|
On September 21 2019 08:19 Charoisaur wrote:
1. INnoVation
2. Maru
3. sOs
4. Zest
5. Life
6. Mvp
7. Stats
8. soO
9. Classic
10. herO
11. Rain
12. Dark
13. MMA
14. MC
15. TaeJa
16. Polt
17. Rogue
18. Serral
19. NesTea
20. PartinG
sOs third ? Is it a joke ? X)
Zest before Life, MC and MMA before Taeja, sure it is.
No need to point serral's ridiculous ranking (at least he comes ahead of nestea) to see the absurdity of this.
|
On September 21 2019 11:29 Xain0n wrote:Show nested quote +On September 21 2019 09:02 Anc13nt wrote:On September 21 2019 08:36 Charoisaur wrote:On September 21 2019 08:28 Wombat_NI wrote:On September 21 2019 08:19 Charoisaur wrote:
1. INnoVation
2. Maru
3. sOs
4. Zest
5. Life
6. Mvp
7. Stats
8. soO
9. Classic
10. herO
11. Rain
12. Dark
13. MMA
14. MC
15. TaeJa
16. Polt
17. Rogue
18. Serral
19. NesTea
20. PartinG Serral at 18 cant see this having any kind of blowback The only one I can see Serral being ahead of is Rogue (even though I disagree) but there's no way he's above any of the others. TaeJa, Polt, MC and MMA have won around the same amount of tournaments as Serral who were on average harder to win so there's no argument there. Putting him above any of the top 10 is ridicolous. At least when we go by pure achievements without giving extra points for dominance which the Serral fans desperately want to do to justify him as the GOAT. I can't argue with a lot of your placements but I am awaiting Xain0n's response to this lol. I just think that Life and Mvp should be swapped with Zest and sOs) and Rain is a bit too high. I would personally put Serral above Polt, Rain and Rogue (but that may change if Rogue wins GSL and Serral doesn't do too well in Blizzcon) but I have hard time saying he is better than the other players you've mentioned. Edit: Yeah I agree that Classic and herO are a bit high (I'd replace them with Taeja and the WoL greats listed below them). Also, I think someone before mentioned that herO didn't actually have very difficult (relatively speaking) IEMs so I might see Serral being above him too but probably not Classic (although I think they are pretty close at this point). I'd put him somewhere between like 12th-15th place. Nakajin's GOAT contest voters ranked Serral 18th before he won two WCS, lost one in the finals, got another one GSL vs the World title and reached ro4 at Asus Rog; that was a pretty low placement already, there is no way Serral can be considered #18 in a GOAT list at the moment. Now, let's imagine an alternate reality in which WCS doesn't exist: Serral's placement seems reasonable at once. There would still be extreme anti-WoL and pro-KeSPa(Maru excluded) bias in Chairosaur's list, I would never agree with it, but Serral's being ranked 18th would not be a point in contention. Too bad that only in the world of Charoisaur(and of few other elects), WCS is worth nothing; six titles put Serral in a tier with Zest, Stats, MC, MMA, maybe Classic and Rain. And I'm not even considering the extra features that Wombat mentioned and that would easily place Serral at the top of this tier.
I mean, I personally rate WCS win as a GSL semi (if it were worth almost nothing I would have put him around 18-20). But having 6 GSL semi + WESG final + Blizzcon + 2 GSL vs World + 2 HSC is probably enough to put him above Rogue, Polt and maybe Rain and herO but I don't think it puts him ahead of other players.
i know you think the WCS are more important but I wonder what you think WCS is worth and why. I am assuming you value it like an ASUS ROG ot GSL final or Super Tournament win, etc.
|
On September 21 2019 07:57 Z3nith wrote:Show nested quote +On September 21 2019 07:34 Locutos wrote:On September 21 2019 05:30 Regisko wrote:
Life > Serral
Top 5:
1. Life
2. Innovation
3. Maru
4. soO
5. $o$
One more blizzcon can get Serral 5th. One more year of domination will get him top-3.
For me it's not only about skill (you can have dozens of different calculation approaches), but also charisma. Then MC is Top 1 This.
