Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
Posted on 19-06-25, 13:38 in Mozilla, *sigh* (revision 1)
Stirrer of Shit
Post: #441 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Posted by Screwtape
This is a hilariously bad assumption to make for all kinds of reasons, some of which you point out yourself.

It's definitely a common assumption, and it probably even works out in > 50% of situations, but it's also the kind of assumption that causes heartache and strife for the people it's applied to, and confusion and frustration for the people who apply it without thinking too hard.

As a general life rule, I try to avoid making generalisations about people, or at the very least keep in mind that generalisations are often wrong.

But about the group of workers?

This is just me trying to argue for an anecdotal observation - the web developers I've met have indeed (in general) been less competent and passionate. Of course this doesn't go for all of them; if you'd model it mathematically I reckon you'd get two partially overlapping distributions with different averages, so that you could find a web developer who is better than a given programmer, but the web developers would still on average be worse than the programmers.

If I've learned anything from my time on the Internet, there is nothing so obscure or commonplace that it doesn't have a community of enthusiasts. [...]

In comparison, web-development is one of the most rewarding technical things you can do - a little time investment, and you can make things 80% of the quality of most professional works and immediately show them off to all your non-technical friends all over the world. It's like like learning BASIC for the Sinclair Spectrum in the UK in 1982, but even better.

Do people do it for the results, though?
Anecdotally, most programmers to some degree enjoy programming. They however surprisingly often don't care a whole lot about the practical utility/attention they can extract from their creations. Whereas, for web developers it would be the opposite if it is as you claim, no?

What actual shortcomings does MS-DOS 1.0 have? Is there anything intrinsically wrong with it?

(To skip to the end of this Socratic dialogue, not everybody has the same needs, but network effects mean people want their tools to be compatible with the tools other people use. Every tool must compromise between supporting a specific use-case, and being compatible with other tools. If you can afford to wall yourself off from the rest of the world and only use Windows XP, or only MS-DOS 1.0, or only flathead screwdrivers, then that's cool, but don't be surprised or angry that other people can't afford to do so, or choose to spend their resources in other ways.)

MS-DOS lacks good graphics APIs, networking, and filesystems, for instance. This seems pretty intrinsic to me. I don't know what such flaws XP has. The security stuff is vastly overrated.

But if it's a matter of network effects, then implying the downsides are intrinsic is not really accurate. My point was that sticking with XP would have been a good idea if everyone did it. (indeed, if I wish to do so unilaterally I can just download the ISO and go, no need to ask anyone else)

Another advantage: With XP still alive, you'd have SOME underpowered netbooks (not just third world, but also kids and cheapskates) to which you'd have to adapt, whereas with Windows Vista developers could assume the computer is at least fast enough to run Vista, giving slower applications.

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-25, 14:35 in Something about cheese!
Stirrer of Shit
Post: #442 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Posted by wertigon
Not true. Please refrain from spreading this lie further.

Their method makes no sense at all. Why didn't they just separate out the male and female distributions, if they had the raw data? The fit is quite bad. If I were feeling conspiratorial, I would say they were deliberately obfuscating a politically sensitive issue in exchange for headlines and funding.

Also, why only look at the top 100 players?

If you compare the share of women in the 80th percentile to the share of women in the 60th percentile, for instance, you find that the latter is lower, and you get more women the further down you go.

I don't know how to get ahold of the German chess database, but FIDE is public. And there you can verify this yourself, without having to rely on kook scientists.

The easiest way however is probably to go at it this way: if there are 10x more male players than women, then–correct me if I'm wrong–you can fit ten men inside one women's rank to get the same percentile. From that it follows it follows that a woman with female rank k should have the same Elo as a man with male rank 10k. In other words, the relationships between the ranks should be linear, with the slope of the regression being the same as the inverse share of women.

Whereas, if the discrepancy was a result of differing abilities, then the slope would be sharper than 10, and a linear regression wouldn't be a good fit.

Now, which one is it? https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2894898/

Of course, this will never be reported on because of course women are better at everything. But the reporting doesn't change the matter of things.
... Because women are NOT drilled to the same extent as men, they do not have to meet the same standards, hence they don't. And also statistics e.g. Lower size of women in general are interested in Soccer => lower number of really good players => lower interest and worse games => lower interest in Soccer => ...

Well, these are two different explanations.
One, women play less soccer. But even if it's just 3% as many, that should mean the USWNT (~300 million) should be able to beat the soccer team of any country with ~10 million, assuming equal participation rates.

This doesn't hold true. In the example of the female soccer teams vs. youth teams, the pool for the WNT would be (female population * share of women who play soccer), while the pool for the youth team would be (city population * share of youth in city * share of youth who play soccer).

Dallas' metropolitan area has 7.2 million people. In Dallas County, you've 88k 10-14 year olds, so around 22k, of which half are male, so 11k, out of a population of 2.3 million. That'd give a pool of 68k 14-year old boys for the whole of Dallas.

USWNT players are 20-30 years old. There are just north of 20 million women that age in the whole US.

That gives the USWNT about 325x larger base player pool. Now how many of the youth in Dallas play soccer? How many American women play soccer?

Say 1% of USA women do it. Then ~340% of Dallas boys would have to play soccer. As you might imagine, this is not possible.

The second one. Do you think 14-year olds are drilled harder than the national team? That would be absurd - for starters, 14-year olds do not typically play soccer professionally.

People in general will only do as much as is expected of them and not an inch more. If you expect more of them... They will magically give more. Until they reach their threshold, but that's usually very far above what is expected of them.

Sure, but this is the national team we're talking about. Surely, at least one country out of the many who has them would have figured out that by now? The USSR had infamously brutal training for its athletes, yet their women's soccer team was positively mediocre.

That is also what is wrong with the current school system. If you tell a kid (s)he's dumb enough times he/she will eventually believe it, and thus no expectations are put on, well, any kid to achieve. This is what feeds the "dumb as bricks" students we see exiting the classroom of today, most are stuck in a negative feedback loop that tells them they are dumb as bricks which makes them dumb as bricks.

