2370: Prediction

Explain xkcd: It's 'cause you're dumb.
(Redirected from 2370)
Jump to: navigation, search
Prediction
You'd think it'd be easy to just bet money against these people, but you have to consider the probability of them paying up.
Title text: You'd think it'd be easy to just bet money against these people, but you have to consider the probability of them paying up.

Explanation[edit]

This comic is about misunderstanding probability. Sometimes people will incorrectly assume that if two events are possible, and one of them is more likely than the other to occur, then the first event WILL occur; or, that if one names two or more outcomes they are equally likely to occur when in fact they might have different probabilities.

Saying that one event is more likely to happen than another is not the same as saying that the first event is definitely going to happen. A statement like "event A has a 70% probability of happening" often misleads people into believing that event A is inevitable, while in fact 3 times out of 10 event B will happen instead of A.

Some don't like probability statements because they are not definite and therefore cannot be proven wrong. For example, if a probability statement says "event A has a 1% probability of happening" and event A actually happens, that does not prove the statement wrong, because the statement admits of the possibility of event A happening.

For example, FiveThirtyEight famously gave Trump a higher odds of winning the 2016 U.S. presidential election than most other models did just before the election, but still not more likely than his opponent. (28.6%). However, many readers at the time interpreted that as "Trump is definitely going to lose", and after he won that election, blasted FiveThirtyEight for getting its prediction "wrong". However, that interpretation is mistaken. 28.6% means Trump had a real chance at winning: if you could put election results in a hat and draw them at random, he would win one out of every three tries. For another example, in tabletop gaming terms, Trump's likelihood of winning was slightly lower than that of passing a flat check with a DC of 15 (6/20 or 30%).

The correct interpretation of a probability statement like "event A has a 70% probability to happen" is that in the long run, about 70% of events with this probability end up happening. If, for example, 99% of those events ended up happening, the 70% probabilities you gave those events may likely be wrong (you should've given probabilities closer to 99%), even though you "called" almost all events correctly (in the sense that 70% means the events are more likely to happen than not to happen, and almost all of them happened). Looking back at your predictions and seeing if the results are what you should expect is called calibration (example).

In the last panel, it is shown that Cueball anticipated this lack of understanding, so he plays pre-recorded audio of his prediction for the conversation.

The title text says that these people are gullible enough to the point that they would accept a disadvantageous bet. However, it also says that the probability that they might not actually go through with paying the bet if they lose brings into question whether to propose the bet is actually worth it. Randall has previously made allusions to betting on fallaciously claimed probabilities in comics such as 1132: Frequentists vs. Bayesians and 955: Neutrinos.

The comic doesn't rule out the possibility that event A and event B aren't directly related. For example, it is more likely to flip a coin and get a head than to roll a 6-sided die and get a 6. This is a fairly pointless observation in most cases, except perhaps if one is trying to explain the probability of an unfamiliar event by comparison with something very familiar.

At the time of writing, the 2020 United States presidential and congressional elections are less than a month away. This is a time when polls showing one or the other candidate leading are common, and may be misinterpreted to mean that the candidate is certain to win. Additionally, after the 2016 election saw Donald Trump, the trailing candidate in the polls, winning, many also interpreted this to mean that the polls were useless and/or wrong, or even go beyond this and take an adverse poll prediction as a perversely authoritative indication that the exact opposite result (which they would favour) is now a certainty. Cueball has previously shown an interest in U.S. election polling, for example in 500: Election.

In early October, famous statistician Nate Silver explained on his podcast "Model Talk" that, according to his model, Donald Trump had a 17% chance of winning reelection in 2020. That seems low, but it's a one in six chance, the odds of Russian roulette, the practice of shooting oneself in the head with a six-bullet barreled pistol with only one chamber loaded: it only has one chance in six to kill the person doing it. Would anyone in their right mind play Russian roulette? The answer he was implying was no. This illustrates how one chance in six is very real. While 17% seems low, it can absolutely happen.

Transcript[edit]

[White Hat and Cueball standing next to each other. Cueball has his palm out.]
Cueball: Event A is more likely than Event B.
[White Hat touches chin thoughtfully]
White Hat: So you're saying that Event A will happen.
Cueball: No, Event B could also happen.
[A frameless panel]
White Hat: So you're saying it's 50/50.
Cueball: No, it's definitely not 50/50.
[Cueball produces a phone]
White Hat: Sounds like you have no idea what will happen.
Cueball: And yet I knew exactly how this conversation would go. Here, listen:
[Cueball clicks a button on his phone]
*Click*
Phone: Then you'll say, "So it's 50/50"


comment.png add a comment! ⋅ comment.png add a topic (use sparingly)! ⋅ Icons-mini-action refresh blue.gif refresh comments!

