Oct 28, 2020
I recently received a text message wondering where the Presidential Election betting articles were at, and the truth of the matter is that I hadn’t put much thought into it this year simply because I really don’t know what’s going to happen. For the most part, I also felt as though I didn’t have the same sort of value insight that I had last year.
For the most part, I’m going to try to avoid political commentary not directly related to the betting aspects, but as everyone knows, it’s really tough to perfectly separate the two things. Given that this is an article, there’s also a little more flexibility in this regard. One thing that I will ask is that any comments try to focus pretty strictly on the betting aspects whilst avoiding opinion-based discourse as relates political matters.
DON’T TRUST THE POLLS
I said this four years ago when it was Trump and Clinton and I’m saying it again now. If you want to make any straight up bets on who is going to win the Electoral College, then the polls should be only a minor factor in that sort of decision-making.
The first thing that everyone should recognize is that the polls are extremely fallible and are all based on limited sample sizes. Another thing is that I used to be a supervisor at a telemarketing place and I can say for sure that many of the questions that pollsters ask are phrased in such a way as to lend themselves to certain results. Basically, results-oriented polling.
Another thing to keep in mind is that those conducting the polls will often do what’s called, “Dispositioning calls,” incorrectly. While you’re obviously not supposed to do it, people who are having difficulty getting ahold of leads will often just click, “Enter,” on the keyboard all the way through and the calls will be dispositioned as whoever’s name happens to be first. Generally, we would occasionally fudge our actual number of successful surveys, but even then, we coached the representatives to do it in such a way that they’d basically reflect polls already conducted…”Six this way, four that way,” and that sort of thing.
Another thing that people should keep in mind is that the order of the candidates when someone is asked can have an unintentional influence on some people. The best polls would randomize this order, but many polls do not and the person with his or her name listed first tends to have a slight advantage over the other candidate with those undecided. Some people aren’t really inclined to answer the poll, but wish to be polite (and not get called nineteen more times that day), so they’ll just default to saying whatever the first option the representative gives them is. I would say that a fair 25% of the calls would just have the respondents going through the motions in this way.
One aspect that it’s important to keep in mind is that even when the polls are technically right, the results won’t be what they indicate. Basically, people saw the 2016 results and came to the apparent conclusion that the polls missed wildly in favor of Clinton. However, when considered in the aggregate, the polls actually ended up being within the margin of error.
If you’re considering getting some late money down on the Electoral College result, another thing to consider is that the National Polls matter very very little. The Electoral College is not won based on National Popular Vote, so rather than a Trump win being based on a serious aggregate national polling miss, you only need the polls to miss in a few individual states...and from what I can tell...only Pennsylvania would have to miss (in Trump’s favor) outside of the aggregated margin of error for the state.
As always, 538 has the most sophisticated model for extracting the reality from polling results, but even that has its flaws.
The 538 model aggregates the results of all of the polls taken, both nationally and on the state levels. In addition to this aggregation, the model also gives slight preferential mathematical treatment to more recent polls, as that might reflect a switch from one candidate to another.
Another thing that the 538 model does, and this is the part that I don’t like, is that it essentially uses the national polls to make certain assumptions (or weighs in) as to the results of individual states. I’ll try to give an example:
Let’s say that a state tends to be five points more Republican leaning than the rest of the country. In addition to the polls for that state, what the 538 model will do is attempt to take the national polling results, but then use those state historical tendencies in their calculations to attempt to make assumptions about how that state will relate to the rest of the country.
Additionally, 538 has no idea what to do with, “Undecided Voters,” not that anyone really does. For example, let’s look at some of the problems with Pennsylvania.
The first problem here is that the 538 model assumes that more than 99% of voters will automatically vote for one or the other, which is not a safe assumption. As you can see from the final projections, 538 has 99.3% of the vote in the state going for one or the other.