Not cool, man!

Correlation doesn't imply causation. It makes far more sense that dumb kids would get told that because they are.

I agree with you that higher pressure ought to be put on the students, but this would exacerbate inequality and not decrease it. I disagree with the part where we should spend money in inverse proportion to aptitude - if you did that, then you'd get everyone performing close to the average, meaning very few prominent scientists and such, whereas if you spent money in proportion to aptitude, you'd get the middle people as they were before, more brilliant scientists, and slightly worse trained retards. Provided the latter is close to zero-cost (say $1000 a month, that's $12k a year) and a brilliant scientist an enormous gain for society (surely more than $12k a year), it stands to reason inequalities ought to be exacerbated rather than alleviated.

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-26, 13:22 in I have yet to have never seen it all.
Stirrer of Shit
Post: #443 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Man, FUCK the Washington Post
We noticed you’re browsing in private mode.
Private browsing is permitted exclusively for our subscribers. Turn off private browsing to keep reading this story, or subscribe to use this feature, plus get unlimited digital access.


It's not even ad blocking, it's straight up private mode. Aren't journalists supposed to at least nominally care about privacy? I'm using perfectly bog-standard Tor Browser with JavaScript on, the same as their dissidents in China or whoever it is they pretend to be deeply concerned about for the moment.

It is innovative though, I'll give them that - "cough up or we'll give your data over to the Google". It sure is something to sell ads to someone who's already paying for your disgrace of a newspaper. Then again, the people who pay to read newspapers online (that's like three layers of WTF at the very least) don't generally tend to be very bright and would probably pay $90 a year because they thought the EU law forces them to do it.

The blocking was also slightly better than usual. It wasn't enough to delete the element, you actually had to open it up in view source and to a CTRL-F for the article content.

Here is the (strikingly dishonest) "article" in its full "glory":


On a side note: has UI gone too far?


(note to self: if you ever make a phishing website, just have a really big HTTPS cert that blocks the address bar)

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-26, 13:41 in Something about cheese!
Stirrer of Shit
Post: #444 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Posted by wertigon
Now you are just plain and simple deliberately misreading statistics. I see you never took a proper statistics course - or if you did, you must've never passed it. :)

If you would have bothered to find and read the original paper, you would have found this excerpt in the discussion (emphasis mine):

This argument sounds reasonable but it rests on a controversial assumption ... which lacks conclusive evidence ... A recent study of 647 young chess players, matched for initial skill, age and initial activity found that drop-out rates for boys and girls were similar (Chabris & Glickman 2006).


Cool it with the personal attacks, will you?

I did read the paper. Since we are trying to ascertain whether there indeed are intellectual differences, it doesn't form a cohesive argument to claim that since there aren't intellectual differences, there can't be differences in chess aptitude, and thus there aren't intellectual differences. This is just plain circular reasoning.

Don't see the the relevance of the last bolded sentence. I wouldn't think girls drop out more than boys.

I read appendix A, it seems like a needlessly convoluted method. Surely, if you have 100 numbers, no matter the distribution, and pick one at random, you will have a 1% chance of picking the highest number, assuming there is such a number. And from this it should follow that the highest number forms the 100th percentile. Or more formally, from a sample of n items from a distribution, the sample's highest value's expected position (e.g. cumulative likelihood of obtaining the value) within the parent distribution is 1-1/n. And if there are 10x as many male players as women players, then provided they are drawn from the same parent distribution, that should mean that nmale = 10nfemale.

Then that would give the highest male player's expected relative position as 1-1/nmale, and the highest female player's expected relative position as 1-1/nfemale. And likewise, the second highest ranked player's as 1-2*(1/n) = 1-2/n, up to the nth highest ranked player's (e.g. the last ranked player's) as 1-n/n = 1-1 = 0.

Perhaps more intuitively: If you have 100'000 people, half men and half women, and 10k of the men go on to play chess and get ranked and 1k of the women do so, then if you'd randomly sample 1% of the men and 10% of the women, you'd end up with 1k men and 1k women. If the gender-wise distributions of ability are equal, then those two groups ought to be indistinguishable short of minor sampling noise, no?

If you have taken such proper statistical courses, then maybe you could tell me what is wrong with such methods and what the need would be to attack the problem from such a convoluted manner as they do, when they have far simpler methods (e.g. taking the mean of both distributions, applying a X2-test, and calling it a day) at their disposal?

The study is legit and has been successfully defended a number of times, and the formulas used to check out the conclusions hold up (check appendix A). At this point, if you are going to argue with science then go right ahead, but I will laugh at you. :)

I don't know what these successful defenses are. I can't find any rebuttals to https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2894898/, which seems to point out several rather large flaws in the study's methodology:
What these authors did not mention, however, is that their model predicts a rating of 3031 for the best male German player and a rating above 2700 for the 16 best male German players. Currently, there are only 33 players in the world with a rating above 2700 and there is no German belonging to this elite group. The highest rating ever achieved by a human player is 2851, which is significantly lower than the expected rating of 3031 predicted for the best German player according to the model of Bilalić et al. (2009). Therefore, this model seems inadequate to describe the upper tail of the distribution of ratings of German chess players.

For what it's worth, that also seems to use a needlessly convoluted methodology, although one which ends up making the same predictions (e.g. linear model with slope equal to representation difference should fit) as mine.

As for women soccer, well, there are a ton of improvements happening already and as time move on, things are bound to get better. :)

Do you have any sources for this?

Assuming no or at least fewer improvements happen in normal soccer, that should imply female win rate in mixed matches approaches 50% as time goes on. I can't actually find any statistics on mixed matches, since they're not held often. But based on other sports, we can guess that if the win rate were anything close to 50%, so that it would be possible to engineer such a match that appears fair at first glance and that gives the girls a good chance at winning, then such a match would indeed be engineered in order to "prove" that women's soccer is in fact a legitimate sport.