Discussion

Is that a JoJo's reference?!1!! 172.68.142.213 23:18, 9 October 2020 (UTC)

Dunno who or what JoJo is (unless Jojo Siwa? But how would that be relevant?), but this is at least 70% likely to be a reference to the current election season in the USA and 538's (and others') predictions of Donald Trump's chance of winning the election in 2016 and 2020. In 2016, if I recall correctly, Trump had about a 30% chance to win (and thus Clinton had a 70% chance to win), and when 538's model launched earlier this year, the chances were basically the same (28-71 (with a 1% of an electoral college split because the USA's election system is phenomenally stupid)). Since then, Trump's chances of winning legitimately (538 does not attempt to model the chances and effects of election interference or votes not being counted) have slipped to about a 15% chance of winning which sounds bad, but will still happen in approximately 1 of every 7 tries, or about the number of Mondays in a week. Not great, but not impossible, either....)162.158.75.194 23:36, 9 October 2020 (UTC)
Yare yare, not knowing about Jojo's Bizzare Adventure. I can see where OP is coming from but I don't think Randall watches anime...
Now, let's talk about another misconception: lot of people intuitively think that an if event has chance of 1/7, it will almost surely happen at least once in seven tries. In reality, that chance is just 66%. -- Hkmaly (talk) 00:17, 10 October 2020 (UTC)
The 1/7 thing surprised me, which is good for me. Of course I checked the math. (6/7) ^ 7 = 34%, about. So it is not 50-50, it is 66-34.
It may not be critical or may be obvious but in the comic, it seems to be understood that Event A or Event B may happen but not both, which "of course" affects how probability works. The difference between flipping one coin or two coins.
If it is about the election, then if most people decide early to vote or not, and for whom, then the election isn't really random. However, the poll is random; they pick a few people out and ask their intention. The picking is unreliable even in a large sample (another probability surprise) and so is whether the people picked answer truthfully about their voting intention. Robert Carnegie [email protected] 141.101.98.155 11:42, 10 October 2020 (UTC)
Yup, this is another politics comic. It's very similar to 1131: Math, and also reflects similar frustrations as the more recent 2357, although from a different angle (2357 was about lack of respect for polls, while this one's about poor grasp of odds and probability in the context of election models). Pelosujamo (talk) 04:09, 10 October 2020 (UTC)

I don't know of Randall's got a series of White Hat comics sitting ready (the last one being 2368) but he didn't want it to look like a 'series' so padded with something else. If we've got another such dialogue before the end of next week, it may be a sign he's recently had a particularly bad conversation/message-session with someone and just wanted to vent a bit. And I wouldn't blame him, if that's so. 141.101.107.82 00:24, 10 October 2020 (UTC)

(PS, pre-post edit, but not for the want of trying: CAPTCHA wanted me to identify tractors. Two obvious tractors, no tractors on any of the other tiles (definitely) but one of them had a road-roller. Refused to accept the two tractors only, and I'm refusing to support the presumably incorrect Id of the roller, so come back to edit this in, do my own venting, and perhaps I'll get a better CAPTCHA when I retry in a moment... (Thanks to an Edit Conflict after I was finished fighting the Captcha I'm able to come back to tell you that the next one was Stairs, and I aced that one! But gotta suffer at least one more, yet...)) Also 141.101.107.82 00:24, 10 October 2020 (UTC)
According to the latest episode of QI XL (oh, hang on, [have a linky]) they suggest that being asked to do an actual Captcha for "I'm not a robot" means that you weren't exhibiting human-enough browsing activity before clicking that box. 141.101.98.154 20:41, 10 October 2020 (UTC)

Probabilities are hard to understand intuitively when you're actually talking about a one-time event. If you roll a d6 a whole bunch of times, you'll get each face about 1 out of 6 times. But it's not like we can hold the election 100 times, and then we can see if Biden wins around 52 of them to prove Nate Silver right or wrong. Also, elections aren't random processes like rolling dice -- there are human beings making conscious decisions how to vote, and we like to believe that we understand human motivations and can predict what people will do, at least in aggregate (fields like economics and marketing depend on this). Unfortunately, it's tough to make predictions, especially about the future (thank you, Yogi Berra). Barmar (talk) 05:44, 10 October 2020 (UTC)