However, if you were to look at polls that actually include the names of all the candidates, what we find is that Libertarian candidate Jo Jorgensen (who is hardly the only third-party candidate) has polled at a steady 3%. Even if we make the assumption that she loses a third of that support, she’d still be polling at 2%, which means that only 98% of the vote, ignoring all other candidates, would be available for Trump or BIden.
The second step in 538’s process also involves an EQUAL allocation of undecided voters. My question is simply this: WHY?
Even when it comes to voters who are truly undecided, (other than those who are voting for third-party candidates and are mostly ignored as a result) what evidence is there that these voters do not favor one candidate over the other? Granted, if you held a gun to my head and forced me to make an assumption, roughly equal is the assumption that I would make---but nobody is holding a gun to my head. The fact of the matter is that it remains possible that almost everyone who was going to vote for Biden has already decided to do so and is being represented---perhaps more than represented---in the polls.
Which means that it’s possible that the so-called, “Undecided Voters,” some of whom are again actually firmly voting for third parties, are really just deciding whether they want to vote for Trump or someone else, but have already ruled Biden out.
One thing that I’ve noticed about polls that refer to voters as, “Undecided,” which some just automatically do if the respondent is not voting for one of the two major party candidates, is that such polls never seem to ask this question, “Are there any candidates that you will absolutely NOT vote for?”
While it would be difficult to pull off, the kind of poll that I would want to see in Pennsylvania would include a sample size of several thousand undecided voters and would ask only this question, “Of Trump and Biden, is there a candidate you will absolutely NOT vote for?”
If you can get a meaningful sample size of such answers...and actually include third-party voters as third-party voters rather than undecided, then you’ll end up with a pretty good idea of the percentage of people voting for either one or the other and where people who are undecided are more likely to break. I tend to feel like the people who want Trump out REALLY want Trump out and most of those remaining are deciding whether to vote Trump, third-party or just sit this one out. I do want to say that’s just a gut feeling based on my conversations with folks in the state.
The third step that 538 does, and probably the most objectionable, is that they will take the results after allocating undecided voters and then adjust those numbers based on demographics?! What the hell kind of scientific method goes into that? The polls are the polls, so what is it that 538 is saying---that the people being polled are wrong about who they say they are going to vote for?
I totally get that they want to factor in demographics and historical tendencies of a state into the equation, but there comes a point where you’re just being, “Too smart for your own good,” which is exactly why Trump won in 2016---and why everyone was so surprised by it---to begin with.
In other words, I would sooner take the aggregated polling at face value rather than make what amounts to theoretical adjustments that swing the numbers by 1.2% from one to another.
Almost SIX MILLION PEOPLE voted in Pennsylvania in 2016 with the difference between Trump and Clinton being fewer than fifty thousand votes. What this, “Demographics Based Adjustment,” is doing is essentially casually taking more than 60,000 votes---more than the 2016 margin of victory in the state---and casually tossing them from one candidate to the other.
Perhaps worse than that, that adjustment is tossing those 60,000-some votes from Trump to Biden? Wait a minute, didn’t the polls from 2016 miss in favor of Clinton? Why the hell are we doing adjustments to the aggregated polling numbers to intentionally favor Biden?
In short, this particular adjustment, in my non-expert opinion, does not make very much sense. Demographics don’t vote; people vote. Perhaps it’s possible, if not likely, that one candidate or another appeals to certain types of people in a non-quantifiable way that supersedes the supposed importance of demographics.
The next thing that 538 does is a fourth step that throws some votes back over to Trump. What they do here is take people who weigh the economy as the most important issue (these types of voters slightly favor Trump) and factor in an incumbency advantage. With all of that, that particular model says that not only would Trump win, he would carry more than 50% of the popular vote (HIGHLY UNLIKELY) in the state. For that reason, 538 only counts this as 4% towards the final analysis and 96% everything else.
And...the final analysis of this writing: Trump 47%, Biden 52.3%.