Since no such matches have been engineered and the only mixed matches we hear about are national teams getting curb-stomped by obscure regional teams' youth divisions with players missing, we can deduce that this is not the case, and that any improvements would have been utterly insignificant and presumably will continue to be so. Or, sure, you could claim that things will sharply get better in the future, but as they say. extraordinary claims require extraordinary evidence.

Further circumstantial evidence for this can be found in the extremely contrived reasons given for why the WNT should not play the national team. Since they can't be taken at face value but the view itself appears to be strongly held, you can draw the conclusion that there's some other reason for it that they don't want to share. And the only such reason I can come up with is that such a match would permanently shatter any delusions of women being even halfway decent at soccer, which would be a politically uncomfortable position to be in.

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-27, 15:24 in Ubuntu: x86_32 is dead because WE SAY SO!
Stirrer of Shit
Post: #445 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Happy that you're holding back the ONE Linux distro and really the ONLY distro that has and can bring Windows and Mac refugees into the Linux fold? You Geeks and Phreaks are Linux's BEST and WORST enemy!

EVERYONE has gone 64 bit. IDE'S AND COMPILERS are dropping 32 bit support left and right. BUT OMG WINE AND MY 32 BIT VERSION OF QUAKE 2 IN UBUNTU?? TO ARMS...TO ARMS!! HUZZAH !!

Or perhaps he's wondering how someone would bring into the fold a Windows user, by breaking the compatibility of his apps?

Jokes aside, I am curious what compiler doesn't support 32-bit. GCC does, Clang does, and I would think MSVC and ICC does, but even if they didn't they're utterly irrelevant and a far greater edge case than i386 ever will be.

Containers are pretty cool though. AppImages are almost exactly what I'd want in an application. No dependencies, just download and run. Could be made better, but so could everything.

You might complain that they're bloated, but remember that every instance of productive bloat reduces the space available for non-productive bloat, and thus helps improve software quality.

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-28, 21:27 in Something about cheese!
Stirrer of Shit
Post: #446 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Posted by wertigon
Actually, your reasoning is not correct, and the same can be said for Knapp.

The criticism of Knapp is threefold. The first is that the maximum theoretical player does not have the expected score of the real world player. This may seem relevant, but since the error is *uniform*, it does not matter for the intended purpose, which is to show that simple statistics make up the absolute majority of the explanation.

The error isn't uniform at all:


Furthermore, it doesn't look like a "random noise" kind of error, but rather a "horribly broken model" kind of error. So it's definitely a relevant criticism.
The second problem is that he questions the assumption that given a large enough dataset, each individual sample will fall within a bell curve. I do not see how this assumption is even questionable, any and everyone who has ever taken a statistics course knows this is the truth. The keyword being large enough dataset.

I believe you're thinking of the central limit theorem. It makes no predictions with regard to the number of samples, but rather the number of variables.

For Elo ratings, it's a reasonable criticism.
Posted by https://en.wikipedia.org/wiki/Elo_rating_system#Most_accurate_distribution_model
The first mathematical concern addressed by the USCF was the use of the normal distribution. They found that this did not accurately represent the actual results achieved, particularly by the lower rated players. Instead they switched to a logistic distribution model, which the USCF found provided a better fit for the actual results achieved.[citation needed] FIDE also uses an approximation to the logistic distribution.[17]


Anyway, by my understanding you'd just end up forcing it to a normal distribution, and the inaccuracies would just end up as a lowered predictive value (e.g. noise) which shouldn't have any systematic effects on the ranking. Maybe it would break the ridiculously complex formulae if he did get it wrong though.

The third and final problem is the same trap you fall in; This is not a linear model. It is an exponential model. That is what you need all those complex calculations for. Again, you are clearly not a statistician, else you would've understood this. It is taught in the first college-grade statistics course. If you want to do more reading look up the official method in this book:

Order Statistics, Herbert A. David, Haikady N. Nagaraja, ISBN: 0471654019

It is considered customary to tell the reader at which page the citation is to be found. You would not be so kind so as to bestow upon me your advanced statistical education, would you?

I don't know what exactly is an exponential model here. I don't think the 2009 study's is, according to figure 3 they model the expected performance of the best performers with what looks decidedly logarithmic.

A linear model does seem perfectly reasonable for the correlation of the internal rankings in a smaller sample drawn from a larger sample and the rankings of the items in the smaller sample computed taking into account the entire sample.

You can model it yourself in Excel. Paste in the following in an empty sheet on A1 through F1:

=RAND();=RAND();=NORM.INV(A1,100,15);=IF(B1<0.25, C1,"");=RANK(D1,$D$1:$D$1000,);=RANK(C1,$C$1:$C$1000,)

Drag down to D1000. Then do a correlation graph of E and F.
Looks pretty linear, no?

Regardless, what Belalic did was prove that there is a statistical model that, when crunched with real numbers data, holds up. The model also explains the gender gap to 96% accuracy.

Sure, but that isn't very hard to do with the definition of "explain" and "accuracy" used in the paper:
Averaged over the 100 top players, the expected male superiority is 341 Elo points and the real one is 353 points. Therefore 96 per cent of the observed difference between male and female players can be attributed to a simple statistical fact—the extreme values from a large sample are likely to be bigger than those from a small one.

351/341 = 96.6%, which they round down to 96%.

(a serious model might have used R2 to estimate the fraction of variance explained, for comparison)

With that definition, I could obtain a model which explains 100% of the gender gap without using any complicated statistical methods: Model the score distribution for women as a normal distribution with μ = 1350 and σ = 350. Model the score distribution for men as a normal distribution with μ = 1650 and σ = 350. Compute the gender gap as μmf.


As for women soccer games, why should they be able to beat men in order to provide an entertaining game?

No reason, my point was just that they were intrinsically inferior at it. It could still be entertaining, just like how people watch children's sports (well, at least when it's their kids playing) and find it entertaining, but this doesn't put children on equal footing with adults.

