2019 Judging and Tech calls discussion

[eppen]

I saw this recently, but did not have time to write about it before now. https://medium.com/@renatamachado_7...he-figure-skating-judging-system-54ef2e033096

The “data science” caught my eye in the first place and got me to click it – something I am interested in for my work life and something that often gets me infuriated and/or amused. A lot of the big data or similar data analyses with “amazing” results esp. when they concern human activities and culture particularly. That is because often it seems that the researchers have not really thought what they are studying and what their results really might mean… This blog left me with similar feelings. Amused because the intention was so clear – Nathan Chen is overscored! – and infuriated because the analysis was really not doing what it was claiming to do.

The author took the results of 10 male figure skaters who have done well in the past two years, isolated the GOEs from their results for a number of years, calculated means and did cool diagrams and scatterplots to visualize the results. She used the final GOEs calculated as a percentage of the BV, not the raw values of -5 to +5.

The premise for the conclusions is that GOEs are a bit like PCS, they get better with time and experience. To a certain extent that is correct, particularly when it comes to steps, spins and chore sequences (IMO). There are rarely such major errors as falls in steps and spins which means that you can follow the development in them. They don’t vary that much year by year (I would imagine without looking at any data actually). Jump GOEs can also be seen as developing with time and practice. A skater introduces a new jump element, it can be not very brilliant at first, but gets better with time and experience, adding more difficult entries is possible etc.

However, when it comes to jumps the GOEs do not follow such a linear development as perhaps imagined by the author. The risk is higher than in the other elements and major mistakes are easier to make. You can have a perfect 3Lz usually, but if you fall you still get -5s and -4s whenever you do that mistake. Skaters can struggle with their jumps one season and be realizing their jump potential to the fullest in another. And it is not a given they will ever do well with the difficult jumps they are desperately trying to do for many seasons (just look at Kevin Reynolds’s track record with quads!).

So, basically calculating the GOE means for jumps per season and comparing them shows how well the skater has been doing jumps that season. If they do really well, the mean is high; if they do hot and cold, it will be average; it they do bad, it can be even in the negative. In addition, using the final GOE means that if a skater does many quads and does them well, the final GOE is very high as the BV is high. For example, Hanyu’s 4Lz got 3,94 as GOE and his 3Lz 1,60 (slightly lower raw scores though) – more than two times higher for the more difficult jump. If a skater with high BV jumps is consistent in many competitions through the season, the GOE mean is going to be also quite high. If a skater does only triples, the GOE will not be that high even though they would get +4s and +5s for every one of them.

The fishy case of Nathan Chen’s GOEs in the last two seasons can be explained by looking at the protocols and realizing that in 2018-9 he did no jump very well, esp. his big jumps were often in the negative side – he was not rewarded unjustly for the poor execution. His only really good competition when it comes to jumps was the Worlds. He won every competition, but mostly because his rivals were doing worse than he was. This fall, on the other hand, Nathan has been pretty consistent in the SP compared to last season. In the FS, not always brilliant with his jumps, but slightly better than last season. The GOEs are generally on the plus side. His jump repertoire consists mostly only of the most difficult jumps, so when he does well, the GOE reward is big.

Also, according to the Skating Scores stats, Nathan’s jump GOE mean in ISU events for 2018-9 was 1.45 and for this season 2.19, which is not even close to the 300% increase as claimed by the author? Which one is the correct one? There seem to be other differences also when one looks at the years in the blog’s per year/diagram, so I wonder if her 2018 actually means calendar year 2018, not season 2018 which would seriously mess up the +3 and +5 systems. If so, forget about this analysis.

Then there is Hanyu who has been getting the top GOE raw scores for quite a few seasons whenever he pulls off any jump. His seasons have been fairly similar for a number of years now – good and bad performances making the GOE totals and means high but not as high as Chen’s this season. If he did a season of mostly good and consistent competitions, he would get a higher GOE mean also. With the 4Lo and 4Lz and a perfect program, probably higher than Chen’s current record. But the difference between his seasons might not be as big as that of Chen’s 2018-9 and 2019-20, because his average is quite high. (And btw, looking at the Skating Scores means for seasons, Hanyu’s GOEs for jumps in 2014-15 0.41 and for 2015-16 is 1.26. And that IS an increase of more than 3 times between two seasons.)