I guess third-party voters don’t exist anymore. Another thing about this analysis is this is exactly what they arrived at when they decided that nearly 100% of people were going to vote for one or the other AND that undecided voters were going to break exactly equally.
Well, what the hell was the point of Steps 3-5 just to get to where Step 2 had us?
Another factor that should be mentioned is that people who consider the economy most important AND incumbency, in theory, should have already been factored into the ACTUAL POLLING. That’s why you ^&$*$$*%& do polling in the first place.
Anyway, the final result there just seems really bizarre and extremely results-oriented.
The final thing that should be mentioned is that, while rare, sometimes recency bias can be a good thing to have. Simply put: Polls taken several months ago are absolutely no longer relevant and taking them into consideration underestimates the impact of voters who may have actually switched from one to the other.
PUBLIC SERVICE ANNOUNCEMENT: The only thing that you would really need for Trump to win the State of Pennsylvania is for the polling results to be off on the high end of the margin of error AND for more actual undecided voters to be breaking in his favor as opposed to that of Biden. Another thing that should be mentioned is that Libertarian defectors who currently say they are voting for Jo Jorgensen are going to be slightly more likely to go Trump if they change their minds on her.
WAIT, DIDN’T YOU JUST SAY DON’T TRUST THE POLLS?
I did say I don't trust the polls, but I have to admit that I have slightly more faith in the polls than I do demographic-based hocus pocus that I feel is being used to achieve a particular result. Without getting into any actual political views, I would be remiss not to admit that I find 538 much more left-leaning than they used to be. Granted, 538 readers are going to be highly educated, (which surveys would say makes them more likely to be left-leaning---sorry) but everything that they have written recently seems to be from the perspective of, “How does this work for the Democrats,” or, “What do the Democrats need to do?”
I honestly just feel like their coverage of 2016 at least tried (with a few undertones here and there) to play it more down the middle.
Anyway, the most likely answer comes from something that nobody is really doing---which is to ask a simple, but specific, question of voters who are actually undecided, “Of the two, is there one you would absolutely NOT vote for.” At that point, it becomes a pretty safe assumption that many of these people are going to be voting for the other.
One other thing that I don’t like is the way they play national polls into it and then make those demographic-based adjustments and adjustments based on a state’s voting history. Those things don’t matter. Different election, different candidates. There was a time when West Virginia (who will probably have the second-highest Trump percentage) was considered a safely Democratic state. Hell, Reagan lost West Virginia in his inaugural bid, if that tells you anything.
I don’t know. I tell you what we’ll do: We’re going to do a little experiment this year where I take the 538 projections from the day before the election, compare them to the polls and then compare them to what actually happens. If it turns out that their predictions knock it out compared to the polling, then I’ll happily write a follow-up apology article and eat my humble pie.
I just think what they are doing amounts to (over)educated guessing, but I guess we will find out.
JUST TAKE THE FREE MONEY!!!
Okay, the first thing that I want to do is discuss the fees on Predictit.org before we get into anything else.
The fee structure on PredictIt is both simple and twofold:
- If the binary option resolves in a win, they will trim off 10% of your profits. If the binary option is sold at a higher price, you will lose 10% of those profits.
- The site will take a fee of 5% on any monies withdrawn.
Therefore, any advice given---which is also for entertainment purposes only and is based solely on my own opinion---is going to be better for those already on the site. In fact, nay possible bet is automatically better for those who have money on the site because, even if you deposited $100 and did nothing with it, you would only get $95 of it back if you were to withdraw your funds thirty days later.
As you’ll recall in 2016, I was unsure about who was going to win the Election, but found that there was excellent value on Trump. They;ve done the binary options a bit differently this year, so rather than it being Trump/Biden, it’s Trump Yes/No and Biden Yes/No separately. PredictIt has also decided to offer some normal binary options and specific binary options on individual states.