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-28, 21:35 in Ubuntu: x86_32 is dead because WE SAY SO!
Stirrer of Shit
Post: #447 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Well, sure, but then developers would have to build with static libraries. Because this is hard and also makes for big applications, they compile with dynamic libraries instead and then create one copy of the full library set for each application. They then assume that some other layer for which they conveniently are not responsible would handle that de-duplication, in accordance with a long-standing software developer tradition. ("don't worry man, the compiler will optimize it")

Since dynamic libraries save space with more than one application using them, and dynamic libraries always have more than one application using them, or else they wouldn't be dynamic libraries, this means they always will save space, and thus must be used for AppImages.

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-29, 17:04 in Something about cheese!
Stirrer of Shit
Post: #448 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
>Is the study perfect? No, of course not. But both you and Knapp have not understood the argument that is made.

>The argument being presented is, given the top 100 men and the top 100 female chess players, there is a skill difference that is proportional to the number of men and number of women playing chess.

Well, the skill difference is a function of the relative under-representation of women among chess players; it's not a linear model. Nitpicks aside, I understand the argument being made full well.

>This is then backed up with a model that predicts ELO and subsequent skill gap based on the number of people in both pools, in a theoretical scenario.

They present a model, at least.

>This is then compared with real world data, and concludes that there was a 96% correlation (on average) for this particular data set. If run on another set, it provides a similar outcome.

You've misread it. Nowhere in the paper do they claim a 96% correlation.

Here are all occurences of the string "96" in the paper body:
Although the performance of the 100 best German male chess players is better than that of the 100 best German women, we show that 96 per cent of the observed difference would be expected given the much greater number of men who play chess.
and higher participation rates of men (Charness & Gerchak 1996; Chabris & Glickman 2006)
Our approach is based on the work of Charness & Gerchak (1996)
the estimate of the extreme value from a sample tends to be highly variable (Glickman & Chabris 1996; Glickman 1999).
Averaged over the 100 top players, the expected male superiority is 341 Elo points and the real one is 353 points. Therefore 96 per cent of the observed difference between male and female players can be attributed to a simple statistical fact
we have the recurrence relationship (Harter 1961)


The second-to-last quote is the only one which has anything explaining how they came up with the 96% figure. The word "therefore" in the sentence "Therefore [sic] 96 per cent of the observed difference between male and female players can be attributed to a simple statistical fact" ought to refer back to the previous sentence, "Averaged over the 100 top players, the expected male superiority is 341 Elo points and the real one is 353 points," and the only way to arrive at that figure from that clause is by computing 341/353 (0.966) and then rounding down.

This isn't a correlation. In fact, the term is used only once in the whole paper:
The rating itself is based on the same assumptions as the Elo rating used by the international chess federation. The two correlate highly (r=0.93).


> Note that what was measured and predicted here, is the difference in skill, not ELO as such. This difference is also assumed to follow a bell curve (but ELO in and of itself isn't). The model subsequently calculates what the n:th person of a certain pool should have as skill gap, vs what the n:th person of a different larger pool have.

>Or in other words: The n:th player in Pool A of size K has ELO x. The n:th player in Pool B of smaller size L then should have ELO y. Difference in skill should be z. Let's check by plugging in the values for the first 100 in pool A and B. Oh look, real world data matches the predictions by an average of 96% (worst case 80+%)!
Not quite sure I follow. The paper deals with differences in Elo rating and nothing else. See for instance the Y axis of the second graph.

My question then is why such a complicated formula is needed. Shouldn't the nth highest ranked player in pool B of size L have the same Elo score as someone with rank (K/L) in pool A, if A and B have the same mean and variance, regardless of what distribution they would obey?

Anyway, the real-world data does not match the predictions by "an average of 96%". It would be more accurate to say that the average of the real-world data matches the average of the predictions by 96%. As I am sure you know, this is not the same thing.

>At this point you're arguing facts. You don't happen to believe the world is flat and the moon landing is a big hoax either, do you? :)
The only facts at play here are the Elo databases' scores, which I have not disputed. The rest are conclusions, which may be either right or wrong.

Here's a far simpler analysis:
1) download the same database they used from https://dwz.svw.info/services/files/export/csv/LV-0-csv.zip
2) open spieler.csv in Excel
3) cut off the zero-scorers at the bottom (optional)
4) in O2 punch in "=IF(E2="M",IF(F2="D", I2, ""),"")", in P2 "=IF(E2="W",IF(F2="D", I2, ""),"")", double-click on O2 and P2.
5) enter in some cells "=MEDIAN(O:O)," "MEDIAN(P:P)," and "=MEDIAN(O:O)-MEDIAN(P:P)" and find that, for German players, the median male score is 1560 while the median female score is 1254 (gap = 306)

This should take you around two minutes to do, and completely falsifies their explanation which took four people what looks like a few months to come up with. Yet that study got published, and this is just a throwaway post on an obscure forum.

See, this is why I don't Fucking Love Science™ one bit. It's all a god-damned fraud, that they use big words to claim it provides some kind of value to society. The only reason why you'd want to go to school is either for entertainment or for the degree. Which, mind you, are both perfectly valid reasons, but it's hardly something to be proud of, and certainly not worth $30k a year or whatever Americans spend on it. If they'd have invested $120k in the stock market and spent four years messing with Perl, they'd have been far ahead of their peers. Yet people claim going to college makes you smart, or proves you're smart. Go figure.

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-30, 12:23 in Something about cheese!
Stirrer of Shit
Post: #449 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Posted by wertigon
Sigh...

The model explains the skill gap to 96% on this particular dataset. Why can you not accept that? This is fact, proven by the paper.

Now you are just arguing semantics, and trying to argue math to fit your world view. In essence, you are saying 2 + 2 = 5 if you squint look at it from a certain angle. That is not only dishonest and misleading, it is trying to warp the facts to your own world view.

It doesn't, I'm afraid. If it had a good fit, they would use the far more honest measure of R2. Since it doesn't, they resort to calculating another measure and trying to pass it off as a correlation.

