­ting on Bi­den then at those odds was still the in­tel­li­gent thing to do, since you’d make money most times you did that, or that we sim­ply got the un­likely re­sult with things go­ing in a spe­cific way. But I can’t think of any­thing ra­tio­nal­ists should have done differ­ently. • Peo­ple were quoted who thought Bi­den’s P(win) was 96%. Please provide ac­tual quotes. I don’t know any­one who held that be­lief. Not even Nate Silver held that be­lief if you take into ac­count elec­tion shenani­gans. No­body in the thread men­tioned any­thing close to that prob­a­bil­ity. Most ex­pressed prob­a­bil­ities I’ve found by ra­tio­nal­ists are some­thing in the 70-80% range for Bi­den, which at this point still doesn’t strike me as un­rea­son­able. • We are in a sce­nario Nate Silver ex­plic­itly de­scribed: But what’s tricky about this race is that — be­cause of Trump’s Elec­toral Col­lege ad­van­tage, which he largely car­ries over from 2016 — it wouldn’t take that big of a pol­ling er­ror in Trump’s fa­vor to make the elec­tion in­ter­est­ing. Im­por­tantly, in­ter­est­ing isn’t the same thing as a likely Trump win; in­stead, the prob­a­ble re­sult of a 2016-style pol­ling er­ror would be a Bi­den vic­tory but one that took some time to re­solve and which could im­peril Democrats’ chances of tak­ing over the Se­nate. On the flip side, it wouldn’t take much of a pol­ling er­ror in Bi­den’s fa­vor to turn 2020 into a his­toric land­slide against Trump. As of writ­ing this com­ment Bi­den has sev­eral paths to vic­tory and is heav­ily fa­vored to hit at least one. I think he is about 75% to win at this point, de­spite all the bad news. We are in a world where Bi­den might win de­spite los­ing all of PA FL OH NC. Very lit­tle had to go right for Bi­den to win. Nei­ther my­self nor Nate Silver made any state­ments say­ing the elec­tion would definitely be a land­slide. I told peo­ple to bet on Bi­den to win not to bet on Bi­den win­ning in a land­slide. I will say there were some peo­ple push­ing more ag­gres­sive Bets (Rain­bow Jeremy comes to mind) and we should definitely up­date against their point of view. But Nate re­mains cred­ible. • Here’s what would make me ques­tion Nate’s cred­i­bil­ity: If we look at the 538′s former pro­jec­tions and com­pare them to the bet­ting mar­kets prior to elec­tion, make Kelly bets, and lose money.[1] Note that 538 can be on the wrong side of the mar­ket some of the time and still be prof­itable. Say­ing Nate lacks enough cred­i­bil­ity when he is well cal­ibrated and on the right side of the mar­ket doesn’t make any sense and seems like an ex­treme case of hind­sight bias. For ex­am­ple, you use the 2016 elec­tion to un­der­mine 538′s cred­i­bil­ity, but (if I re­call cor­rectly) 538′s fore­cast gave Trump a greater chance than the mar­ket’s odds im­plied, so if you bet based on 538′s fore­cast you would have bet on Trump and made money. I could be re­mem­ber­ing wrong, so if some­one links this data, they’d get an up­vote from me. 1. Look­ing at all or­der­ings of elec­tions ↩︎ • I’m too lazy to look it up, but I did re­search this a cou­ple of weeks ago and found that 538 had in­deed out­performed the mar­kets both in 2008 and 2016 (I wasn’t able to find data for 2012). This is not very in­for­ma­tive, though, since it’s just a hand­ful of cases. Much bet­ter is to look at the state-level pre­dic­tions and use brier scores as a mea­sure of fore­cast­ing perfor­mance. • 538 to­tally out­performed in 2012 on in­trade—it seems like there were whales push­ing up the rom­ney price on in­trade. • And this year too. I agree about the small sam­ple size if we’re just ask­ing “Who will be pres­i­dent?” We should use ev­ery one of 538′s pre­dic­tions that could have been used to make a bet in a liquid bet­ting mar­ket. If a prof­itable set of Kelly/​Markow­itz bets (un­der all elec­tion or­der­ings) has a worse Brier score than the mar­ket’s Brier score, I’d be more in­ter­ested in the profit of the bets, since we’re talk­ing about bet­ting money. • Here are the re­sults we cur­rently have put into 538′s in­ter­ac­tive map. I’ve been strictly kind to Trump, mean­ing I’ve put him as win­ning some races that aren’t yet called, while only giv­ing Bi­den races that have been called. 538 thinks this comes out at ~86/​7/​7 Bi­den/​Trump/​Tie (these are simu­la­tions and the num­bers seem to fluc­tu­ate). And I be­lieve Ne­braska’s sec­ond dis­trict is also won by Bi­den (I’m con­fused about that one, can’t find an offi­cial source), in which case all the ties to go Bi­den. I’m not con­vinced this looks bad for Nate Silver. We seem to be in the pes­simistic half of the dis­tri­bu­tion (from Bi­den’s pov), but not that far. It’s too early to be sure, but there seems to have been a mas­sive pol­ling er­ror in Florida, and some smaller er­rors in other swing states in Trump’s fa­vor. Again, that’s worse for Bi­den than the me­dian out­come, but not crazy worse. • I’m not a fan of in­vest­ment ad­vice on LessWrong. If these mar­kets are sig­nifi­cant (they aren’t—there’s just not that much money chang­ing hands, and the fees and terms make it unattrac­tive to pro bet­tors), there’s some­thing to ex­plain, be­fore ad­vis­ing to ex­ploit. I think you’re well off the mark when you claim that “ra­tio­nal­ists … have faith”. • This is ex­tremely false; I know of mul­ti­ple pro­fes­sional poli­ti­cal bet­tors, and plenty more who make 5+ figures bet­ting on elec­tion years. Also, Pre­dic­tIt has$113m matched on the “Who will be the next pres­i­dent?” mar­ket alone, which is a lot of money by any rea­son­able stan­dard…