When it comes to sports betting, research and insight helps skilled bettors to take advantage of market inefficiencies. When it comes to PredictIt, you’ve got people, “Betting with their hearts,” even more so than with sports betting, so often, just basic understanding of extremely fundamental facts will help you take advantage of market inefficiencies. I’m going to list the bets that are out there that I like as of the time of this writing:
Which Party Will Win California in 2020: Democrat $0.94, Republican $0.08
This is free money, but only if you already have money on Predictit. There is no way in a million years that Joe Biden loses in California.
The only problem with this option is that you can’t really deposit funds to make this bet. The reason why is because you would bet $94 to purchase 100 shares. WHEN you win, 0.6 cents of your profits will be subtracted from your result, but that’s going to round up to the nearest penny and be a loss of $0.01/share.
After that, you will pay $0.05 per every dollar you withdraw, which will be a total of $5. In other words, this is a mortal lock that will result in you not only ending up exactly where you started, but also in having to wait thirty days to get your money out.
That being said, if you ALREADY have money in Predictit and want to add to your balance for free, this is one way to do it.
Which Party Will Win Hawaii in 2020: Democrat $0.94, Republican $0.06
This is the same question as California in an even safer state for him than that one. I don’t know how a state can BE safer than California, but based on state polling, this one is.
Again, this is only useful if you already have funds in PredictIt and just want to grow your balance.
Which Part Will Win Ohio in 2020?: Democrats $0.30, Republican $0.73
I don’t love this if you’re looking for an outright win, but Ohio has been something of a bellwether state in that it has only missed correctly picking the President twice since 1896.
There’s no question that Biden is favored to win overall, so that being the case, Ohio should be considered more than 30% likely (implied odds) to go for Biden. 538 has the probability of a Biden win at 42% in Ohio and the polls have it really close, based on in state polling with Biden leading in many of the polls.
I’m not personally putting anything on this one, just identifying value where I see it. This would also only be a value bet ONLY if you already have money in PredictIt because you would effectively only be paying the additional 5% cash out fee:
A.) If you withdraw to begin with.
B.) Effectively, only on your winnings. The reason why it’s only on the winnings is because you would be paying 5% withdrawal fee on your original money if you cashed out right now anyway.
Therefore, you win $0.70 and take your 10% haircut that knocks you down to a win of $0.63 and an additional 5% of that which is another 3.15 cents for a total effective win of $0.5985. If we look at a 42% probability of Biden winning the state:
(.42 * .5985) - (.30 * .58) = 0.07737
Which represents an expect profit of about seven cents on a thirty cent bet for an advantage of:
37.7737/30 = 25.9%.
Let’s suppose that I think Biden is only 40% to win Ohio, just replace:
(.40 * .5985) - (.30 * .60) = 0.0594
35.594/30 = 1.18646666667
Which means that I think I have an advantage of about 18.65% on the bet.
Even if I thought the probability of Biden winning Ohio was only 35%, I would still enjoy a nearly 5% edge on this bet, so this one might be worth it for you depending on how likely you think he is to win the state.
I’d like to reiterate that I’m not playing this one. The market inefficiencies that can regularly be found on Predictit, often on sure wins, are such that I’m really not interested in playing most long-shot type things unless the value is extremely good.
Which Party Will Win Pennsylvania in 2020?: Democrats $0.60, Republicans $0.44
Say what you want, but of all of the state polls, only three in as many months have had Trump winning. Not only would the polls (in aggregate) have to miss outside of the margin of error for Trump to win Pennsylvania, but there would also need to be a sort of systemic polling bias that’s experienced amongst all pollsters such as to render polls meaningless.
538 gives Biden this state 85% of the time, but I like ¾ or 75% for Biden to win. We know that we will lose $0.04 for every $0.40 won if this goes the way of the Democrats, so that knocks us down to a $0.36 gain per share. We will lose another 1.8 cents of that if we cash out (assuming you already have money in there) for a final gain of 34.2 cents per share if it wins.