If this were a CO2 study, you'd be arguing there is no global warming because the model predicted this year would have an average temperature of 20 degrees celcius, while in fact it only had 19 degrees celcius.

On the contrary. My issue isn't that their predicted gap is not exactly equal to the gap found in the data, but rather that they take this as evidence of their claims. If you compute it the way they did, the model could be wildly off for all the data points but still end up with an average close to the data's such. See below for such an example.

[edit]
Aaaand your math still fails you.

Article: The model predicts the difference gap between the n:th ranking male and n:th ranking female skill as a function of skill pool size and rank.

With you so far.

Let X be the number of males available.
Let Y be the number of females available.
Let n be the rank we choose to look at (1 - best male vs best female, 50 - 50th best male vs 50th best female).

gap = model(n, X, Y);


Let's code it:


Pool X = getMalePool()
Pool Y = getFemalePool()
abs_error = 0
sum_predicted = 0
sum_real = 0

for n in 1..100:
predicted_gap = model(n, X.size, Y.size)
real_gap = X.members[n].skill - Y.members[n].skill
abs_error += abs(predicted_gap - real_gap)
sum_predicted += predicted_gap
sum_real += real_gap

print("Absolute error: %f", error)
print("Average accuracy: %f", sum_predicted / (sum_real*1.0)) // This amounts to ~96%


Could you please point out which line of this algorithm is incorrect?

[/edit]

The last line, as it uses an incredibly poor not to say misleading definition of accuracy.

Say for instance the "model" would just be this:
int gap(int n, int X, int Y)
{
if (n == 50)
return 35000;
return 0;
}

I'm sure you would agree this is a very poor model. But it would give the same results - 96% or thereabouts.

The proper way to do it would be something like this:
for n in 1..100:
predicted_gap = model(n, X.size, Y.size)
real_gap = X.members[n].skill - Y.members[n].skill
sum_predicted += predicted_gap
sum_real += real_gap

avg_predicted = sum_predicted/100
avg_real = sum_real/100

for n in 1..100:
predicted_gap = model(n, X.size, Y.size)
real_gap = X.members[n].skill - Y.members[n].skill
sstot += (real_gap - avg_real)^2
ssres += (real_gap - predicted_gap)^2
fraction_variance_explained = 1.0 - (ssres/sstot)


There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-30, 19:32 in Something about cheese!
Stirrer of Shit
Post: #450 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Posted by wertigon
Ok, you have the model, you have the data, go nuts. Prove him wrong. :)
I did, two posts ago:
for German players, the median male score is 1560 while the median female score is 1254 (gap = 306)

This could hardly be caused by any outliers, now could it?

Protip: It's built in to the model.

What is?
Also, seems you have confused accuracy and causation.

How?

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-30, 23:37 in Something about cheese!
Stirrer of Shit
Post: #451 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
OK, challenge accepted.

Going to rename c1 and c2 to Magic1 and Magic2, since C1 and C2 are cells, approximate H(x) by ROUND(Gamma+ln(x)), and for σ and μ calculate those on basis of the players with Elo > 0.


(Mean + Magic1*Sigma) + Magic2*Sigma*(FACT(Count)/POWER(FACT((Count-Rank)),Rank))*(LN(Count)-(ROUND(Gamma+ln(Rank-1))))


This gives the expected ranking diff for pair number Count as

((Mean + Magic1*Sigma) + Magic2*Sigma*(FACT(Count_M)/POWER(FACT((Count_M-Rank)),Rank))*(LN(Count_M)-(ROUND(Gamma+LN(Rank-1)))))-((Mean + Magic1*Sigma) + Magic2*Sigma*(FACT(Count_F)/POWER(FACT((Count_F-Rank)),Rank))*(LN(Count_F)-(ROUND(Gamma+LN(Rank-1)))))

Now, these FACT expressions cause Excel to overflow big time, so we replace them with Stirling's approximation, and compute the whole expression as a logarithm:

n! / (n-k)! * n^k

ln(n!) - ln((n-k)! * n^k)

ln(n!) - (ln((n-k)!) + ln(n^k))

ln(n!) - (ln((n-k)!) + k*ln(n))

n! ~= sqrt(2*pi*n)*(n/e)^n

n! ~= (2*pi*n)^0.5*(n/e)^n

ln(n!) ~= ln((2*pi*n)^0.5*(n/e)^n)

ln(n!) ~= ln((2*pi*n)^0.5)+ln((n/e)^n)

ln(n!) ~= 0.5*ln(2*pi*n)+n*ln(n/e)

ln(n!) ~= 0.5*(ln(2)+ln(pi)+ln(n))+n*ln(n)-ln(e)

ln(n!) ~= 0.5*(ln(2)+ln(pi)+ln(n))+n*ln(n)-1

substituting back into the previous expression:

0.5*(ln(2)+ln(pi)+ln(n))+n*ln(n)-1 - 0.5*(ln(2)+ln(pi)+ln((n-k))+(n-k)*ln((n-k))-1 + k*ln(n))

and back to Excel:

0.5*(LN(2)+LN(PI())+LN(Count))+Count*LN(Count)-1 - 0.5*(LN(2)+LN(PI())+LN((Count-Rank))+(Count-Rank)*LN((Count-Rank))-1 + Rank*LN(Count))


And back again:

(Mean + Magic1*Sigma) + Magic2*Sigma*EXP(0.5*(LN(2)+LN(PI())+LN(Count))+Count*LN(Count)-1 - 0.5*(LN(2)+LN(PI())+LN((Count-Rank))+(Count-Rank)*LN((Count-Rank))-1 + Rank*LN(Count)))*(LN(Count)-(ROUND(Gamma+ln(Rank-1))))


And to get the difference:

