If there is one thing about Advantage Play that often proves true, it’s that a savvy Advantage player definitely keeps an eye out and considers every possibility, even those that seem unusual. This was no less true during the Presidential Election for the U.S.A. this year, which, as surprised many people, saw President-Elect Donald Trump defeating Hillary Clinton and winning the election.
My looking for opportunities in strange places, as I often do, started with an Article that I wrote for our sister site, WizardofVegas.com, on the Presidential Election.
The long and short of that was that I determined, as of July 28th, 2016, that Betfair was offering +245 if someone wanted to bet that the Republican Party (read: Donald Trump) would win the Presidential Election.
For someone that has an interest in the electoral process to begin with, this became an interesting case-study for me from a betting standpoint. When it comes to the actual probability of a candidate winning the election, one source that I trust pretty implicitly is www.fivethirtyeight.com. The website in question was formally owned by Nate Silver (who remains Editor-in-Chief) and is now a subsidiary of ESPN. The focus of the site is to use statistics and probability to determine things such as elections, and they also cover all of the other major sports using statistical and probabilistic angles that are interesting and unique.
At the time that I first found the +245 line for someone wanting to bet on Donald Trump, FiveThirtyEight had his probability of winning at 42.3%, if we use this MoneyLine converter.
Based on the 42.3% probability of winning the election, one can conclude that a, ‘Fair Line,’ for Donald Trump to win the Presidential Election, at the time, would have been +136.407. In other words, for a $100 bet, the pay being offered by Betfair at the time was in excess of $100 more than would have been fair based on the probability suggested by FiveThirtyEight.
That would obviously lead some people to ask the question: What if FiveThirtyEight is wrong?
That’s certainly a fair question to have asked, at that point, and it was a question that I had been posed in the Comments section of that Article. However, when it comes to the actual probability that Donald Trump was going to win, FiveThirtyEight would not just have to be wrong... they would have to miss it by a mile!
The methodology used by FiveThirtyEight is that they used the polling results of National Polls as well as State Polls (for the states that did polling, and as polls come in) in order to come up with an overall national popular vote advantage, as well as to attempt to derive a probability of winning for each state. In some cases, they also use historical state (party) voting data for states in which not much polling has been conducted and they compare that data to the popular vote advantage held by one candidate or another. The ultimate goal, at all times, is to end up with a conglomeration of data that yields a reasonably likely probability of winning for individual states as well as the national vote at-large.
Furthermore, FiveThirtyEight does not simply take a mean average of the polls. Due to different polling methodologies that different pollsters may use, FiveThirtyEight determines that there is a concept called, ‘House Effects,’ which means that the polling data may swing one way or another from what is actually expected to happen. For example, if one pollster has missed in its polling, (as almost all pollsters do, to some degree) but that miss has almost always been in favor of a specific party, (Democrat, Republican) then FiveThirtyEight will attempt to adjust its matrix to account for the house effect of that pollster in determining the probability of winning. Furthermore, the site will also tend to value newer polling data as opposed to older polling data and weight that accordingly, primarily because newer polling data reflects the fact that individuals may have changed their minds, and furthermore, it also reflects individuals who were previously undecided deciding on one candidate or the other.
Anyway, having determined that FiveThirtyEight was giving Trump a chance in excess of 40% compared to a line of +245, which implies a probability of winning of roughly 28.99%, (same source) one comes to the conclusion that for a bettor to have an advantage of the +245, the site in question would not just have to be off, it would have to be WAY off!
It is also important to remember that the probability is not based upon a single pollster, but it is rather based on every single pollster that there is. It is not just based on the National Popular Vote polls, it is based on individual state polls, as well.
The Other Bettors:
If there is one thing that is true about, ‘Ploppy Sports Bettors,’ (read: Those that have little to no knowledge and do not play with an advantage) it is that they often tend to favor whatever outcome has the greatest probability of happening. It is for that reason, when betting NFL Games, sharp bettors can more often find advantageous lines taking the underdog than they will taking the favorite. The natural tendency of people is to take a look at the most likely result and subjectively infer from the fact that it is the most likely that it is also the best value, despite what they may have to lay to take that position.
In other words, many inexperienced bettors, and we are talking about people who may not bet sports AT ALL, but were only interested in betting the Presidential Election (and perhaps other political bets) were pretty clearly swinging the line to give Hillary Clinton a larger advantage than she really had. In fairness, someone might ask, “Now, Brandon, you can sit there and talk about this now that Donald Trump actually won, but how can you say you were right at the time?”