(34.2 * .75) - (.25 * 60) = 10.65
That represents an expected profit of 10.65 cents per share for an expected final result of turning sixty cents into 70.65 cents, so:
70.65/60 = 1.1775
Which represents a 17.75% edge and I will say that I’m on the fence about taking it. It’s hardly a sure thing, but what I do see is recent line movement towards the Republican side and I think an extreme outlier poll released close to the election (that favors Trump) might cause the Democrat side of this line to panic and sell off. I’ll probably wait to see if I can take advantage of an event along those lines, and if not, oh well.
Vermont and Washington
***Vermont and Washington are basically variations of Hawaii and California for the same reasons. It’s not much free money (and it ONLY works if you already have money on the site) but it’s there for the taking if you want it.
Popular Vote Question---Which Party Will Win the Popular Vote?
There are too many possibilities here to list as they all include percentages.
As of the time of this writing, don’t take any of them. Instead, look at this:
The total on Democrats winning the popular vote in the percentage question is $0.83 if you want to cover every possible way they could do it, whereas it’s only $0.82 to say that Trump will NOT win the popular vote.
Listen, it ain’t happening. If Trump didn’t beat Hillary Clinton (the second least favorable candidate---by favorability rating polls---in history) then he’s not going to beat somebody who actually has a net positive favorability rating.
This is another good one if you already have money on the site. In my opinion, Democrats should be 95%---at an absolute minimum---to win the popular vote. Because of that you will win $0.18 per share if that happens of which you will lose 1.8 cents in fees for gains of 16.2 cents per share. You will lose 5% of that (if you already have money on the site---otherwise---you’re also losing 5% of whatever you must deposit) for another .81 cents leaving 15.39 cents.
(15.39 * .95) - (82 * .05) = 10.5205
Therefore, I put my expected profit at 10.52 cents, per share, on something that costs me $0.82/share to get, therefore:
92.5205/82 = 1.1283% (Rounded)
In other words, I feel like I have a 12.83% advantage on this one and think that Biden is REALLY going to run up the numbers in states such as California and New York to an even greater extent than Clinton did in 2016. The polling also indicates that Biden is nearly a ten point favorite, so this would have to be a miss of historical proportions, and without precedent, for Biden to actually lose the national popular vote.
It’s not the best value so far, but it’s the one I feel the most comfortable with. I also consider my assumed probability of 95% to be extremely conservative.
***But if you want a sure thing…
The one thing I noticed about the percentages ones is that you can cover every possibility EXCEPT anything that would have the GOP winning by 4.5%+ of the popular vote for $0.91. This has a 99.9999999% chance of occurrence and SEEMS like a great value bet!
IT IS NOT DO NOT DO THIS!!!!
The reason why not is that this series of bets would cover such things as the Democrats winning by 1.5%, or less, which can be bought alone for $0.06 on the yes. The problem here is that you would pay 9.4 cents per share on the commissions if you did win, so just the commissions on the win end up being more than the total of 91 cents per (all) shares that you would be betting. Granted, you would not be losing that six cents, but you would also face the 5% withdrawal fee on cashing out...so the whole thing ends up basically being a wash.
Electoral College Guaranteed Win?
Okay, this is going to be a really weird one, so check this out:
GOP 280+--No $0.97
GOP 210-279---No $0.97
GOP 150-209---No $0.95
GOP 100-149---No $0.87
GOP 60-99---N0 $0.85
GOP 30-59---No $0.90
GOP 10-29---No $0.95
GOP 0-9---No $0.97
Dems 1-9---No $0.98
Dems 10-29---No $0.96
Dems 30-59---No $0.95
Dems 60-99---No $0.93
Dems 100-149---No $0.87
Dems 150-209---No $0.85
Dems 210-279---No $0.89
Dems 279+---No $0.90
Okay, so if you took every single one of the Nos at those prices, you would pay a total of $14.76 per set of shares as long as you can get these prices.
