Alternative rating system to goratings.org

I get a sense that, based on your opinion, most of statistics (the academic discipline as we know today) and most of what we today call data science is not “science”. Physic’s burden of proof is like X-sigma which is a statistical measure.

My rating even publishes the uncertainty surrounding the ratings, so I am pretty sure my ratings are quite scientific. Also, my data, method and code are in the open. So ppl can make up their own minds.

Not at all, most statisticians go to great lengths to eliminate bias, explain what methods they used, clarify with each decision they make why it is reasonable to make it, how it compares to contemporary research, etc.

You’re not doing that at all: you made a rating system based on some seemingly arbitrary choices, don’t explain how it works (if it’s machine learning, do you even know yourself how it works, and what exactly it is measuring), are biased yourself (since there’s no way a Japanese pro can be rated this high without winning an international title) and seem to make no effort to eliminate that bias, and then act like we’re crazy for asking you why your rating system would be any good.

The fact that you think that adding some uncertainty (uncertainty in what? what does this measure exactly?) around ratings makes it scientific, is quite unsettling.

3 Likes

can you name a rating system that doesn’t have what you call “arbitrary choices”? Name one. Just one. So it’s a “feature” of all rating systems. Don’t make it sound like it’s something special about mine.

Lol. I already explained the methodology. In short, it’s a simple application of logistic regression.

lol. I’ve been very respectful, I’d say. based on the below

lol. this is just an example of being scientific, there are many other such examples in my methods. i get a sense that you focus on tiny little things and start trying to twist every word or sentence I say or try to blow them up beyond proportions.

@Vsotvep He has no interest in scientific accuracy, statistics, logical reasoning, or anything of the sort.

I hate to say it, but I think you’re wasting your time here.

3 Likes

yep, you can go into personal attacks. Lol. I’ve been fair and open as ppl can see in this thread. I’ve also been respectful. As I said, my method, code, data are all open. I also pointed out some ways to assess the accuracy of my models vs goratings.

so i think when those studies are done, we will know for sure the relative accuracies of the methods.

The WHR rating you hate so much. It has a 12 page pdf explaining the choices, and includes experimental results based on a sample from KGS, comparisons with other rating systems and explains the algorithm in detail.

No, explaining it would answer the question WHY you’re doing what you’re doing. Have you ever even read a methodology section in an actual scientific paper before? Just stating “it’s regression” is not explaining your methods.

Yes, I feel so indeed.

I’m not going to respond anymore here, unless @xiaodaiboy starts explaining some of the “why” questions.

3 Likes

The choices are also arbitrary. If you put on a nitpicking hat, why does kgs correlate with pro play? Also, have they exhaustively checked vs all methods? Are their method more accurate over all time frames, why didn’t test using one year or two years of data? If they did, why not 3 years 5 years. 3.5 years. There are infinitely many things you can nitpick on.

logistic regression has been known to be a good model for estimating binary outcomes… so why not? i don’t get this question. there are infinitely many methods with which you can derive a rating. I just choose one that has proven to work and I just happen to choose a 1 year time frame.

as i explained, there is an objective method to assess if 1 year or 2 year or whole history would lead to better estimates. that remains to be studied. but i don’t see how someone could reasonably say that 1 year is definitely a bad choice. vs what they choose in goratings.

as i pointed out, i have reason to believe the goratings are not right. it’s an opinion. if ppl don’t hold that opinion and ask me to jump thru a much higher bar than gorating has to then it’s not reasonable. gorating has published a pdf for sure which i haven’t. so maybe it’s just more work to do some more studies.

as i said, at least one other person has a similar result to mine. so to me, i don’t see why goratings would necessarily be the better rating.

anyway, I’ve done what i needed, which is to raise awareness of my ratings.

have a good day ppl.

That’s a good idea.

OGS is used by lot of science people who care more about your methodology as about results. In the very last post you start to explain it when in my opinion you should have start with it.

Happy programing!

Very unbecoming of OGS members to be such sticks in the mud. Tsk tsk.