So basically, when you do your any kind of analyses, quantitative or qualitative:
1) know what you are analyzing: what GOE measures and how it works in different contexts 8Spins vs jumps), how its evaluation has changed in the rules over the period you are looking at (and here it is important to maintain the seasonality even with the old +3 system, rule changes apply to seasons, not calendar years)
2) make sure your algorithms work right and don’t mess up different seasons

She did a lot of work of her own when she could have just checked the stats from Skating Scores (which she used to collect the data) and compared them... The sad thing is that this will probably float around the cyberspace for some time and might be used to argue for god knows what...

E

 

[Mishaminion]

Very interesting!

 

[1111bm]

Very interesting!

I think such an amount of ignorance, lack of logical thinking and just... plain stupidity is really rather sad (and annoying). And frightening, when you realise how many people jump onto such 'scientific' analyses without questioning what they're being fed (you know, because "it's been proven by science!") and how powerful a tool this is.

This blog left me with similar feelings. Amused because the intention was so clear – Nathan Chen is overscored!

Ha, yeah... that person describes herself as an asian culture lover, and call me prejudiced, but that already gives me a certain idea when a figure skating fan is concerned.

I literally have no idea what those GOE plots/data collections are supposed to tell us. Why on earth would you calculate the average of the GOEs? And across several different types of jumps, too?

As you already said, this is particularly stupid for an element such as jumps, which because of their risky nature allow for a big range of possible GOEs for one skater. Avering that makes no sense whatsoever, you're literally erasing informational content from your data instead of revealing hidden information and making it more accessible and obvious.

So, basically calculating the GOE means for jumps per season and comparing them shows how well the skater has been doing jumps that season.

Not quite, since the GOE is still awarded by the judges. And just as we shouldn't presume that the judging is biased, we shouldn't do the opposite either. So we can't really draw any conclusions from the GOE re the actual execution of a jump. There's bound to be a correlation of course, because there are certain rules in place, but we can't discount the element of subjectivity and possible bias either.

But yes, it makes absolutely no sense to not take the actual execution of the jumps into account when looking at the GOE.
Of course you will never succeed at complete objectivity, so it's really difficult to incorporate that info into an analysis like this, but it's possible for at least the more obvious mistakes such as falls, two-footing, step-outs and maybe URs/downgrades too (assuming that the judges have to trust these calls and score the GOE accordingly per the guidelines).

She used the final GOEs calculated as a percentage of the BV, not the raw values of -5 to +5.

Using the actual percentage would actually make sense, but only since the introduction of the +5/-5 system, because it's now interchangable with the average of raw GOE (after the lowest and highest score have been eliminated), except in the case of the ChSq. But you have to take the percentage, not the absolute points added on top. Did that person seriously do that? I'm speechless at so much stupidity...

Oh, and just as you, I have also noticed that she seemed to have looked at the GOE per year, not per season... what the...

Also, according to the Skating Scores stats, Nathan’s jump GOE mean in ISU events for 2018-9 was 1.45 and for this season 2.19, which is not even close to the 300% increase as claimed by the author? Which one is the correct one?

I'm not surprised... I see this all the time, resentful fans on the internet, filled with 'righteous anger' getting the numbers wrong (usually giving their claims more weight of course, and making the perceived injustices a lot more outrageous).

No idea if they're lying on purpose to further their agenda or what's happening there. I mean, what's so difficult about quoting/copying the actual numbers correctly from the scoresheet?

 

[CanadianSkaterGuy]

I saw this recently, but did not have time to write about it before now. https://medium.com/@renatamachado_7...he-figure-skating-judging-system-54ef2e033096

The “data science” caught my eye in the first place and got me to click it – something I am interested in for my work life and something that often gets me infuriated and/or amused. A lot of the big data or similar data analyses with “amazing” results esp. when they concern human activities and culture particularly. That is because often it seems that the researchers have not really thought what they are studying and what their results really might mean… This blog left me with similar feelings. Amused because the intention was so clear – Nathan Chen is overscored! – and infuriated because the analysis was really not doing what it was claiming to do.

The author took the results of 10 male figure skaters who have done well in the past two years, isolated the GOEs from their results for a number of years, calculated means and did cool diagrams and scatterplots to visualize the results. She used the final GOEs calculated as a percentage of the BV, not the raw values of -5 to +5.

The premise for the conclusions is that GOEs are a bit like PCS, they get better with time and experience. To a certain extent that is correct, particularly when it comes to steps, spins and chore sequences (IMO). There are rarely such major errors as falls in steps and spins which means that you can follow the development in them. They don’t vary that much year by year (I would imagine without looking at any data actually). Jump GOEs can also be seen as developing with time and practice. A skater introduces a new jump element, it can be not very brilliant at first, but gets better with time and experience, adding more difficult entries is possible etc.