I can say that I was right at the time because I was right at the time! Even in that Article, I suggested that I did not believe that Donald Trump would actually win the election, (I guess I was wrong about that) I simply suggested that there was a lot of value in Donald Trump at +245 and that it was a good bet. Once again, any opinion to the contrary, at that time, would just be people thinking the favorite is going to come in by virtue of being the favorite.
It really is quite that simple, people often lay money on the favorite in a certain proposition without fully appreciating the concept of what they are laying and what that means for the implied probability of their bet. They just figure that the event in question has a greater than 50% chance of happening, and this usually gets ascribed to people who may even believe the event to have a 75% (or better) chance of happening, and they are not afraid to lay accordingly even if what they are laying is a bad bet compared to what they think the probability of winning would be.
For proof of concept, let’s take a look at the Seattle. V. Carolina NFL game for Week 13. At the time that the game opened, the MoneyLine for Seattle was -275, which means that they were implied to be a huge favorite. The night before the game, the Vegas Insider consensus line sits at -360 with some casinos requiring a player to lay as much as -400 if they want to take Seattle straight up. While that has changed, the point spread has changed from -6.5 to -7.5 from the time that it opened until the night before the game.
According to the MoneyLine, Seattle went from an implied 73.3333% probability of winning the game to an implied probability of 78.26% or 80% exactly, depending on whether you are using the consensus line or the fact that some books have them at -400. What is especially intriguing about that is the fact that if we look at those point spreads of -6.5 and -7.5 and compare them to what is shown on Wizard of Odds for the expected probability of winning.
The Odds site shows an estimated probability of winning for Seattle of 70.3% based on the opening the point spread as well as 73% based on changing the point spread to -7.5 instead.
The question is: Why would the implied probability of Seattle (a huge favorite no matter how you look at it) of winning be less based on the Point Spread than it would based on the MoneyLine, is the house taking more juice on these bets?
Granted, the house taking more juice can be part of it, but let’s also consider that the difference between the two at the time that the lines opened was an implied probability 3% greater on the MoneyLine than on the Point Spread, but after both of those things moved, the MoneyLine reflects a probability of winning in excess of 5% greater than that reflected by the Point Spread according to the VI consensus figures for both the Point Spread and the MoneyLine. If you use the MoneyLine being offered at some casinos at -400, then that is more than double what the implied probability difference was based on the MoneyLine against the Point Spread.
The question then becomes: Why should the implied probability of Seattle winning on the MoneyLine improve moreso than the implied probability on the Point Spread, and why does it start out higher?
For one thing, point spread bettors (as a group) are generally more sophisticated than MoneyLine bettors, at least in terms of having a better understand of the proposition at hand. When it comes to huge favorites on the MoneyLine, there is a tendency for someone to say something like, ‘I think Seattle is almost definitely going to beat Carolina,’ and that is good enough for them. The difference between smart bettors and bettors who do not bet so well is the simple fact that Seattle will probably win the game (which is true) is not going to be enough for an intelligent bettor. What the intelligent bettor wants to try to figure out is Seattle’s actual probability of winning and what that means for how much money has to be laid.
The mistake that we see here is that many players are willing to lay increasing amounts of money in order to get on the side of the proposition that says Seattle is going to win the game outright. We saw this with the Presidential Election to an even greater extent because Hillary Clinton, at almost all times, was the favorite to win. The reason it happened to a greater extent is because you had people placing bets on Hillary Clinton that, generally speaking, are simply less well-versed when it comes to the concepts of probabilities and how they effect how much money someone has to lay on a, ‘Yes/No,’ proposition, in general.
Different Types of Betting:
The Presidential Election betting had some serious ups and downs when it came to the betting lines, in fact, the ups and downs were among the quickest that we have seen in recent memory, especially when it comes to the downs on the Donald Trump side. The reason that was the case was the tendency for Donald Trump to say things that would be considered gaffes, and the way that he was (perhaps rightfully) slammed by the media at-large for doing so.
The general course that the Presidential Election implied odds, as well as those probabilities being reported by FiveThirtyEight took was that Donald Trump would make what was considered a serious blunder, Hillary Clinton’s lead in the polling would improve substantially, Donald Trump was getting several news reports that cast him in a negative light very quickly, and as a result, the implied odds would move very quickly to Hillary Clinton. As any negative publicity absorbed by Trump started to die down a bit, Trump’s position against Hillary Clinton would gradually begin to improve again.