Ok, agreed
|
On September 21 2019 19:52 Anc13nt wrote:Show nested quote +On September 21 2019 11:29 Xain0n wrote:On September 21 2019 09:02 Anc13nt wrote:On September 21 2019 08:36 Charoisaur wrote:On September 21 2019 08:28 Wombat_NI wrote:On September 21 2019 08:19 Charoisaur wrote:
1. INnoVation
2. Maru
3. sOs
4. Zest
5. Life
6. Mvp
7. Stats
8. soO
9. Classic
10. herO
11. Rain
12. Dark
13. MMA
14. MC
15. TaeJa
16. Polt
17. Rogue
18. Serral
19. NesTea
20. PartinG Serral at 18 cant see this having any kind of blowback The only one I can see Serral being ahead of is Rogue (even though I disagree) but there's no way he's above any of the others. TaeJa, Polt, MC and MMA have won around the same amount of tournaments as Serral who were on average harder to win so there's no argument there. Putting him above any of the top 10 is ridicolous. At least when we go by pure achievements without giving extra points for dominance which the Serral fans desperately want to do to justify him as the GOAT. I can't argue with a lot of your placements but I am awaiting Xain0n's response to this lol. I just think that Life and Mvp should be swapped with Zest and sOs) and Rain is a bit too high. I would personally put Serral above Polt, Rain and Rogue (but that may change if Rogue wins GSL and Serral doesn't do too well in Blizzcon) but I have hard time saying he is better than the other players you've mentioned. Edit: Yeah I agree that Classic and herO are a bit high (I'd replace them with Taeja and the WoL greats listed below them). Also, I think someone before mentioned that herO didn't actually have very difficult (relatively speaking) IEMs so I might see Serral being above him too but probably not Classic (although I think they are pretty close at this point). I'd put him somewhere between like 12th-15th place. Nakajin's GOAT contest voters ranked Serral 18th before he won two WCS, lost one in the finals, got another one GSL vs the World title and reached ro4 at Asus Rog; that was a pretty low placement already, there is no way Serral can be considered #18 in a GOAT list at the moment. Now, let's imagine an alternate reality in which WCS doesn't exist: Serral's placement seems reasonable at once. There would still be extreme anti-WoL and pro-KeSPa(Maru excluded) bias in Chairosaur's list, I would never agree with it, but Serral's being ranked 18th would not be a point in contention. Too bad that only in the world of Charoisaur(and of few other elects), WCS is worth nothing; six titles put Serral in a tier with Zest, Stats, MC, MMA, maybe Classic and Rain. And I'm not even considering the extra features that Wombat mentioned and that would easily place Serral at the top of this tier. I mean, I personally rate WCS win as a GSL semi (if it were worth almost nothing I would have put him around 18-20). But having 6 GSL semi + WESG final + Blizzcon + 2 GSL vs World + 2 HSC is probably enough to put him above Rogue, Polt and maybe Rain and herO but I don't think it puts him ahead of other players. i know you think the WCS are more important but I wonder what you think WCS is worth and why. I am assuming you value it like an ASUS ROG ot GSL final or Super Tournament win, etc.
I agree, if WCS were worth nothing it would be reasonable to rank Serral being at #18-20; and, to someone, WCS really is worth nothing. I can understand your point of view and I think your ranking is consistent with your ideas.
I think both ASUS Rog and Super Tournament are more valuable than WCS, which is roughly equivalent to a Code S final to me(absimally better in my opinion, but I would value them the same).
While a second place is a very good result and, sometimes, extremely good money, winning a title is the big deal.
I can't justify losing being more valuable than winning when we compare tournaments belonging to the same tier(granted, not all the tournaments that share a tier are equal, but they should be close enough); for a second place to be more valuable than a title, the level of competition where the final is reached must be much higher(and I would question the fact that those two tournaments belong to the same tier).
There were times in HoTS in which this would have been true, if WCS were a tournament devoid of top tier koreans; in 2018-2019, WCS players have improved consistently and I feel that the distance between GSL and WCS is getting thinner and thinner(for example, Reynor vs Serral is a better ZvZ matchup than Dark vs Rogue, and it could very well have happened at a superior level than a Code S semifinal), definitely not enough to think that a second place in Code S is better than a WCS title in my opinion.
|
I was a bit surprised by this article's conclusion. Of course all models/stats are wrong but some are useful.
Saying they depend on start date and timeframe is right, but doesn't exonerate from looking deeper into these. You can average/bootstrap over start dates or pick a sensible one (the first of Jan springs to mind). Similarly, timeframe is very important too - but 6 months or six years are equally inane, so I went ahead and picked a full year, safe in the knowledge you can average those to get 2 or 3 year estimates.
But of course Aligulac's rampant ratings inflation issue makes comparing players across eras apples-to-oranges.
So in order to objectively quantify domination, I've taken an hour that would otherwise have been spent posting, and coded up the pulling out of 1 year rolling, January-starting, offline winrates versus Koreans only, for twenty fantastic SC2 players. (Some players in the lists above don't show rates much above 55% and are therefore not good candidates for inclusion; in consequence I just cut them off the discussion, but I might have forgotten some, do let me know).
First, the winrate in series, as it's more discriminative :
![[image loading]](https://i.imgur.com/T442n6A.png)
Second, the winrate in games :
![[image loading]](https://i.imgur.com/ZhYGSZy.png)
Hopefully those should be legible via browser zooming. These # include this morning's Maru vs Trap series. I deliberately am not presenting my own interpretation of such Stats (hmm) for now, so that you can draw your own conclusions; but note for instance that 2/3 in series seems to be the 'magic' threshold above which players are able to win major tournaments repeatedly. I'll update these numbers after Blizzcon of course, and hope to do a writeup time allowing. I thought TL writers would do this 
|
Northern Ireland25344 Posts
That’s some good reading right there, cheers for putting that together.