However, when it comes to jumps the GOEs do not follow such a linear development as perhaps imagined by the author. The risk is higher than in the other elements and major mistakes are easier to make. You can have a perfect 3Lz usually, but if you fall you still get -5s and -4s whenever you do that mistake. Skaters can struggle with their jumps one season and be realizing their jump potential to the fullest in another. And it is not a given they will ever do well with the difficult jumps they are desperately trying to do for many seasons (just look at Kevin Reynolds’s track record with quads!).

So, basically calculating the GOE means for jumps per season and comparing them shows how well the skater has been doing jumps that season. If they do really well, the mean is high; if they do hot and cold, it will be average; it they do bad, it can be even in the negative. In addition, using the final GOE means that if a skater does many quads and does them well, the final GOE is very high as the BV is high. For example, Hanyu’s 4Lz got 3,94 as GOE and his 3Lz 1,60 (slightly lower raw scores though) – more than two times higher for the more difficult jump. If a skater with high BV jumps is consistent in many competitions through the season, the GOE mean is going to be also quite high. If a skater does only triples, the GOE will not be that high even though they would get +4s and +5s for every one of them.

The fishy case of Nathan Chen’s GOEs in the last two seasons can be explained by looking at the protocols and realizing that in 2018-9 he did no jump very well, esp. his big jumps were often in the negative side – he was not rewarded unjustly for the poor execution. His only really good competition when it comes to jumps was the Worlds. He won every competition, but mostly because his rivals were doing worse than he was. This fall, on the other hand, Nathan has been pretty consistent in the SP compared to last season. In the FS, not always brilliant with his jumps, but slightly better than last season. The GOEs are generally on the plus side. His jump repertoire consists mostly only of the most difficult jumps, so when he does well, the GOE reward is big.

Also, according to the Skating Scores stats, Nathan’s jump GOE mean in ISU events for 2018-9 was 1.45 and for this season 2.19, which is not even close to the 300% increase as claimed by the author? Which one is the correct one? There seem to be other differences also when one looks at the years in the blog’s per year/diagram, so I wonder if her 2018 actually means calendar year 2018, not season 2018 which would seriously mess up the +3 and +5 systems. If so, forget about this analysis.

Then there is Hanyu who has been getting the top GOE raw scores for quite a few seasons whenever he pulls off any jump. His seasons have been fairly similar for a number of years now – good and bad performances making the GOE totals and means high but not as high as Chen’s this season. If he did a season of mostly good and consistent competitions, he would get a higher GOE mean also. With the 4Lo and 4Lz and a perfect program, probably higher than Chen’s current record. But the difference between his seasons might not be as big as that of Chen’s 2018-9 and 2019-20, because his average is quite high. (And btw, looking at the Skating Scores means for seasons, Hanyu’s GOEs for jumps in 2014-15 0.41 and for 2015-16 is 1.26. And that IS an increase of more than 3 times between two seasons.)

So basically, when you do your any kind of analyses, quantitative or qualitative:
1) know what you are analyzing: what GOE measures and how it works in different contexts 8Spins vs jumps), how its evaluation has changed in the rules over the period you are looking at (and here it is important to maintain the seasonality even with the old +3 system, rule changes apply to seasons, not calendar years)
2) make sure your algorithms work right and don’t mess up different seasons

She did a lot of work of her own when she could have just checked the stats from Skating Scores (which she used to collect the data) and compared them... The sad thing is that this will probably float around the cyberspace for some time and might be used to argue for god knows what...

E

The author fails to take into account that GOE criteria have changed in the latest iteration of the IJS - not just the scale from -3 to +3 and -5 to +5.

The author also really fails to take into consideration, Nathan has gotten much cleaner and more consistent in his jumping over the years, particularly this year. So if you're taking the averages from the past season (especially 2019 where he delivered two essentially perfect competitions at Worlds and the GPF) then obviously the GOE will skew higher in 2019.

In 2019 Chen competed at 2019 Worlds where he was essentially clean, WTT, Skate America, IdF and the GPF. Compare the performances from these events to the seasons prior and you'll see that his average GOE increased because he was simply landing his jumps more consistently, and with greater speed/transitions. Even those who don't like Chen have admitted that his jumps have improved over the years, and add to that tangible things like transitions being added before his quads, and being a popular skater (Uno/Hanyu/Chen all benefit from it...) and it's unsurprising to see the spike in GOE.

