Board 8 > Politics Containment Topic 268: Hope & Change

(+) Yeah! (+)

By the way, why don't people in this topic like Bloomberg? I know next to nothing about him and it's seeming more likely he will actually be relevant.
---

You guys just don't get that everything is a 50/50 chance because it can either happen or not

---

KamikazePotato posted...

One of the websites that gave Clinton 99% posted an article pre-election that bashed Nate Silver, saying he was inconsistent for doing things like 'not sticking to one prediction forever' and 'updating projections over time'. Nate got really upset at them over social media, then started making fun of them once Trump won. They proceeded to post an ENTIRE SALT-FILLED ARTICLE about how much of big dumb meanie Nate Silver was.

Was the only part of that entire part of history that made me smile.

And that site should be thinking about where they went wrong with their prediction. Not saying, well, Trump had a 1% chance! It was just this time!

---

If something happens once out of a hundred times, it cannot ever happen

---

TheRock1525 posted...

So when a predictive model said the Atlanta Falcons had a 99% chance of winning when they went up 28-3 and ultimately lost, does that mean the model was broken?

Yes, it failed to account for the fact that Shanahan is a moron.

But, no, that model is predicated on a ton of similar scenarios (though you could argue it could have been altered some with some information they couldn't have quantified in some way). There is no similar scenario to the presidency.

---

Panthera posted...

You guys just don't get that everything is a 50/50 chance because it can either happen or not

This was the premise of my absolute favorite Daily Show field piece where John Oliver interviews a guy who was suing over the LHC possibly creating a black hole and destroying the earth

He claimed there was a 50% chance of it happening using exactly that logic

---

pyresword posted...

By the way, why don't people in this topic like Bloomberg? I know next to nothing about him and it's seeming more likely he will actually be relevant.

He has a very troubled history on race: Stop and frisk, the audio coming out of him saying they should just "xerox descriptions of 16-to-25 year old minorities to police," the audio of him coming out and saying the cause of the 2008 financial crash was because we ended the practice of redlining.

He also ran as a Republican as mayor in New York, then when he hit term limits he strong-armed his way into getting another term. And there's also the extreme distaste of the fact that he's literally buying his way into this election on a level that makes even Steyer look tame.

There's more, I'm sure and others can cover it if they like (I recall there also being quotes of him making gross objectifying comments towards women) but that's all just off the top of my head.

I actually hope he makes it to a debate stage (and wonder if he actually wants to avoid them) so the other candidates can nail him on all this. They've been doing it in interviews and social media but man it would be great to see Warren decide to go out exposing him in a national debate.

---

They had to rig the RNG for Fire Emblem to make the most likely outcome more likely because when a 10% chance happens roughly one time in ten, people think it's rigged.

---

Manipulating RNG speedruns should show you guys that some %rates of stuff in RPGs actually were bullshit.

How many of you were the kid who reset a save 500 times just to have the same outcome happen each time you reloaded! Talk about some bullshit RNG system. Lol

---

pyresword posted...

By the way, why don't people in this topic like Bloomberg? I know next to nothing about him and it's seeming more likely he will actually be relevant.

As far as his stated policy platform goes I think he's mostly fine, but in addition to stuff like stop and frisk he sank like a rock for me when I learned that he has an extensive history of being sued for sexual harassment and discrimination (I believe there have been 40 such lawsuits). I'd still vote for him over Trump because Trump is strictly worse on the axis of every Dem candidate's worse flaws including this one, but I can't support someone like that in a primary.

---

Corrik7 posted...

Manipulating RNG speedruns should show you guys that some %rates of stuff in RPGs actually were bullshit.

How many of you were the kid who reset a save 500 times just to have the same outcome happen each time you reloaded! Talk about some bullshit RNG system. Lol

In FE (not sure if this is true in all of the games, but it definitely was in the GBA games) everything is deterministic, so if you reload the exact same thing will happen.

---

That's not bullshit RNG. That's just usually the game having determined its rolls beforehand and stuffing them in a table or something. If it determines the seed it uses to roll RNs first and then goes through them as the rolls come up.

RNG manipulation then in this case is usually predicated around "using up" bad rolls on meaningless actions (also something you can do in FE7 to some extent because there's a trick with how it determines pathing that relies on the RNG so depending on how it resolves you can know if given rolls are low (1-50) or high (51-100) and then reset and act accordingly.

It's a function of truly random number generation being a far more difficult computing problem then you think so they rely on other methods to simulate it which depending on implementation can be gamed.