The ex­pla­na­tion for the mis­pric­ing is re­ally sim­ple: Trump sup­port­ers tend to think that the polls are rigged and he’ll win, and Bi­den sup­port­ers tend to think the polls are off/​Trump will rig the elec­tion/​voter sup­pres­sion will cost them the elec­tion/​ etc etc so 90% of the square money is on Trump and there’s not enough sharp money to bal­ance it out. Elec­tions are just too in­fre­quent for bi­ased bet­tors to lose money quickly enough that prices are effi­cient, and it doesn’t help that Trump un­ex­pect­edly won the 2016 elec­tion, trans­fer­ring a lot of money to bi­ased bet­tors.

• If it’s that ob­vi­ous, though, why does it per­sist? Elec­tions might not be fre­quent enough to bankrupt the sources of dumb money, but why have the pros not eaten it up by now? Even if there aren’t that many of them, and de­spite their skills they’re not very rich yet, why haven’t the peo­ple who do have enough cap­i­tal to bring the mar­kets back to their senses got in­volved?

• That’s a good ques­tion; I as­sume it’s be­cause they don’t have enough do­main ex­per­tise to make a con­fi­dent judge­ment about the mar­ket, and they have more valuable things to do with their time than ac­quire that ex­per­tise.

• it doesn’t help that Trump un­ex­pect­edly won the 2016 elec­tion, trans­fer­ring a lot of money to bi­ased bet­tors.

I mean they got the an­swer right so it seems a lit­tle ar­ro­gant to call them bi­ased. Maybe they just had a good heuris­tic to do with prefer­ence falsifi­ca­tion?

• Prices have been bi­ased to­wards the GOP ever since 2016, and I’m pretty sure it’s be­cause most of the Trump bet­tors in 2016 were GOP par­ti­sans rather than so­phis­ti­cated an­a­lysts. “Maybe they just had a good heuris­tic to do with prefer­ence falsifi­ca­tion?” what does this mean?

• Thanks for the data!

I do know a num­ber of pro­fes­sional bet­tors (poker and sports/​horse, not poli­ti­cal), who are only lightly bet­ting on the poli­ti­cal sites—the terms and vig are very unattrac­tive to them.

I stand cor­rected if there’s \$113M on that spe­cific bet, though—that’s enough (pre­sum­ing it’s fairly ac­tive, not all placed months ago and just forced to re­solve rather than hedg­ing/​ad­just­ing) to in­di­cate pretty broad be­lief. That does change it from “small-stakes very likely ex­ploitable” to “rea­son­able stakes, but one should an­swer WHY there’s a dis­crep­ancy be­fore com­mit­ting very much”.

