Guess you’d better find a new rating system then 
Iyama Yuta made it to top 20 of the rating system fair and square due to recent good results
https://daizj.net/baduk-go-weiqi-ratings/
That’s the great thing about an objective system. Ppl can accuse you of trying for an outcome and then proven wrong when evidence suggest otherwise.
1 Like
I’m not sure why you think this proves any of the objections wrong.
The objection isn’t that your system rates Japanese pros consistently low, it’s that you claim to have a better rating system. The problem is that you’re not making any effort to substantiate that claim by making a proper comparison with other systems.
6 Likes
Do you think goratings is not objective?
5 Likes
Not sure I intentionaly claim that. But I think there are issues with goratings that it rates Japanese players too highly. Like Iyama is top 10.
So I just tried to fit a model using last 365 days worth of games. That’s all. The results are reasonable.
if ppl think that Japanese players deserve to be in the top 10 then i dont want to argue with that. it’s their opinion. In my opinion, i dont think that’s true so I tried to see where they would land using what I think is a reasonable set of data. Does that mean i think my system is “better”? maybe. i definitely thinks it makes sense where it’s putting the Japanese players.
1 Like
You’re still not giving any arguments. Saying something is “reasonable” does not make it so.
Yes, you’re claiming that your rating system is more accurate than goratings: you created it because you were unhappy with goratings.
You give the reason for why you think it’s bad: not because it is inaccurate or because the way it computes rank is incorrect, but because of your own seemingly racist bias that Japanese players cannot be rated that strongly
And you keep claiming that your results are reasonable, but you are giving no comparison showing that this is actually the case!
The only argument you give for your system being good, is that it does indeed satisfy your expectation that Japanese pros should be ranked lower. I’m willing to bet that you would not even have published your system if it turned out that your method rated the Japanese players as high or higher than goratings: you would instead think “this method must be bad, I’ll try something else”
The worst part, is that it is really easy to show if your system is any good: take a snapshot right now of both your system and go-ratings, and look at the next month of pro games being played, and see which of the two rating systems gave a more accurate prediction of the match results.
But you don’t, you were happy with your system as soon as you saw it ranked Japanese pros low, and gave up on any justification.
It’s not an opinion, the whole point of a good rating system is to not have an opinion, but instead a fact-based estimate of how strong a player is.
You had an expectation, created a system that satisfies that expectation and then call it “reasonable” and “objective”. It’s not (unless you can give some statistical evidence that it actually is).
You’re a fraud.
7 Likes
Come on now, I think you’re being a bit hard on xiadaiboy. They clearly stated their intention, namely creating a rating system that ranks japanese players lower. I doubt xiadaiboy was trying to make anybody believe they are unbiased - that would be a bit silly, no?
But I also can’t imagine that anybody will be interested in using a system that was created from such a motivation 
5 Likes
I don’t know, they keep repeating that their system is objective and reasonable, and the only reason given appears to me to be of a racist nature.
Showing that it’s objective is impossible: the system wasn’t objectively chosen.
Showing that it’s reasonable is possible, by making a comparison: but that doesn’t happen
Even refuting the racism claim is easy, but it doesn’t seem that Xiaodaiboy is interested in refuting that claim, despite several accusations.
I stand by my words, this system is not objective, I’m not sure what value it has, and anybody trying to convince others that the system is any good without making an effort of actually showing that it is, is therefore a fraud.
7 Likes
I was (trying to) be a bit sarcastic. Sorry 
“I can’t believe Austrians are rated low in the European Go Database. Better make my own rating system that ranks them higher (and thus others lower). I’m sure people will love it.”
6 Likes
Sarcasm on internet is often not noticed and taken seriously.
But there are sarcasm icons.
By the way is there someone who can develop a rating system for me and make me (don’t want to be too greedy) a 5D?
Thanks!
4 Likes
Ppl can accuse you of trying for an outcome and then proven wrong when evidence suggest otherwise.
That doesn’t prove anybody wrong. You were indeed trying for an outcome, you say so yourself.
5 Likes
Yeah. You can do that piece of work too.
I’m not returning to this topic twice after all discussion had ended to claim that my rating system is superior: that’s you doing that. I’m just asking you to prove it, but you’re just stubbornly stating that it’s not necessary and that we should do it ourself.
Again, it’s you who is trying to convince us you made a good rating system, not the other way around.
3 Likes
i dont have to spend time proving toi you anything. To me, the fact that the rating i made does not inflate Japanese players like goratings is good enough for me. If you genuinely think Iyama Yuta should be in the top 10 (as is what gorating is suggesting now), and I still need to prove anything, then good for you. Like Japan hasn’t won a world title in something like 15 years and 4 wins straight in Nongshim is their FIRST ever. To me, it’s obvious that Japanese players can not be rated so highly.
To me, I am already satisfied based on the above. A more objective study is good, but I am not going to spend time on it.
I don’t need to convince anyone, take it what you will. If you ppl think goratings is better, good for them. But I bet the silent majority trusts my rating better. Again that’s just my opinion. You dont have to agree with it, and I am not going spend effort to convince you or anyone for that matter.
Also, I will stay respective and not call someone a “fraud”. Lol
You are again confusing a group of people with particular members from that group. That a group of people is on average weaker doesn’t mean that one or two people from that group cannot stand out.