---

I've seen "procedural != random", but never thought I'd need to say "seeded != unrandom"

---

I sort of see corriks point here. Nate Silvers stuff isnt quite like a fire emblem chance to hit because FEs attacks are repeatable so you can see over time that there actually was that 1% chance or whatever, but the election just happens once so its impossible to prove that running the election 100 times would result in 99 trump losses or whatever the actual prediction was

so silvers predictions look like making a bet you cant technically lose no matter the outcome even if the outcome is the one he said was incredibly unlikely. If the things he says are less likely to happen consistently end up happening it casts some doubt on the process by which hes making these predictions (I have no idea how often this is the case for him though)
---

LordoftheMorons posted...

In FE (not sure if this is true in all of the games, but it definitely was in the GBA games) everything is deterministic, so if you reload the exact same thing will happen.

It's true in Three Houses as well. Level-ups are rolled at mission start so if your main character's next level up is a bad level up, rewinding time won't help you.

It's also true in a general sense. If you rewind time then play out the exact set of actions the outcome will not change. However as I noted, if you alter a few actions so that the various rolls are being used differently you can end up on a different place on the table when you get back to the spot you revert from and possibly get a new result. On the same front, if you understand what rolls are occurring and can keep track of how many are being used/how many you rewound you can tell where that particular roll was and plan around it. Though again, level-ups are in their own area and you can't game them this way (but you can in FE7!)

---

NFUN posted...

I've seen "procedural != random", but never thought I'd need to say "seeded != unrandom"

definitely haven't hanged around enough fire emblem discussions then to hear the complaints about pseudorandom number generators (though almost never in those actual terms because if you're that deep into it then you understand)

---

MoogleKupo141 posted...

I sort of see corriks point here. Nate Silvers stuff isnt quite like a fire emblem chance to hit because FEs attacks are repeatable so you can see over time that there actually was that 1% chance or whatever, but the election just happens once so its impossible to prove that running the election 100 times would result in 99 trump losses or whatever the actual prediction was

so silvers predictions look like making a bet you cant technically lose no matter the outcome even if the outcome is the one he said was incredibly unlikely. If the things he says are less likely to happen consistently end up happening it casts some doubt on the process by which hes making these predictions (I have no idea how often this is the case for him though)

Pretty much. Is it, hey you are just wrong and this is how it happened? Or would this really have happened 1 time in a 100?

And, I assume he means 1 time in a 100 with this data, but again, the variables are all relatively different from 4 years to 4 years in an election.

So basically, your highest probable outcomes are what you are saying your result is to be. If it is some outlier, then it isn't well it would have happened the 99 other times. It just didn't happen. There is no way to prove the other end, yet it is handwaved as a once in a hundred happening.

---

xp1337 posted...

It's true in Three Houses as well. Level-ups are rolled at mission start so if your main character's next level up is a bad level up, rewinding time won't help you.

It's also true in a general sense. If you rewind time then play out the exact set of actions the outcome will not change. However as I noted, if you alter a few actions so that the various rolls are being used differently you can end up on a different place on the table when you get back to the spot you revert from and possibly get a new result. On the same front, if you understand what rolls are occuring and can keep rack of how many are being used/how many you rewinded you can tell where that particular roll was and plan around it. Though again, level-ups are in their own area and you can't game them this way (but you can in FE7!)

That really strikes me as a bullshit way to do RNG lol. Imo

---

pyresword posted...

By the way, why don't people in this topic like Bloomberg? I know next to nothing about him and it's seeming more likely he will actually be relevant.

because, on top of his record that's already been brought up, fuck billionaires trying to buy their way into the race and flood the airwaves with ads while skipping out on things like debate so that their narrative doesn't actually have to get challenged.

MoogleKupo141 posted...

I sort of see corriks point here. Nate Silvers stuff isnt quite like a fire emblem chance to hit because FEs attacks are repeatable so you can see over time that there actually was that 1% chance or whatever, but the election just happens once so its impossible to prove that running the election 100 times would result in 99 trump losses or whatever the actual prediction was

so silvers predictions look like making a bet you cant technically lose no matter the outcome even if the outcome is the one he said was incredibly unlikely. If the things he says are less likely to happen consistently end up happening it casts some doubt on the process by which hes making these predictions (I have no idea how often this is the case for him though)

I mean yeah there are fundamental limits on being able to tell if you forecast a one time event correctly. I don't think that's a reason to conclude that it's fundamentally unknowable and that we can't obtain some information out of polls.