=((Mean + Magic1*Sigma) + Magic2*Sigma*EXP(0.5*(LN(2)+LN(PI())+LN(Count_M))+Count_M*LN(Count_M)-1 - 0.5*(LN(2)+LN(PI())+LN((Count_M-Rank))+(Count_M-Rank)*LN((Count_M-Rank))-1 + Rank*LN(Count_M)))*(LN(Count_M)-(ROUND(Gamma+LN(Rank-1)))))-((Mean + Magic1*Sigma) + Magic2*Sigma*EXP(0.5*(LN(2)+LN(PI())+LN(Count_F))+Count_F*LN(Count_F)-1 - 0.5*(LN(2)+LN(PI())+LN((Count_F-Rank))+(Count_F-Rank)*LN((Count_F-Rank))-1 + Rank*LN(Count_F)))*(LN(Count_F)-(ROUND(Gamma+LN(Rank-1)))))


Unfortunately, this function doesn't compute. I don't think there are any errors in it, but I get the #NUM! error, which is about as descriptive as it sounds: https://support.office.com/en-us/article/How-to-correct-a-NUM-error-F5193BFC-4400-43F4-88C4-8E1DCCA0428B

But I would think the most likely explanation is it doesn't like big formulae. Because the individual results pop up in the function inspector, except for some of them, like ROUND(Gamma+LN(Rank-1)) which sometimes gives #NUM!, and sometimes different values. Not always the same, which seems like a bug in LibreOffice. On another note, opening that spreadsheet now causes LibreOffice to hang.

(If you can get that function to run in real Excel, please post your results. All you have to do it rename the cells and paste. Count_F is ~6k, Count_M ~60k, and Rank is just a counter from 1 to 100)

Instead, we shall have to eyeball it from the graph provided in figure 3.
From n=0..20 it looks about flat at 275. Then linear up to 425 at 100.

Unfortunately, since the values for N have changed with the male superiority now being even higher, this doesn't work. So you'd have to actually run the function.

In want of this, we can just take a look at figure 3 and notice that it looks extremely close to a linear trend line. And if we use such a trend line, it looks about the same as in their study:


R-squared of 0.74, and perhaps more importantly we can observe the same thing in the paper: this is clearly not the right regression to use. If I just put in random trend lines, "Power" gets a r2 of 0.92, and actually tracks the only prominent feature in the graph. This is not to say that power is appropriate, but if a random regression model from Excel has a stronger predictive power than your model you should probably rethink your life.

Now let's see what the ordinary hypothesis (e.g. women are worse at chess) would predict, using parameters from the data:
nm = 58301
nf = 3666
σm = 335
σf = 364 (!)
μm = 1594
μf = 1335

Surprised about the standard deviation, women IQs have less variation than male such. But this might be a function of the odd cutoff effects in the sample.

Anyway,

Follows the shape much better, R2 = 0.87. Overpredicts the gap though. This might be of the range restriction effects, which seemed to take far more women out of the sample than men.

Probably someone with an actual degree could redo this with fancier software, but it stands to reason that the study is an utter fraud. Why else would you use so complicated methods that they can't even run in Excel?

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-06-30, 23:41 in Board feature requests/suggestions
Stirrer of Shit
Post: #452 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Code tags break the board layout, shouldn't they get a scroll bar if they're too big?

Example.

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-07-01, 13:05 in Board feature requests/suggestions
Stirrer of Shit
Post: #453 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
That IS the best practice. The non-best practice would be something like
=(($J$9+$J$10*$J$12)+$J$11*$J$12*EXP(0.5*(LN(2)+LN(PI())+LN($J$13))+$J$13*LN($J$13)-1-0.5*(LN(2)+LN(PI())+LN(($J$13-D2))+($J$13-D2)*LN(($J$13-D2))-1+D2*LN($J$13)))*(LN($J$13)-(ROUND($J$14+LN(D2-1)))))-(($J$9+$J$10*$J$12)+$J$11*$J$12*EXP(0.5*(LN(2)+LN(PI())+LN($J$13))+$J$13*LN($J$13)-1-0.5*(LN(2)+LN(PI())+LN(($J$13-D2))+($J$13-D2)*LN(($J$13-D2))-1+D2*LN($J$13)))*(LN($J$13)-(ROUND($J$14+LN(D2-1)))))


See how readable it becomes when you use variable names?

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-07-01, 13:26 in Something about cheese!
Stirrer of Shit
Post: #454 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Posted by wertigon
So with your own calculations you find that the mean R² is about 0.74.

This means in your own words that the model explains for 0.74 proving, at the very least, that there is a strong correlation between pool size and skill gap.

Thank you! :)

Not the mean R2, just the R2.

What does the second sentence mean? That the high R2 proves the model is accurate, even if it happens to perform worse a model which does take into account the skill gap?


(X = Elo binned in groups of 50 points, Y = share of population in bin)

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-07-01, 20:43 in Something about cheese!
Stirrer of Shit
Post: #455 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
R2 is fraction of variance explained, there's never any average calculated. So I don't see how "mean R2" is a meaningful term. You square each individual deviation from the model, then sum them all up and divide by the sum of deviations from the mean.

The correlation is about 86%, sure, but what does this have to do with anything?

Oh, and your "other hypothesis"? Women are worse at chess? Yes, this is already a fact. What is your explanation to this? That women inherently are worse at chess? Nope, not holding up according to multiple studies.

Which "multiple studies"?
The one you just linked doesn't seem to make predictions which hold up very well, and perhaps more importantly offers no explanation for the difference in means, which ought to be unrelated to how many they are. (If you've got any theories here, I'd be interested in hearing them, but it seems unlikely)

That women are intrinsically worse at chess seems fairly reasonable to me. It's a well-known fact they have worse spatial intelligence, and according to some observers lower intelligence in general. Chess g-loads at 0.3, so it isn't super high, but I'd reckon visuospatial intelligence plays a higher role than e.g. verbal intelligence in it. And then it ought to follow women would be worse, which is also what the data shows.

I don't see what evidence there is against this, other than cack-handed attempts to come up with some alternative explanation and then claim the field is closed for further study.

And yeah, there's a reason I prefer MATLAB (actually Octave) or Numpy over shitty, shitty excel. :)

Then perhaps you could use your expensive software then and calculate that expression I posted?