Vote folks. Boatster (talk) 07:14, 10 October 2020 (UTC)

How do we know that Cueball doesn’t have a number of pre-recorded messages and he just chooses the one that suits the situation? 162.158.155.54 09:26, 10 October 2020 (UTC)

We don't! 02:42, 11 October 2020 (UTC)
So that's impossible then? :p 162.158.158.103 09:20, 11 October 2020 (UTC)

For the first time since I've been reading this site, the explanation has left me more baffled than the comic. Five hours previous to me making this comment, somebody edited the explanation to add a highly technical reference that I assume may be British English, because it sure ain't American. What does "passing a flat check with a DC of" refer to? What is the formula for a flat check? And is it any different than what we in America would call a rubber check, and is passing one (essentially committing a form of counterfeiting) illegal where you are from?Seebert (talk) 13:53, 11 October 2020 (UTC)

It's a role-playing or board game reference. I googled it, and found this: https://2e.aonprd.com/Rules.aspx?ID=333 Barmar (talk) 18:24, 11 October 2020 (UTC)
Also, in the UK, it’s “cheque” not “check” 141.101.98.52 11:15, 15 October 2020 (UTC)

Roll Charisma. DC 15. Ok. *rolls* NAT 20! --108.162.216.202 18:19, 11 October 2020 (UTC)

The first sentence of the last paragraph is in the past tense, and will be correct when the U.S. elections are over. The next sentence is in the present tense. I'm not sure which is better, but we should be consistent. BunsenH (talk) 18:31, 11 October 2020 (UTC)

I feel compelled to point out that if the conversation had not gone as he ‘predicted’ he never would have mentioned his ‘prediction’ at all. Responses to this comment will be exactly what I predict, you’ll see when I tell you what I predicted. 172.68.189.161 19:32, 11 October 2020 (UTC)

This reminds me of hearing about several years ago in Italy there were a bunch of little earthquakes, and they asked some scientists if that meant there were was a big earthquake coming, and they said probably not but they couldn't be certain, then a big earthquake happened and some people died, then they put the scientists on trial for manslaughter.--162.158.75.36 07:48, 12 October 2020 (UTC)

Interesting. Here I was thinking this was related to Trumps assertion that he doesn't think Scientists have any idea what will happen with the global climate situation, but no one has made any reference to it at all so far. 162.158.166.167 09:23, 12 October 2020 (UTC)

My immediate thought, as others have pointed out was that this was a commentary on the criticism of 538 (and other statistical models) after the 2016 election. For example, Donald Trump tweeted out a criticism of 538 yesterday talking about how wrong they were and Nate Silver responded. https://twitter.com/NateSilver538/status/1315296563212832768 108.162.216.228 12:25, 12 October 2020 (UTC)

People who have been declared stupid have no obligation to react sensibly. People feel no obligation to try to educate stupid people. But declaring oneself to be on the clever side feels so good. --162.158.92.194 11:14, 12 October 2020 (UTC)

The title text implies that you are discussing with a dishonest person ("you have to consider the probability of them paying up"). So I think this comic is about people who purposely distort facts and claim that the opponent are confused. A red herring. 141.101.105.244 11:22, 12 October 2020 (UTC)

Rather, it's just saying again more about how you need to factor in probability into things correctly, and in general an idiot like this isn't trustworthy. It's not about them being dishonest about whether they are misunderstanding probability though. It does raise the question though about what the terms of the bet are so that it would be advantageous for you, and how to consider that they are basing their predictions of what will happen on a simplification and misunderstanding of predictions that you give them yourself, and they'd thus be reluctant to bet with you when they think you just said which way it is going to happen, and thus any such bet you make with them afterwards, they'd either think the way you want to bet is guaranteed for you to win, or it wouldn't make sense as you would be trying to bet the way you said wouldn't happen, which would likely make them think you are being dishonest somewhere, unless they are even more crazy and stupid than shown here. At best you would bet on something you convinced them is 50/50 with you taking the more likely side of the bet.--162.158.74.245 11:10, 14 October 2020 (UTC)