---

Reg posted...

because, on top of his record that's already been brought up, fuck billionaires trying to buy their way into the race and flood the airwaves with ads while skipping out on things like debate so that their narrative doesn't actually have to get challenged.

I know the common refrain on Twitter is "DNC is changing the rules to get Bloomberg in the debate! rigged!"

...but I feel like the DNC needs to get him in the debates. He is getting nothing but positive coverage through his ads and never has to defend his record or positions in front of the media. He's dodging the debates because he doesn't need donors due to how rich he is, but his polling puts him way above a lot of people that were and are making the debates.

I think it would be "better" to force him to debate at this point. I don't know for sure, but avoiding them clearly isn't hurting him in any way.

---

xp1337 posted...

It's true in Three Houses as well. Level-ups are rolled at mission start so if your main character's next level up is a bad level up, rewinding time won't help you.

It's also true in a general sense. If you rewind time then play out the exact set of actions the outcome will not change. However as I noted, if you alter a few actions so that the various rolls are being used differently you can end up on a different place on the table when you get back to the spot you revert from and possibly get a new result. On the same front, if you understand what rolls are occurring and can keep track of how many are being used/how many you rewound you can tell where that particular roll was and plan around it. Though again, level-ups are in their own area and you can't game them this way (but you can in FE7!)

Yeah I have fond memories of using that fact to farm infinite Boots in Sacred Stones and having 15 movement units wreck everything...!

Corrik7 posted...

That really strikes me as a bullshit way to do RNG lol. Imo

Well, in a practical sense you can't realistically get real random numbers for a game (you could use an actual quantum process like getting a radioactive source and measuring decays or something I guess!). Typically if you want to make things not be deterministic from run to run you'd seed the pseudorandom number generator differently based on, e.g., the system clock.

---

Suprak the Stud posted...

I know the common refrain on Twitter is "DNC is changing the rules to get Bloomberg in the debate! rigged!"

...but I feel like the DNC needs to get him in the debates. He is getting nothing but positive coverage through his ads and never has to defend his record or positions in front of the media. He's dodging the debates because he doesn't need donors due to how rich he is, but his polling puts him way above a lot of people that were and are making the debates.

I think it would be "better" to force him to debate at this point. I don't know for sure, but avoiding them clearly isn't hurting him in any way.

No, I'm with you 100% on this. Warren recently said something about thinking he shouldn't be able to buy his way onto the debate stage and in a vacuum that's true but I think she's wrong in this instance. Bloomberg is being aided by being able to go unchallenged. I want him in a debate so the other candidates can expose the awful parts of his record that aren't breaking through because he's flooding the airwaves with his own ads.

---

Suprak the Stud posted...

I know the common refrain on Twitter is "DNC is changing the rules to get Bloomberg in the debate! rigged!"

...but I feel like the DNC needs to get him in the debates. He is getting nothing but positive coverage through his ads and never has to defend his record or positions in front of the media. He's dodging the debates because he doesn't need donors due to how rich he is, but his polling puts him way above a lot of people that were and are making the debates.

I think it would be "better" to force him to debate at this point. I don't know for sure, but avoiding them clearly isn't hurting him in any way.

I agree with this. Normally being in the debates is a strict positive, but that assumes that your campaign would otherwise be suffocating for attention. Bloomberg's unlimited money lets him bypass that while avoiding any scrutiny.

---

LordoftheMorons posted...

Well, in a practical sense you can't realistically get real random numbers for a game (you could use an actual quantum process like getting a radioactive source and measuring decays or something I guess!). Typically if you want to make things not be deterministic from run to run you'd seed the pseudorandom number generator differently based on, e.g., the system clock.

There isn't a way to just internally roll a dice every time it needs a random result?

---

Suprak the Stud posted...

I know the common refrain on Twitter is "DNC is changing the rules to get Bloomberg in the debate! rigged!"

...but I feel like the DNC needs to get him in the debates. He is getting nothing but positive coverage through his ads and never has to defend his record or positions in front of the media. He's dodging the debates because he doesn't need donors due to how rich he is, but his polling puts him way above a lot of people that were and are making the debates.

I think it would be "better" to force him to debate at this point. I don't know for sure, but avoiding them clearly isn't hurting him in any way.

This is absolutely and objectively correct.

The rigged notion isn't wrong either, and its a very textbook example of him just buying influence in all the ways that are awful for this country. But you're still correct.

I'd rather have a person be able to buy themselves into the debates as opposed to buy themselves into an election without ever having to appear in person.