In the 1980s, Japan was dominating History of Go Ratings but it doesn’t mean all top players had to be Japanese.
4 Likes
Maybe not directly, but what about the people who made goratings.org? If you claim their rating system is scewed toward ranking japanese players too high, aren’t you accusing them of fraud? Without evidence no less, I would call that disrespectful.
4 Likes
I disagree here, thinking goratings is wrong does not mean it’s a fraud. It could be a good faith effort to make a good ranking which, unfortunately, suffer from certain shortcomings.
I think @Vsotvep’s accusation on this alternative rating was based on the perception that @xiaodaiboy did not make a good faith effort.
1 Like
Ok, I want to know whether your system is rubbish or not, so I made some preparation to test the predictive qualities of your rating system.
First, I took the data from your rating, the one from goratings.org and the one from http://mamumamu0413.web.fc2.com/rating/world/ranking.html that you mentioned earlier, which I’ll refer to as “mamumamu”. I tried my best to connect the three datasets and ended up with a list of 242 players present in each database.
There are a lot more than these 242 players that are present in each ranking, but due to name mismatch (either the English transliteration, or due to differences in Chinese hanzi / Japanese kanji), they aren’t all linked properly, and your database is excluding a lot of pros for which you have less than 12 game records over the last 12 months, making comparison a bit harder. I did some manual work to fix some unmatched hanzi / kanji pairs, but it’s a lot of work. I’m still working on it to connect the missing names.
I then went on and looked at players that are rated unusually high or low in your ranking compared to the other two. Again, it’s hard to measure this, since your ranking is missing about 600 pros that are present in the top 900 of the other two systems. I tried to adjust for this, but it’s hard to compare really. The metric I ended up using is taking the distance between your rating and the average of all ratings, scaling that by dividing it by the average, and subtracting 1/600th of your rating to roughly account for the missing pros.
Anyways, I found the following players that are big outliers (in brackets the number of games played):
| yours | gorating | mamumamu | |
|---|---|---|---|
| Ke Jie (54) | 6 | 2 | 2 |
| Xie Erhao (35) | 38 | 22 | 14 |
| Gu Zihao (46) | 2 | 4 | 5 |
| Tuo Jiaxi (36) | 48 | 32 | 24 |
| Tong Mengcheng (43) | 50 | 33 | 26 |
| Li Qincheng (39) | 9 | 21 | 12 |
| Byun Sangil (81) | 4 | 6 | 8 |
| Huang Yunsong (57) | 15 | 24 | 29 |
| Shi Yue (37) | 45 | 40 | 19 |
| Zhao Chenyu (45) | 37 | 29 | 20 |
| Wang Xinghao (39) | 21 | 38 | 35 |
It think that it makes sense to only look at the outliers from your system, since this is where your system may shine brightest: it’s where your system least agrees with the other two systems. If your system is worse, we’d expect this to be the places where it will fail most spectacularly. I’ve excluded those players that played less than 30 games, to make sure that we only have outliers that your rating is quite certain about (since @claire_yang already demonstrated that your system suffers from this). This excluded the following players, who may be ranked inappropriately only because of the small timewindow of a year:
| yours | gorating | mamumamu | |
|---|---|---|---|
| Chen Haoxin (14) | 30 | 136 | 118 |
| Lee Yeoungkyu (20) | 139 | 87 | 86 |
| Jin Yucheng (15) | 55 | 113 | 131 |
| Choi Jaeyoung (13) | 41 | 71 | 90 |
| Chen Yunong (17) | 43 | 115 | 52 |
I’m not sure how many games each of the players will play over the next (few) month(s). I will track all games played by these pros until I have gathered 120 games. I expect that to take about 3 months, hopefully.
Each time player A wins from player B and A is ranked higher than B in rating system C, I’ll give system C a point, for correctly predicting the outcome. As a very basic test, we’d expect the system with the best predictive capabilities to score highest.
This is pretty rudimentary, though, I’m happy to use a different scoring, as long as it is reasonable and decided upon before the results are in. I’m also happy to reevaluate which players to choose as outliers, or any other critique you (or others) may have.
11 Likes
huh, no japanese players in that list
3 Likes
The most prominent outliers according to my metric that are playing in Japan are:
| xiaodai | gorating | mamumamu | |
|---|---|---|---|
| Fukuoka Kotaru | 136 | 320 | 275 |
| Suzuki Ayumi | 168 | 365 | 410 |
| Hane Naoki | 92 | 151 | 171 |
| Fujisawa Rina | 121 | 193 | 234 |
| Xie Yimin | 182 | 431 | 424 |
None of them are particularly extreme outliers, considering that there are about 900 players in the latter two rating systems, and only about 300 in the first. However, it does believe all these players are rated too highly in xiaodai’s system, compared to the other two, ironically. It should be taken with a grain of salt, though, none of these players are top 100, except perhaps Hane Naoki, and up in those ranks, rating points get very close to each other (it’s sort of like Elo after all, you expect the difference between rank 1 and 10 to be larger than between 11 and 20, and especially than between 251 and 260).
Both Iyama Yuta and Ichiriki Ryo aren’t outliers, since they are rated higher on gorating, and lower on mamumamu. It’ll be fun to track them regardless, though, since they have been the main reason that this ranking was created in the first place.
5 Likes