This article had something good going there with the stats, but this article is so blatantly biased towards certain skaters that it quickly devolved into pretty much clickbait to attract certain people. The language used throughout is not fairly balanced and very anti-Chen, and it's laughably obvious the way it's written that the author's intent was not just to objectively present data. Look at the concluding statement "Nathan Chen GOEs scores increased enormously in a short period of time, among the currently 10 figure skaters still competing, including two times Olympic champion and 2018 silver medalist, never before a skater presented such an increase on GOE score." To the author, enjoy your JSF cheque, and to Medium, congrats on the click-attracting article that will surely be shared as nauseum for the “point” it’s trying to make.

p.s. where was the outraged article in 2015 when Hanyu's GOE went up 185% from 2014 to 2015? Well duh. Look at his GP events in 2014 vs his GP events in 2015... events like CoC2014 or NHK2014 would skew the average GOE down, while events like NHK2015 and GPF2015 would skew the GOE up.. so obviously he was way cleaner in 2015 and his average GOE would spike up from one year to the next! Compare Chen's events from last year to this year and you can see why his average GOE is much higher in 2019 (especially since its his cleanest “year” so far - since the author is taking into account calendar years not seasons) compared to 2018.

Had Chen fallen on all his jumps at the GPF, and got awarded all -5s and it brought his average GOE for 2019 much lower, suddenly the author’s argument falls apart. In other words it’s less to do with worse judging and more to do with better execution on Chen’s part.

 

[1111bm]

This kind of analysis is also too simplistic in how it's not taking into account and not checking certain additional factors, which can possibly influence the scoring, especially since it's a judged sport.

Averaging GOEs across several competitions for instance. There's a reason why it's often not advisable to compare between competitions.

What about the field of participating skaters i.e. the level of competition and also the actual performances of everyone else in a given competition? This can make a skater stand out more/less or appear in a more favourable/negative/triumphant light in comparison (Nathan at Worlds 2018 anyone? ). I've always felt that this has an effect on the scoring and it would be interesting to check if the numbers support that observation.

Or looking at the GOE of elements across several competitions, independent of the program a given element was performed in and the quality of the rest of the elements in that performance.
My impression is, that judges don't score elements in isolation, I actually don't believe humans are capable of this (I think there are even studies that support this, someone with a background/knowledge in psychology might be able to answer this).
So when a skater has a clean skate with stable looking jumps, the GOE of his other elements (spins, StSq, ChSq) seems to go up too compared to other, less successful outings. In case of a messy skate with several jump mistakes, the opposite seems to be happening. Even when I personally can't see a difference in the quality of the elements (doesn't mean it's not there, but still).
Maybe it's the body language of the skater which changes and the judges respond subconsciously to that, or it's just the overall impression which affects their perception of every element.
So if this is true, it could also lead to an overall increase of the GOE for a consistent skater.

 

[Mishaminion]

I found it interesting that this person's obvious bias so blatantly influenced their conclusion. They're not even trying to hide it despite the "science"

They start by making the accusation that "Old and new skating fans realized that one particular skater is being constantly overscored" and make some rather unscientific comments before (unsurprisingly) concluding Chen is overscored.

In fact they begin with a conclusion, make unsubstantiated assumptions about the data and basically "confirm" their already set in stone conclusion.

There is absolutely nothing in that "analysis" to suggest this person watched any of Nathan's programs to compare with the data either.

 

[CanadianSkaterGuy]

This kind of analysis is also too simplistic in how it's not taking into account and not checking certain additional factors, which can possibly influence the scoring, especially since it's a judged sport.

Averaging GOEs across several competitions for instance. There's a reason why it's often not advisable to compare between competitions.

What about the field of participating skaters i.e. the level of competition and also the actual performances of everyone else in a given competition? This can make a skater stand out more/less or appear in a more favourable/negative/triumphant light in comparison (Nathan at Worlds 2018 anyone? ). I've always felt that this has an effect on the scoring and it would be interesting to check if the numbers support that observation.

Or looking at the GOE of elements across several competitions, independent of the program a given element was performed in and the quality of the rest of the elements in that performance.
My impression is, that judges don't score elements in isolation, I actually don't believe humans are capable of this (I think there are even studies that support this, someone with a background/knowledge in psychology might be able to answer this).
So when a skater has a clean skate with stable looking jumps, the GOE of his other elements (spins, StSq, ChSq) seems to go up too compared to other, less successful outings. In case of a messy skate with several jump mistakes, the opposite seems to be happening. Even when I personally can't see a difference in the quality of the elements (doesn't mean it's not there, but still).
Maybe it's the body language of the skater which changes and the judges respond subconsciously to that, or it's just the overall impression which affects their perception of every element.
So if this is true, it could also lead to an overall increase of the GOE for a consistent skater.