I would suggest that the best type of betting in order to take advantage of these cycles would be Binary Option betting, if someone could find Binary Options through a legitimate source that didn’t charge too much for the trades. For those who were knowledgeable enough, they could BUY the position on Donald Trump at an extremely low cost, and then hold that position as his implied probability gradually recovered and that recovery was reflected in what the position could be sold for. For my part, my tendency would be to buy a position anywhere under 30 (based on 100 max position) and then try to sell it in the low-40’s, whenever possible.
Of course, there was also value, at the time that I wrote that Article (and certainly at other times) simply to bet on Donald Trump straight up on the MoneyLine. Again, I don’t know that those bets were really the best when it comes to probability of winning, (as Trump’s probability of winning only bested 50% for very brief periods) but the goal is to have value when betting, and one does not necessarily have to expect a positive result in order to have value.
The Wizard has done exhaustive research on Half Point Parlay betting in the NFL, and has put a ton of money where his mouth is, to very positive results!.
If there is one thing that stands out about this half point parlay betting, it is that the player can achieve extremely positive value, despite the fact that the player has a very low probability of winning on any individual proposition, which is why I bring this up. The point is the fact that one side or the other having a greater tendency to come in has no impact on the odds whatsoever, how much you are laying (or getting) vs. the actual probability of the event coming to fruition is what makes a bet either good or bad.
Most people are familiar with the game of Craps and understand the concept of a Place Bet, which is simply a bet that either a 4, 5, 6, 8, 9 or 10 will come before the next seven does. These bets can be made individually, or one can choose to Place Bet multiple numbers at once. Imagine that, after the Come Out roll, the point is a four, the player could then make place bets on 5, 6, 8, 9, and 10 planning to take down all of the Place Bets if any one of those numbers hits before a seven. Interestingly enough, a player deciding to make such a bet has a 21/27 probability of winning on the overall proposition, however, each individual Place Bet is a negative expectation bet, therefore, the sum of these place bets is a negative expectation bet. The player might be more likely than not to win on an individual trial, but the fact that the player will lose the sum of the bets 6/27 times is more than enough to trump the fact that the player is more likely than not to have a winning Place Bet occur before a seven.
Again, it was the same case as with the Presidential Betting, because while people had a tendency to look to the National Polls and favorability ratings (Hillary was definitely winning on those, at least, she was less unfavorable) in order to determine what they thought the result of the election would be, what they did ignore was a number of fundamentals. The first fundamental that was pretty actively ignored was the fact that Republicans were pretty much only going to vote for the Republican candidate, at least, for the most part. To that extent, there were a number of, ‘Solid Red,’ states that Trump essentially had in the bag regardless of what happened elsewhere. The fact of the matter is, as much as America prides itself on Democracy, only a handful of states really have any meaningful effect on deciding who the next President will be. The vast majority of states always go one way or another.
So, What the Hell Else Happened?
Inevitably, it wasn’t simple ignorance of the fundamentals that allowed people to get their money in good on Presidential Election betting, nor was it ignorance of the fundamentals that led to Donald Trump being the new President-Elect. There were a number of things that happened in the meantime.
For one thing, it turned out that the national polling averages were somewhat off, but they were not off by nearly as much as people thought they would have to be for Donald Trump to win. Hillary Clinton won the Popular Vote by 1.9% and will likely win it by a full 2% when all is said and done. The National Polling averages had Hillary Clinton winning the popular vote by about 3.9% when they closed, so that is hardly an unprecedented polling miss. Furthermore, the National Polls had Gary Johnson (Libertarian) with nearly 5% of the vote, and he didn’t come anywhere close to that.
Gary Johnson, more than anything, was something of a protest choice against both Hillary Clinton and Donald Trump for many people. However, the tendency for third-party candidates in general, and especially in the case of Gary Johnson, is that they tend to fade as the election draws closer. It could be an effect of people not wanting to, ‘Waste a vote,’ (I voted for Gary!) as they might say, by voting for a candidate that, ‘Has no chance of winning.’ For someone thinking along such lines, they have to vote for someone (or not at all) and so they are left choosing one of the two major party candidates. For whatever it is worth. Libertarianism is essentially economic conservatism combined with being liberal on social issues, but when it comes to people who were not fans of ObamaCare, (fiscal conservatives) it makes more sense that they would gravitate towards Trump.