The way this is going to work is kind of weird. The worst result possible for you on these sixteen unique bets is that you will lose $0.98 on one of them, which will leave you with $13.78 of the original monies bet. However, that will mean that the other fifteen bets win by default, which results in total winnings of $1.22 per set of shares.
Of those winnings, you will have to pay 12.2 cents in commissions and would also eat 5% of the remaining $1.098/set of shares, which is an extra $0.0549 per set of shares. That leaves a total of about $1.04 in winnings per set of shares that you are essentially trading for a loss of $0.98 for a guaranteed gain of about $0.06/set of shares. In other words, it’s a guaranteed profit if you can get it at exactly those prices...which is probably going to be pretty tough to do as large share purchases will cause movement all over the place.
I just thought this was an interesting thing to point out if you can figure out a way to guarantee yourself those share prices all at once.
While that might be a tough thing to do all at once, similar situations will occasionally pop up as a natural consequence of just watching the line movements on things for the longer running propositions.
Kanye Third Best?
You can get the, “No,” on Kanye West getting the third most popular votes for $0.93/share. I swear on my life this is a thing as of the time of this writing. Jo Jorgensen will get the third most votes and you should collect some free bucks on this Kanye West, “No,” if you already have money on the site. It’s not much, but it’s being given to you.
Instead of Biden…
This is not actually advice on making a bet as I have no position on this one, but one thing that I did notice is that, if you want to take the, “YES,” on Biden to win, the cost is $0.63/share. However, as of the time of this writing, if you want to take the, “YES,” that there will be a female Vice-President, the cost is only $0.59/share. Biden’s running mate is female, therefore, it’s the same question.
Oddly enough, the Trump YES market and female Vice President NO markets match perfectly.
New York for Biden---YES $0.92.
This is better than the California and Hawaii ones. Take this if you already have money on the site as this is 99.99999% likely to happen. Even if you don’t have money on the site, you could deposit and make a small guaranteed profit with this one.
$0.08 gain per share, $0.072 after the site’s cut. You will turn $0.92 into $.992 of which you will lose 5% when you withdraw which will leave $94.24 (assuming you get 100 shares) for gains of $0.0224 per share on this free money, even if you have to make an initial deposit.
.9424/.92 = 1.02434782609
Okay, so the advantage is only 2.43478% assuming a win probability of 100%, but free money is free money and it’s not often that you’ll get an opportunity like that with having to make a deposit. For the record, 538 puts the probability of Biden winning the state at YES, he will, why are you asking?
Who Will Win the Presidency?
I have no idea, but I do think that the winner is he who wins Pennsylvania, one way or the other. I guess I would say that I think Biden is the favorite, but the betting markets seem to have it about right and I think 538 is too low on Trump this time around. It was the other way around in 2016 where 538 gave Trump a pretty good chance while the betting markets hated him.
Once again, this article is strictly for entertainment purposes, so any bets that you make are fully at your own risk and I am just sharing my opinions on some of the Predictit lines and market inefficiencies that I think I have identified. I would definitely do more research aside from this article before getting any actual money down if I were you.
In the meantime, please leave a comment as to your thoughts on some of my ideas as well as any lines that you like. There will be a follow-up article after the election where I compare 538 against the polls as well as the actual results, so that will be fun. In that article, we will also take a look at how my picks went and what the gains would have been, if any. Maybe the losses (Gasp!).
"Bush v. Gore ... was a decision of the United States Supreme Court on December 12, 2000" per Wikipedia
I'm not in a state where I can bet on anything like this, and from what I can tell, the last time I ever checked anyway, it is actually not that easy to even find out what the odds are on anything. But I wonder if there is a bet you can make whether or not the election is decided by the SCOTUS this time ... and when
I'm thinking the date the election is finalized is going to be later than Dec 12th this time, but I'm not sure how a bet could be designed seeing as how Trump or Biden has to concede if it doesn't go to court. Some people think there will be more than one lawsuit going on
methodical about polls but too mad to realize the methods are flawed.