Well said. Often when a skater is skating lights out, things look effortless so the GOE tends to be higher. But when a skater is eking out their program or looking tired, the overall impression of their individual elements is less impressive which can diminish GOE. And it’s true, some skaters dwell on mistakes or are feeling off and they can’t shake it off so you see the tension in their shoulders and in every landing and spin position. Landings that normally get held and have nice posture and look smooth end up being saved or skid on the landing or the carriage is off.

 

[giulia95]

numbers , numbers and naughty math, Pythagoras's"Musica Universalis"

 

[DSQ]

I’m so terrible at maths, thanks everyone for such and interesting discussion and explaining it so easily.

 

[wintersmith]

I've seen some pretty "interesting" twitter users pretending to objectively present evidence for Chen's overscoring. For instance, edited videos contaning many arrows, comparisons and slow mo shots. Of course they show everything Hanyu does better in a jump, but curiously they leave out the landings and musicality, and claim Nathan's preparation time is longer, while showing said preparation in slow motion, making it impossible to accurately compare.

These kinds of posts feed a lot of outrage with many #boycottISU type of hashtags. So basically, if you want to be biased against Nathan, you can easily find "justification" for it, unless you spend a few seconds evaluating if the information is correct or skewed. Which some people don't.

 

[Mishaminion]

I've seen some pretty "interesting" twitter users pretending to objectively present evidence for Chen's overscoring. For instance, edited videos contaning many arrows, comparisons and slow mo shots. Of course they show everything Hanyu does better in a jump, but curiously they leave out the landings and musicality, and claim Nathan's preparation time is longer, while showing said preparation in slow motion, making it impossible to accurately compare.

These kinds of posts feed a lot of outrage with many #boycottISU type of hashtags. So basically, if you want to be biased against Nathan, you can easily find "justification" for it, unless you spend a few seconds evaluating if the information is correct or skewed. Which some people don't.

I've also seen people measuring jump height and distance to the finest degree and claim that because Yuzu's measure more, that means Nathan shouldn't get the GOE bullet for good height and distance.

 

[gkelly]

This is why I keep saying, look at the whole field. Don't just focus on two top skaters in isolation. Especially if you're starting from a position of strong preference for one of those skaters.

 

[Mishaminion]

This is why I keep saying, look at the whole field. Don't just focus on two top skaters in isolation. Especially if you're starting from a position of strong preference for one of those skaters.

Agree.

That is why I started to compare the PCS given to all 6 GPF finalists across both programs and taking into account how well (or poorly) they skated.

 

[CanadianSkaterGuy]

Agree.

That is why I started to compare the PCS given to all 6 GPF finalists across both programs and taking into account how well (or poorly) they skated.

According to my highly scientific analysis the PCS tends to be lower than usual when the average GOE is lower than.

Add it to groundbreaking findings from the Medium article, such as a skater’s average Grade of Execution being higher when they execute everything well, compared to when they poorly execute everything.

 

[Mishaminion]

According to my highly scientific analysis the PCS tends to be lower than usual when the average GOE is lower than.

Add it to groundbreaking findings from the Medium article, such as a skater’s average Grade of Execution being higher when they execute everything well, compared to when they poorly execute everything.

But that was a highly scientific and unbiased article don't you know? From someone who obviously knows a lot about how scoring works.

 

[CanadianSkaterGuy]

But that was a highly scientific and unbiased article don't you know? From someone who obviously knows a lot about how scoring works.

I feel like I should just start writing biased articles analyzing tech calls that are made specifically to pander to those who love the most popular/buzz worthy skaters, and make money off all the site traffic. It would be so easy and one could make a killing because, instead of considering balanced analysis, people are chomping at the bit to read articles/stats that reinforce their views especially ones that are presented/formatted in such a way to make it seem like they have actual scientific basis (see above). And once I’ve collected a decent sum for all the inevitable frenzy of traffic/retweets/etc., be like “Just kidding... I actually think the judging wasn’t that bad... thanks for all the hits!” It’d be manipulative but rather lucrative too.