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-07-02, 00:15 in Something about cheese!
Stirrer of Shit
Post: #456 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Posted by wertigon
5. Result: The result shows the model has around 96% accuracy from the dataset, and a very strong correlation coefficient of around 86%.

6. Conclusion: The data suggests that the skill gap can mostly be explained by the difference in size, though the model should be tested on other, similar data for conclusive evidence.

This is a scientifically sound method.

See, this is the part I am not so sure about. You can't claim it's a valid model solely on basis of the high R-squared values. A model that predicts a gap of 400 points for all players would get a similarly high coefficient of determination, but that would clearly be a rubbish model. A five-second look at the data shows it doesn't track it well at all, which ought to be a red flag. Not to mention the other issue: if differences in group size are the root cause, then how come the means are different too? The model provides no explanation for this observation.

The linked study I showed you lists them in section 4 paragraph 4. It's 9 studies listed there.

6 of those are dead links save for Spelke (2005), Kerkman (2000), and the 2005 Harvard president speech (only accessible through archive.org).

Spelke's study is the typical word salad that psychologists tend to produce, with absoulutely zero quantitative content. Going to skip the parts about infants since I don't know anything about that and haven't claimed so - it would make perfect sense that they develop in puberty, so let's roll with that.

The metastudy makes quite a few logical leaps which underpin the argument:
This problem may be illustrated by a specific example.
Girls consistently perform better than boys on items in
which the student must determine if the data provided in a
problem are sufficient to answer the problem. Such data-
sufficiency items once appeared on the SAT-M, but they
have been eliminated. According to Chipman (2005), the
decision to eliminate these items was justified on pragmatic
grounds, because performance on the items benefits con-
siderably from coaching. Removing a class of items on
which girls score better nevertheless has the effect of
lowering the scores of girls, relative to boys, and it raises a
question: Did this change increase or decrease the fairness
of the SAT-M as a measure of mathematical ability in men
and women? If boys are more talented than girls, then this
change may have increased the fairness of the test. If boys
and girls are equally talented, then this change increased
the test’s bias against girls. Evaluation of the SAT-M
therefore requires an independently motivated account of
the nature of mathematical talent, its component processes,
and its distribution across boys and girls (Willingham &
Cole, 1997). On pain of circularity, SAT-M scores cannot,
in themselves, reveal whether boys or girls have greater
aptitude for mathematics.

Here, they completely avoid evaluating whether it could be the case of girls getting more of said coaching, and throw away quite a wide chunk of quantitative evidence. And again:
Although high school calculus classes once drew more
boys than girls, that gender gap has closed. Boys and girls
take equally demanding math classes in high school, and
girls get better grades.
...

The contrast between the performance of high school
students on the SAT-M and the performance of college
students in mathematics classes suggests that the SAT-M
systematically underpredicts the performance of high
school girls, relative to boys. Further analyses support that
suggestion. When the SAT-M scores of boys and girls are
matched, girls go on to earn higher grades in college
mathematics classes
(see Royer & Garofoli, 2005, for a
review). The SAT-M’s underprediction of girls’ mathemat-
ics performance
is widely known (e.g., Gallagher & Kauf-
man, 2005; Nature Neuroscience Board of Editors, 2005;
Willingham & Cole, 1997) but is rarely mentioned in
popular discussions of males’ and females’ aptitude for
mathematics (e.g., Cronin, 2005; Pinker, 2002; Summers,
2005).

Considering it's a well-known fact that girls get higher grades relative to subject test scores (leaving aside the issue of schools optimizing for female performance, which definitely is interesting though), this seems like as unreasonable evidence to keep.
That women perform worse under pressure is just a matter of common sense, so to say that they score worse on tests could very well be a matter of that. But it's hardly a robust explanation.

I mean, you could rewrite this whole paragraph and draw an entirely different conclusion from the same data:
The contrast between the performance of high school
students on the SAT-M and the grades of college
students in mathematics classes suggests that grades
systematically underpredict the performance of high
school boys, relative to girls. Further analyses support that
suggestion. When the college mathematics classes' grades
of boys and girls are matched, boys go on to score higher
on the SAT-M (see Royer & Garofoli, 2005, for a
review). The grading system's overprediction of girls’ mathemat-
ics performance is widely known (e.g., Gallagher & Kauf-
man, 2005; Nature Neuroscience Board of Editors, 2005;
Willingham & Cole, 1997) but is rarely mentioned in
popular discussions of males’ and females’ aptitude for
mathematics (e.g., Cronin, 2005; Pinker, 2002; Summers,
2005).


They then pull this tiresome rhetorical trick yet again:
Because these conclusions depend on students’ scores
on the SAT-M, they are open to two interpretations: Either
more boys than girls have extreme talent in mathematics, or
SAT-M scores overestimate the abilities of talented boys,
relative to girls. The SMPY data provide a wealth of
information bearing on these interpretations. Benbow and
her collaborators (Lubinsky & Benbow, 1992; Lubinski,
Webb, Morelock, & Benbow, 2001; Webb, Lubinski, &
Benbow, 2002) looked at the school performance of tal-
ented girls and boys. In early samples, more boys than girls
entered the SMPY program, and boys went on to take more
demanding high school mathematics classes. In the later
samples, however, the numbers of male and female partic-
ipants were nearly equal, as were the numbers of boys and
girls in high school mathematics classes. Although boys
outnumbered girls at the upper tail of the SAT-M, the
SMPY girls got better grades in high school mathematics,
as they have in less selected samples. In college, male and
female SMPY veterans continued to take equally demand-
ing classes and got equally good grades, as do college
women and men generally. They also graduated at equal
rates and obtained an equal number of doctoral degrees
(Lubinski & Benbow, 1992; Lubinski et al., 2001; Webb et
al., 2002).


The entire study consists of this. They start off by a priori assuming the proposition is false. Then they use this to throw out evidence which says otherwise, since it is known to be false. They then use whatever evidence they didn't discard to show the validity of the proposition, which vindicates their decision to throw out the opposing evidence.