Not sure “methodology” is the exact word I’d use to describe racial profiling, but I’m glad OP was able to get those obviously terrible Japanese players out of his precious top 20 ratings. They must seem much purer now. No point even comparing the ability for ratings to predict outcomes with results like that.

4 Likes

Your rating system places Byun Sangil #3. Did he win more international titles or participate in more international finals than Ichiriki Ryo? I can’t find the information because I don’t read Korean and Google Translate doesn’t do a good job in translating Korean into English.

More generally, you say that it’s not right to have Ichiriki in the top 20 because he doesn’t satisfy a list L of criteria. What is that precise list of criteria, and how many percent of the top 20 players of your rating satisfy the list L?

1 Like

he did beat Shin Jinseo a few times though. I mean, it’s not about Ichiriki specifically, it’s about Japanese Go in general. They haven’t been able to win title reflects their general lower level of play vs the Korean and Chinese. So if Ichiriki ranks so high, mostly due to his performance in Japanese games. I know Ichiriki did alright last year in international games so he’s about 30-40 on my list, which seems reasonable. But top 20 isn’t.

There isn’t one. The ratings are based on last 365 days of games using a simple logistic regression model. If you think of the ratings as a reflection of the players performance in the last year.

Anyway, I don’t think my ratings is perfect.

Thank you for doubling-down on your confesion that this entire thing is racially motivated.

This ^

5 Likes

@xiaodaiboy You are still not giving scientific arguments. There is no doubt that Japanese Go is currently weaker than Chinese or Korean Go, but Ichiriki is not the average Japanese Go player. Just because a group A is on average weaker than a group B doesn’t mean that a member of a group A can’t be in the top 20.

If a rating placed Ichiriki in the top 3, I would agree that it’s clearly wrong. However as I said, putting him #17 is not unrealistic at all. I am not saying that goratings is necessarily the best system in the world, maybe an equally good rating system would place him #20 or #30, I’m just saying that there is no scientific argument backing up the claim that “Ichiriki shouldn’t be in the top 20”. Or if there is such a scientific argument, you didn’t provide any.

5 Likes

I think it is a fair observation that the Japanese big three have been overrated by GoRatings at times. I do take monthly snapshots (and track their Fox usernames) of Goratings because I find it interesting to track players current form. So I also enjoy this 365 day window as well.

Both are meaningful data and mean different things.

One shortcoming of Goratings (coming from a fan of it and also a subscriber to Go4Go) is the dataset is noticeably smaller than the one @xiaodaiboy is using from Kifu-Depot. It would be informative to see the algorithms used on the same dataset for real discussion on their merits.

I appreciate them both though.

6 Likes

Not talking about a race being inferior in its nature in my opinion. If he was saying japanese go is weak because they are a japonese race i would agree it’s racism.

Your wording seems a bit strong and unaccurate.

6 Likes

But the claim is that Ichiriki Ryo is weak because he is a Japanese player, right? Ichiriki Ryo isn’t considered as an individual here, and personal results in international matchups, like the ones jlt mentioned, are discarded in favour of deciding he is weak based on his nationality / go association.

6 Likes

Also, note that to win an international title, you need to participate. Japanese players tend to not participate in many international tournaments, see this post on L19: Following Iyama Yuta (no world ranking discussions) • Life In 19x19

2 Likes

I give people the benefit of the doubt on things like this.

There is a lot that goes into it. One of my first thoughts is that Ichiriki has less grueling competition to qualify for these international events so may be better at preparing for them. He would absolutely have fewer opportunities and stronger opponents if he were qualifying in China.

The other main knock on Japanese is just sample and algorithm related. Go4Go has a pretty big bias for Japanese games so they are building up rating points in lower leagues by the time they ever face A-league players who steal it from them.

Anyway, think of Japanese player as a player from the Japanese system rather than their race maybe? At least until proven otherwise…

– Just throwing this on: it is another system that places more emphasis on recent results and also places Japanese players very low: http://mamumamu0413.web.fc2.com/rating/world/ranking.html

1 Like