• Have you ac­tu­ally tried check­ing with any­one re­sem­bling a ‘pro bet­tor’? I pre­dict the ma­jor­ity of them would say the bet is in fact at­trac­tive and you should put money on Bi­den. I see one per­son in the thread who is known for reg­u­larly mak­ing money bet­ting on things other than stocks (Zvi). He said ‘yes the odds are in­sane at this point’.

Terms are ac­tu­ally bet­ter right now out­side of Pre­dic­tit but Amer­i­cans can­not legally use those. I have not look into all the op­tions for get­ting mil­lions into the mar­kets (I do not have that kind of money). But you can cer­tainly bet hun­dreds of thou­sands of USD eas­ily if you are not Amer­i­can. This seems more than suffi­cient to be at­trac­tive to pro bet­tors. The things you list in paren­the­sis do not ex­plain the cur­rent situ­a­tion.

• ‘Nate Silver is cred­ible’

He had Trump at 28% chance to win, so while cred­ible he is not 100% ac­cu­rate. Let’s say Trump ac­tu­ally had a 50% chance of win­ning last time, which is not out­landish see­ing that he won the elec­toral col­lege by a sig­nifi­cant mar­gin, so NS was off by ~22+%.

Now he says Trump is 87% likely to lose. Take off 22% and it’s 65%. After costs, taxes, op­er­a­tional slip-ups, your time etc, and you have an ex­pected loss at 60-65% odds.

Also, if Trump wins and you are a Bi­den sup­porter, you lose twice. That would be bad he­do­nics. On the other hand if you pre­fer Trump this could be a form of hedg­ing.

Not only that there is time to the elec­tion dur­ing which a lot can hap­pen. “A week is a long time in poli­tics”, So even if you are right at the mo­ment, you are ex­posed to events lead­ing up to the elec­tion, which could make you wrong.

For the record I think Bi­den has about 70% chance to win, so if you take this bet you will prob­a­bly win. Still a bad bet though IMHO.

Source: I have made a large frac­tion of all pos­si­ble in­vest­ing mis­takes. In bet­ting mar­kets I am ahead, hav­ing only ever bet on one horse race, in which my horse came in first.

• Let’s say Trump ac­tu­ally had a 50% chance of win­ning last time

What do you mean by this? That hu­man be­havi­our is non-de­ter­minis­tic? Or that, given the pub­li­cly available in­for­ma­tion at the time, the best guess was 50 per­cent? If the lat­ter, it’s easy to get a bet­ter cre­dence af­ter the event hap­pened. Look at Nate’s track record. An event that he gives an per­cent chance hap­pens pretty damn close to per­cent of the time. You could’ve just as eas­ily said he un­der­es­ti­mates the win­ners when he got it “right” (i.e. he said > 50 per­cent chance), and there­fore Bi­den has an even higher chance of win­ning.

• > What do you mean by this?

I mean that given he won, the ac­tual odds of him win­ning were ac­tu­ally bet­ter than 10%.

I can­not prove this ex­ter­nally but - be­fore the elec­tion in 2016 I said to sev­eral peo­ple, who re­mem­ber this, that it would be some­where be­tween a nar­row Clin­ton win and a strong Trump win. So it was not out­landish to think Trump could win. The main rea­son I had was what is now known as the “shy Trump voter” effect. Peo­ple did not want to get can­cel­led for ad­mit­ting that they were a ‘fas­cist’.

> Look at Nate’s track record

If Trump wins this one, as looks fairly likely at the mo­ment, NS will be 35 for Pres­i­den­tial elec­tions, no bet­ter than chance.

My main beef with the ar­gu­ment from cred­i­bil­ity is a) NS does not have a long strong track record in this field of pres­i­den­tial elec­tions, b) Cred­ible is differ­ent than to­tally ac­cu­rate. I pointed out if he was wrong by a few tens of per­cents, like last time, his view is not strong ev­i­dence that there is a win­ning bet here.

Fun­nily enough NS re­duced his P(Bi­den) to about 50% on elec­tion day (or maybe the day be­fore).

• Depends on where you bet. If you are in the right coun­try you can make a lot of money if Bi­den is 70% to win. I em­pha­sized pre­dic­tit be­cause I be­lieve most peo­ple on less­wrong are in the USA.