---

Corrik7 posted...

There isn't a way to just internally roll a dice every time it needs a random result?

AFAIK there is no actual way for computers to generate true randomness. We just have best approximations
---

I also don't legit get how a model that predicted a 1% chance for Donald Trump means that a completely different model is invalid. I cite 538 because 538 does really good work and is highly accurate.

---

Jakyl25 posted...

AFAIK there is no actual way for computers to generate true randomness. We just have best approximations

Wow I didn't know that. That's interesting.

---

Also IOWA is old news by now, but still feel obligated to post this:

https://www.nytimes.com/interactive/2020/02/14/us/politics/iowa-caucus-results-mistakes.html

Some exact info on how many mistakes were made (a.k.a. lots), but this is my favorite:

When the reporting app failed on caucus night, the state party was flooded with calls from volunteers trying to submit their results. The people taking the calls had to copy down the results and enter the data by hand, leaving plenty of room for error.

In this example, in Indianolas Second Precinct in Warren County, the party released results showing that Senator Elizabeth Warren of Massachusetts and Mr. Sanders had zero votes in the first round but picked up several dozen each in the final round, a highly unlikely scenario. Meanwhile, the billionaire Tom Steyer and former Gov. Deval Patrick of Massachusetts met the viability threshold in the first round but lost all of their votes in the final round.

The precincts worksheet shows vote counts that make more sense. It appears that whoever entered the data into the partys database switched the votes for Ms. Warren and Mr. Sanders with those for Mr. Steyer and Mr. Patrick.

---

TheRock1525 posted...

I also don't legit get how a model that predicted a 1% chance for Donald Trump means that a completely different model is invalid. I cite 538 because 538 does really good work and is highly accurate.

TheRock1525 posted...

I also don't legit get how a model that predicted a 1% chance for Donald Trump means that a completely different model is invalid. I cite 538 because 538 does really good work and is highly accurate.

I am not saying it is invalid. I am saying the model is pointless if no matter the result you can just be like, yep, that would have happened 1 time in 100! Instead of realizing something was wrong that happened. I agree 538 is probably the most accurate model that exists right now, but I don't agree that all the moving parts are accurate at times either. But, enough is that he is relatively accurate.

---

Corrik7 posted...

There isn't a way to just internally roll a dice every time it needs a random result?

Assuming you're not thinking of a literal physical die inside your computer, no. The program's code is fixed, so the only way you can get different outcomes is to give it different inputs. Pseudorandom number generators are actually deterministic, but the sequences of numbers they give "look" random in the sense that (if the PRNG is perfect) if I get some numbers from the PRNG, I can't predict future numbers any better than I could for a truly random number generator.

---

I legitimately have no idea what point you're trying to make.

Can anyone else explain it to me?

---

Corrik7 posted...

Wow I didn't know that. That's interesting.

I told you true randomness is a much harder computing problem then you would think!

---

Corrik7 posted...

Wow I didn't know that. That's interesting.

I mean its kind of the same for us. When we roll dice it isnt a TRULY RANDOM result. Based on the starting position of the dice, the force with which they are thrown, and the dimensions and texture of the throwing surface, you could in theory predict the result.

Its just impossible for us to calculate all that live and in person
---

TheRock1525 posted...

I legitimately have no idea what point you're trying to make.

Can anyone else explain it to me?

He can never be wrong in that scenario unless he has a truly 0% option that happens.

If he feels a insignificant result happening doesn't mean the model was flawed for that presidential election but just that the 1% finally occured, then he can never be wrong. Which makes it pointless.

---

Jakyl25 posted...

I mean its kind of the same for us. When we roll dice it isnt a TRULY RANDOM result. Based on the starting position of the dice, the force with which they are thrown, and the dimensions and texture of the throwing surface, you could in theory predict the result.

Its just impossible for us to calculate all that live and in person

I suppose that is true.

---

KamikazePotato posted...

Some exact info on how many mistakes were made (a.k.a. lots), but this is my favorite:

IIRC, when we first heard that someone in this topic speculated it was something akin to a spreadsheet error by not having Deval Patrick in since it could be explained by shifting the votes off alphabetically by one.

but yes that kind of mistake was silly and i remember when the data people called it in real time lol when that drop of votes first came out

---

Could God create a random number generator so random even he couldnt predict it?

Makes you think
---

Jakyl25 posted...

Could God create a random number generator so random even he couldnt predict it?

Makes you think

No because he is omniscient.