Furthermore, to the extent that Gary Johnson was a potential protest vote for many people, Donald Trump more closely resembles a protest candidate than does Hillary Clinton. Therefore, for many people who did not want to vote for a major party candidate by way of protest, but made the, ‘Gun to my head,’ decision, Donald Trump being an, ‘Outsider,’ more closely resembles a protest candidate. Furthermore, while the effect may be negligible, Green Party candidate Jill Stein could be more or less inferred to have directly, ‘Stolen,’ all of her votes from Hillary Clinton. There is also something of a question as to whether or not Gary Johnson didn’t hurt Clinton more than he did Trump, that answer is unknown at this time, but Gary Johnson was a good choice for Liberals who are more concerned with social issues that wanted a protest vote.
Michigan is one state, for example, in which Jill Stein may have seriously hurt Hillary Clinton. As of 11/22/16, Donald Trump won Michigan by 0.3% of that state’s vote while Jill Stein managed to achieve 1.1% of the state’s vote. It is almost doubtless that, had Jill Stein not have been in the race (and all other things equal) that Hillary would have won Michigan. Wisconsin is another example of a state in which, if Jill Stein’s votes moved to Hillary Clinton’s column (though, literally, it would have to be almost all of them) Hillary Clinton would have won the state.
One of the many problems with Presidential Election polls is the fact that they do not necessarily list all of the different options, to wit, some of the state and national polls only asked for people to choose between the two major party candidates, and so that is what the people who are completing the polls do: they choose one or the other. Naturally, that doesn’t mean that those people will actually choose one or the other when they are actually in the ballot box touching the screen, punching holes or filling in circles. (It varies from location to location)
For that reason, if there were polls in either of those two states that did not mention Jill Stein at all, and the Green Party being an extremely liberal party, what’s going to happen is that all of the people voting for Jill Stein, when presented only with two choices, are going to go ahead and answer Hillary Clinton for the purposes of the poll. Once again, we still cannot be sure whether Gary Johnson, ‘Pulled more votes,’ from Hillary Clinton or Donald Trump, but if it was Clinton, then that could have also had a detrimental effect on the accuracy of polls that did not include Gary Johnson.
One thing that should be noted is that swinging both Michigan and Wisconsin from Donald Trump to Hillary Clinton is not enough to swing the Electoral College back to Clinton, so the winner would remain unchanged. If we look at some of the other states where Clinton was supposed to either win or be extremely competitive, Hillary Clinton lost by such an amount that even pulling EVERY SINGLE JILL STEIN VOTE over to Clinton would not have been enough to change the outcome in those states. In fact, Pennsylvania is the only one that is even arguable in this regard, and that’s only if you assume that Gary Johnson pulled substantially more voters from Clinton than he did from Trump. Of course, if Clinton had won Pennsylvania, Michigan and Wisconsin (her closest losses) that would have been enough to swing it. An outside argument could also be made for Arizona, but again, Gary Johnson would have had to been pulling almost ALL of his votes from Hillary Clinton...and I seriously doubt that is the case.
I’m not saying that any of these things actually point to an easy Trump win, I’m not even saying that taking these things into consideration would have made Trump implied to be more likely than not to win, I’m simply suggesting that these considerations made Donald Trump appear to have a much better chance than to look at the bare National and State polls which often did not include third-party candidates. When you combine fundamentals with unknowns, I would say that at no point did it look like Donald Trump had less than a 30% probability of winning the election.
Polling Inaccuracies:
While the National Polls, taken as a whole, will have proven to be off by right about 2%, the State polls were all over the place in terms of how erroneous they were. We have already discussed how third-party candidates may not have been taken into consideration in any meaningful way, and how when they were considered and people chose them (but later changed their minds) those who may have been voting for Gary Johnson likely broke for Trump. I would suggest that those who actually did vote for Gary Johnson might have actually favored Clinton slightly, (overall) but those who only liked Johnson as a protest candidate might have broken off and went to Trump. Once again, I don’t even know that Gary Johnson’s actual voters would have made a, ‘Gun to the head,’ decision of going for Clinton over Trump, and I especially don’t know that to be the case in the three to four states in which such a question would even matter, but that’s my working hypothesis.
The question is: Why were the National Polls more accurate than State Polls, and why should this have such an effect overall?
The answer to the first question is actually extremely simple: The conglomeration of State Polls adjusted according to recency and house effects is simply a smaller sample than that of National Polls. Simply put: More National Polls are conducted, so the weighted average of National Polls is more likely to be correct than the combined polls of any individual state. Provided the National Polls are distributing their polling more-or-less according to population, then the results should be closer to being in line with what will actually happen. Again, while a full 2% is a pretty decent sized miss, it is nowhere near unprecedented.