Kerkman's I can't find on sci-hub, so I can only read the abstract. I'm not sure what conclusion I'm supposed to draw from it:
There were no sex differences in accuracy or response time on the possible rotation problems, but males were more accurate than females on impossible problems. Neither auditory nor visual interference affected accuracy. The alleged sex difference in mental rotation problems is largely due to the use of problems that are not actually mental rotation problems.

Surely, men being able to solve some problems better than women and some equally well makes them better on average?

The Harvard speech did get memory-holed. But it hardly supports your claims. Christ, the guy was fired for holding it. He suggested that the cause was due to higher variability. Which also sounds plausible, although this might just be because it's utterly unpalatable to suggest the means are different. This guy was President and he got fired for suggesting the variations differed, so imagine what'd happen to a lowly researcher who'd suggest the means did.

Now, do you have any explanation for why women would score worse at chess on average? Or any actual studies to cite other than a tiresome Gish gallop?

Why? You can use octave and numpy yourself, they are free to use. If anything, excel is the expensive tool here.

Oh yeah, MATLAB is. Octave is free. Then again, so is LibreOffice Calc (in common parlance, Excel), and it's utter dogshit.
Octave seems to have about the peculiarities LibreOffice does, but less crashing and less loading of data:
>> T = readtable(filename)
warning: the 'readtable' function is not yet implemented in Octave

Please read <http://www.octave.org/missing.html> to learn how you can
contribute missing functionality.
warning: called from
__unimplemented__ at line 524 column 5
error: 'readtable' undefined near line 1 column 5

>> factorial(1000)
ans = Inf

How do you get it to load a csv file which contains both int and string values?

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-07-02, 12:03 in Something about cheese!
Stirrer of Shit
Post: #457 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
It's easier to test the model on the same data set, isn't it?

The model purports to explain why you've got so few women in the top 100, and says it's because you've got fewer women which then causes extreme effects at the tails.

And let's say that's true, for the sake of argument. Now how could that possibly explain the difference in means or medians?

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-07-02, 18:04 in Something about cheese!
Stirrer of Shit
Post: #458 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
For a liberal definition of data set, yes.

The model is applied to the top 100 players, and purportedly explains why there is a score gap between the male and female players without having to reach the conclusion that it's because female players are worse.

But if you'd apply it to the whole data set, then the explanation provided ("there are more male players than female, ergo there must be more good male than good female players") would fall apart, since the average male player is also far better than the average female player. If the model still predicts a score gap, the model doesn't model what it claims to model since that would indeed imply female players are worse, and if it doesn't, it's wrong, since there is such a gap in the data.

Both of these outcomes would falsify the study, no?

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-07-03, 00:42 in Something about cheese!
Stirrer of Shit
Post: #459 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
I wouldn't think there's any need to apply the model in this case. If you're okay with assuming it's not completely broken, it stands to reason that the player with a rank of 1/2n would have the median score, no matter the value of n.

And if it would indeed predict this, then this wouldn't match up with the data, since the average man is better than the average woman, with a gap of 300-odd points (see graph above), both for mean and median.

The data set is publicly available, the problem is the model which requires you to compute values far too large for LibreOffice or Octave to handle.

Python too, which I think would rule out Numpy unless it can handle calculations better:
>>> ((Mean + Magic1*Sigma) + Magic2*Sigma*(math.factorial(Count_M)/math.pow(math.factorial((Count_M-Rank)),Rank))*(math.log(Count_M)-(round(Gamma+math.log(Rank-1)))))-((Mean + Magic1*Sigma) + Magic2*Sigma*(math.factorial(Count_F)/math.pow(math.factorial((Count_F-Rank)),Rank))*(math.log(Count_F)-(round(Gamma+math.log(Rank-1)))))
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
OverflowError: long int too large to convert to float


And with the optimized version:
>>> ((Mean + Magic1*Sigma) + Magic2*Sigma*math.exp(0.5*(math.log(2)+math.log(math.pi)+math.log(Count_M))+Count_M*math.log(Count_M)-1 - 0.5*(math.log(2)+math.log(math.pi)+math.log((Count_M-Rank))+(Count_M-Rank)*math.log((Count_M-Rank))-1 + Rank*math.log(Count_M)))*(math.log(Count_M)-(round(Gamma+math.log(Rank-1)))))-((Mean + Magic1*Sigma) + Magic2*Sigma*math.exp(0.5*(math.log(2)+math.log(math.pi)+math.log(Count_F))+Count_F*math.log(Count_F)-1 - 0.5*(math.log(2)+math.log(math.pi)+math.log((Count_F-Rank))+(Count_F-Rank)*math.log((Count_F-Rank))-1 + Rank*math.log(Count_F)))*(math.log(Count_F)-(round(Gamma+math.log(Rank-1)))))
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
OverflowError: math range error


There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Posted on 19-07-03, 11:45 in Something about cheese!
Stirrer of Shit
Post: #460 of 717
Since: 01-26-19

Last post: 1525 days
Last view: 1523 days
Posted by funkyass
you don't need to put the entire formula in a single line.

math.factorial(Count_M)/math.pow(math.factorial((Count_M-Rank)),Rank))


that seems off. is it not n!/((n-k)! * nk)?

thats also probably where the overflows are coming from. you can reduce the size of the number that produces with a bit of work, but not easily expressed in a single line.

It is valid, just a bit clumsily expressed with double parens.
And yes, you can approximate ln(n!), but eventually you'll still have to calculate exp(ln(n!)-ln((n-k)!)+ln(n)*k), which causes the overflow.

I don't know what the model's good for. Because the expected value of a player with rank k/n can be calculated by just plugging the quantile into the probit function, no complicated math needed.

There was a certain photograph about which you had a hallucination. You believed that you had actually held it in your hands. It was a photograph something like this.
Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
    Main » sureanem » List of posts
    Get an ad blocker.