Seems like that the author will benefit this way from an article such as the one they posted, regardless of the litany of variables they are neglecting to mention, and overwhelming bias ... which is worse than many judges whom people condemn for their bias. And since they’re making coin off of it... good for them... I guess?!

 

[1111bm]

I’m so terrible at maths, thanks everyone for such and interesting discussion and explaining it so easily.

Lucky for you there was barely any math involved in that analysis in the Medium article.

I felt very inspired by the methods employed in that article, so I've decided to do a few light calculations of my own.

I compared the (absolute) jump GOE Yuzuru and Nathan both got in their first two competitions this season, Autumn Classic + Skate Canada and Skate America + Internationaux de France respectively.
I've summed up and combined the jump GOEs from SP and LP.

I was shocked at my findings!
Yuzuru scored 1,81 at ACl and 27,31 at SC... That's an increase of 1400 %! And this happened over the course of only two competitions! A certain improvement throughout the season is to be expected, but doesn't it seem very unrealistic that it has happened this fast and by this much? (she asked suggestively )

Nathan in comparison got 22,55 at SA and it went down to 9,16 at IdF, so he lost 60 %. I don't think that such a sudden loss of abilities seems plausible, so I can only conclude that the judging was corrupt.

Based on my results I see a very clear bias towards Yuzuru and efforts to diminish Nathan's accomplishements.

Also, based on my own impression Nathan is totally better than Yuzuru and the difference is as clear as day. And it's not just me, this view is backed up by different people that I know (some even with an actual skating background), who share my opinion.
Therefore it must be a fact.

 

[wintersmith]

Also, based on my own impression Nathan is totally better than Yuzuru and the difference is as clear as day. And it's not just me, this view is backed up by different people that I know (some even with an actual skating background), who share my opinion.
Therefore it must be a fact.

:rofl2:

Literally the kind of stuff I've read underneath Nathan's videos on youtube (except reversed).

 

[CanadianSkaterGuy]

Lucky for you there was barely any math involved in that analysis in the Medium article.

I felt very inspired by the methods employed in that article, so I've decided to do a few light calculations of my own.

I compared the (absolute) jump GOE Yuzuru and Nathan both got in their first two competitions this season, Autumn Classic + Skate Canada and Skate America + Internationaux de France respectively.
I've summed up and combined the jump GOEs from SP and LP.

I was shocked at my findings!
Yuzuru scored 1,81 at ACl and 27,31 at SC... That's an increase of 1400 %! And this happened over the course of only two competitions! A certain improvement throughout a season is to be expected, but doesn't it seem very unrealistic that it's happened this fast and by this much? (she asked suggestively )

Nathan in comparison got 22,55 at SA and it went down to 9,16 at IdF, so he lost 60 %. I don't think that such a loss of abilities seems plausible, so I can only conclude that the judging was corrupt.

Based on my results I see a very clear bias towards Yuzuru and efforts to diminish Nathan's accomplishements.

Also, based on my own impression Nathan is totally better than Yuzuru and the difference is as clear as day. And it's not just me, this view is backed up by different people that I know (some even with an actual skating background), who share my opinion.
Therefore it must be a fact.

Compelling analysis!

Although he needs more YouTube hits and Google searches before we can truly legitimize his skating.

 

[Mishaminion]

Lucky for you there was barely any math involved in that analysis in the Medium article.

I felt very inspired by the methods employed in that article, so I've decided to do a few light calculations of my own.

I compared the (absolute) jump GOE Yuzuru and Nathan both got in their first two competitions this season, Autumn Classic + Skate Canada and Skate America + Internationaux de France respectively.
I've summed up and combined the jump GOEs from SP and LP.

I was shocked at my findings!
Yuzuru scored 1,81 at ACl and 27,31 at SC... That's an increase of 1400 %! And this happened over the course of only two competitions! A certain improvement throughout the season is to be expected, but doesn't it seem very unrealistic that it has happened this fast and by this much? (she asked suggestively )

Nathan in comparison got 22,55 at SA and it went down to 9,16 at IdF, so he lost 60 %. I don't think that such a sudden loss of abilities seems plausible, so I can only conclude that the judging was corrupt.

Based on my results I see a very clear bias towards Yuzuru and efforts to diminish Nathan's accomplishements.

Also, based on my own impression Nathan is totally better than Yuzuru and the difference is as clear as day. And it's not just me, this view is backed up by different people that I know (some even with an actual skating background), who share my opinion.
Therefore it must be a fact.

Sounds legit


Hehehehe
Brilliant!