But then again he is omnipotent so he should have the power to do so.

Okay, fuck that question.


---

Corrik7 posted...

He can never be wrong in that scenario unless he has a truly 0% option that happens.

Duh?

---

Corrik7 posted...

He can never be wrong in that scenario unless he has a truly 0% option that happens.

If he feels a insignificant result happening doesn't mean the model was flawed for that presidential election but just that the 1% finally occured, then he can never be wrong. Which makes it pointless.

Maybe this is the problem. Statistics isnt really about being "right" or "wrong" like you're using it. It's about making reasonable predictions. Statistics basically says you can't know for sure but here is my best guess if you did this thing hundreds of times.

---

HeroDelTiempo17 posted...

Maybe this is the problem. Statistics isnt really about being "right" or "wrong" like you're using it. It's about making reasonable predictions. Statistics basically says you can't know for sure but here is my best guess if you did this thing hundreds of times.

Then is Silver actually more right than the 99% Clinton prediction? What's the point at that point. Maybe Trump winning was really once in a 10p and they were right, and Silver is wrong that it was more likely.

---

LordoftheMorons posted...

No, if he claims 100 races are 70/30, he should be judged on whether the 70 candidates won roughly 70 of those races. If they won 95 or 50 of them, his model was likely wrong.

I am not sure about this. For argument's sake I am going to assume that we are judging the model on its predictive power. If 95% of the 70% candidates won, it seems the model could actually be extremely (like, 95%) predictive. I could create another model that takes the model that says 70% and averages it with 50%. My new model would say that X candidate will win 60% of the time. And it would still be 95% accurate if we assume that it's really predicting that whoever is over 50% in the model will win 100% of the time.

Conversely, suppose we had a model that was wrong 100% of the time. This model is actually 100% accurate at predicting things. Because we can construct another model that is right 100% of the time by just predicting the opposite.

I'm aware that 538 tries to mean what they say. But I'm not sure they can be criticized for getting 100% of the states correct in 2012 for instance when their model was not at that level of confidence. I guess what I'm saying is that this shows that the model acknowledges its own lack of precision, but the actual results aligning with 538's call 100% of the time doesn't mean the model is worse than if it had only "called" 90% of the states correctly. That is sort of a fictitious accuracy, rather like guessing a number between 1 and 100 and getting it exactly right - it means you were lucky.

---

Corrik7 posted...

Then is Silver actually more right than the 99% Clinton prediction? What's the point at that point. Maybe Trump winning was really once in a 10p and they were right, and Silver is wrong that it was more likely.

Again, we wouldn't know unless we ran 2016 hundreds of more times exactly the same way, but given that 538 models DO have situations we can verify like the 2018 example, you can have a little confidence in their methods.

---

Also the example posted was that the 99% Clinton guys said it was bad to account for new data so based on that alone you can guarantee their model has flawed assumptions!

---

To flesh this out a little more, here's a thought experiment. It's late in the evening of election day 2012. 538 has called the first 49 states correctly. The last state to have be called, let's call it Florida for argument, is still outstanding. 538 predicted earlier that Obama has a 60% chance of winning Florida. When Florida is called for Obama, does this increase or decrease your confidence in 538's model?

Surely your answer cannot be that it depends on how 538 did in the other states, and if it did really well in the other states, then its favored candidate in Florida winning means the model is less accurate.

---

HeroDelTiempo17 posted...

Again, we wouldn't know unless we ran 2016 hundreds of more times exactly the same way

Then at that point it is pointless! We can never know who was more accurate! Ever. If Silver has 99% Trump wins 1% Clinton wins and someone has 99% Clinton wins and 1% Trump wins. For all we know in this we will NEVER KNOW SCENARIO is that they either could have been right. Maybe silver is right is happens 99%% of the time. Maybe the other model is right it happens once in a hundred times! Lol.

You can't be more or less accurate in this scenario so it becomes pointless.

That is why I am saying the point is for the actual result to be your likeliest result, and if it wasn't your likeliest or at least another highly likely option then to wonder where your model went wrong.

---

People are misunderstanding this. The Clinton 99% people were obviously wrong. They could have had a great model and a correct prediction based on the information they had, but then their information was lacking.

It's like the question, what are the odds of the next card in a pack of cards being a K? You might say 1/13. But what if the previous cards were A 2 3 4 5 6 7 8 9 10 J Q? Now it's overwhelmingly likely you have an unshuffled pack - and the probability of the next card being K is nearly 100%.

---