The answer is only slightly more complex with respect to state polls. Once again, the prior should be that because fewer states actually matter, fewer state polls actually matter. California, for example, ‘Missed by a mile,’ in state polls as Hillary Clinton won 7% more of California’s vote than the state polls would have led one to believe, but the fact of the matter is that California never mattered, no matter how you cut it. California’s polls suggested that Hillary Clinton would win in a huge landslide, and instead, she just won in an even bigger landslide. That Hillary Clinton would win California was never in doubt.
The biggest state poll miss in a state that actually had Clinton as a favorite that she would go on to lose was Wisconsin, followed by Pennsylvania, North Carolina and Michigan. These state polls all missed in excess of 4%, and with exception to North Carolina, these states are all quite similar demographically. What has since been determined is that these state polls all had one thing in common, the votes of caucasians who are not extremely well-educated were underrepresented in the polls, and that was enough to make a huge difference.
The fundamentals of those states, previous manufacturing-center giants, were all such that they bode a bigger advantage for Trump than other states with similar demographics. Ohio is another example of such a state, but prior to Election Day, Trump was expected to win Ohio, anyway.
In addition to the fact that the polls in Wisconsin missed wildly, (6.1% off) there is also the practical matter of the fact that Hillary Clinton spent exactly zero minutes actually campaigning in Wisconsin. Ironically enough, the biggest reason for that was probably what the polls said about Wisconsin: Hillary Clinton would basically win it in a landslide. It was due to these polls that Hillary Clinton assumed she had it in the bag, so why spent time and financial resources in a state that you have essentially already won?
I am not claiming to have known ALL OF THIS prior to the election and certainly not when I wrote that Article for WoV back in July, but what I am claiming is that I knew that many of these states (and others) had not yet conducted enough polling that they should be considered absolutely conclusive. In fact, states such as Ohio, Pennsylvania, Florida and North Carolina even swung back and forth a few times in state polling, and had swung even prior to that point.
Does that mean that I thought Trump would win the Presidential Election? Absolutely not, in fact, I’m still somewhat surprised. It was only several hours into watching the Election coverage on Election Day that I would have even entertained betting Donald Trump at even money. Specifically, when I saw Trump really start to pull away in North Carolina, I thought, ‘Uh-oh.’
Conclusion:
If there is one thing that should be taken away from all of this, and I am going to refer to Presidential Election Betting as a sort of, ‘Proposition Bet,’ simply because it happens so rarely, it is that people read too much into someone simply being the favorite on a given proposition and they beat accordingly. When I saw, back in July, that the betting markets were giving Donald Trump less than a 30% chance of winning the Election, I thought to myself, ‘The fundamentals basically never agree with that,’ and then I went ahead and conducted some additional research that seemed to not just support my conclusion, but also added evidence in support of my conclusion of which I was previously unaware.
There can often be a good deal of value in taking the underdog, not just in political betting, but in other forms of betting, as well. The question is not whether or not you are likely to win the bet, the more appropriate question is often: Even though I will probably lose this bet, is the value of what I am taking against the actual probability (albeit under 50%) of me winning this bet sufficient that I should choose the underdog?
The only way to answer that question is to start with a position on the matter, and then to investigate for yourself using good sources of proven solid information in order to reach your conclusions. That is not just true of political betting, it is true of every kind of betting that there is. When it comes to sports betting, for example, a bettor may use his information about the teams and delve into the statistics in order to accurately determine what may or may not be a good bet. For proposition betting, it is often a case of statistics.
Here’s a silly little example: Let’s say that you know something that someone else doesn’t know. For example, let’s say that you have noticed that the license plates in your state all start with a number rather than a letter, (I don’t know if that’s actually the case anywhere) if you are aware of that fact and your friend is not, then you might make a wager that the next car that passes the two of you will have a license plate that begins with a number. Your friend may unknowingly accept that bet thinking it is essentially a 50/50 proposition, in reality, you would have a huge advantage.
The key to proposition betting, and any other unique forms of betting in which one might have an advantage, is to be better studied and more knowledgeable about the subject than the person against whom you are betting. There are a surprising number of opportunities to do that, especially when those with less knowledge of the subject are all breaking one way. (often towards the favorite) If you can recognize these opportunities and are willing to do the proper research, it’s just another way for you to make money.