Interesting how Classic’s peak WR still probably doesn’t correspond with his absolute peak as a player, really attests to how well he’s been in set planning for GSL this year. Ofc precious eras there’s another StarLeague plus Proleague, so it’s that little bit easier to focus on build prep nowadays.
Rain’s remarkably consistent through his relatively short span, not a huge amount of deviation.
|
On September 21 2019 22:20 Wombat_NI wrote:
That’s some good reading right there, cheers for putting that together.
Interesting how Classic’s peak WR still probably doesn’t correspond with his absolute peak as a player, really attests to how well he’s been in set planning for GSL this year. Ofc precious eras there’s another StarLeague plus Proleague, so it’s that little bit easier to focus on build prep nowadays.
Rain’s remarkably consistent through his relatively short span, not a huge amount of deviation.
Well, my eye got instead caught by Byun having his best win ratio in 2015 and Maru, Gumiho having theirs in 2016.
|
As expected, Rain's the fucking best
|
Canada8989 Posts
On September 21 2019 22:27 Xain0n wrote:Show nested quote +On September 21 2019 22:20 Wombat_NI wrote:
That’s some good reading right there, cheers for putting that together.
Interesting how Classic’s peak WR still probably doesn’t correspond with his absolute peak as a player, really attests to how well he’s been in set planning for GSL this year. Ofc precious eras there’s another StarLeague plus Proleague, so it’s that little bit easier to focus on build prep nowadays.
Rain’s remarkably consistent through his relatively short span, not a huge amount of deviation. Well, my eye got instead caught by Byun having his best win ratio in 2015 and Maru, Gumiho having theirs in 2016.
For Byun his 2015 was almost only online stuff in Lotv beta/first weeks so it's a bit missleading
|
On September 21 2019 22:04 MyLovelyLurker wrote:I was a bit surprised by this article's conclusion. Of course all models/stats are wrong but some are useful. Saying they depend on start date and timeframe is right, but doesn't exonerate from looking deeper into these. You can average/bootstrap over start dates or pick a sensible one (the first of Jan springs to mind). Similarly, timeframe is very important too - but 6 months or six years are equally inane, so I went ahead and picked a full year, safe in the knowledge you can average those to get 2 or 3 year estimates. But of course Aligulac's rampant ratings inflation issue makes comparing players across eras apples-to-oranges. So in order to objectively quantify domination, I've taken an hour that would otherwise have been spent posting, and coded up the pulling out of 1 year rolling, January-starting, offline winrates versus Koreans only, for twenty fantastic SC2 players. (Some players in the lists above don't show rates much above 55% and are therefore not good candidates for inclusion; in consequence I just cut them off the discussion, but I might have forgotten some, do let me know). First, the winrate in series, as it's more discriminative : Second, the winrate in games : Hopefully those should be legible via browser zooming. These # include this morning's Maru vs Trap series. I deliberately am not presenting my own interpretation of such Stats (hmm) for now, so that you can draw your own conclusions; but note for instance that 2/3 in series seems to be the 'magic' threshold above which players are able to win major tournaments repeatedly. I'll update these numbers after Blizzcon of course, and hope to do a writeup time allowing. I thought TL writers would do this
This is really good stuff.
|
On September 21 2019 22:04 MyLovelyLurker wrote:I was a bit surprised by this article's conclusion. Of course all models/stats are wrong but some are useful. Saying they depend on start date and timeframe is right, but doesn't exonerate from looking deeper into these. You can average/bootstrap over start dates or pick a sensible one (the first of Jan springs to mind). Similarly, timeframe is very important too - but 6 months or six years are equally inane, so I went ahead and picked a full year, safe in the knowledge you can average those to get 2 or 3 year estimates. But of course Aligulac's rampant ratings inflation issue makes comparing players across eras apples-to-oranges. So in order to objectively quantify domination, I've taken an hour that would otherwise have been spent posting, and coded up the pulling out of 1 year rolling, January-starting, offline winrates versus Koreans only, for twenty fantastic SC2 players. (Some players in the lists above don't show rates much above 55% and are therefore not good candidates for inclusion; in consequence I just cut them off the discussion, but I might have forgotten some, do let me know). First, the winrate in series, as it's more discriminative : Second, the winrate in games : Hopefully those should be legible via browser zooming. These # include this morning's Maru vs Trap series. I deliberately am not presenting my own interpretation of such Stats (hmm) for now, so that you can draw your own conclusions; but note for instance that 2/3 in series seems to be the 'magic' threshold above which players are able to win major tournaments repeatedly. I'll update these numbers after Blizzcon of course, and hope to do a writeup time allowing. I thought TL writers would do this
It's also important to make note of the quality of opponents. 65% vs lower rated players, for example, may not be as impressive as 60% vs top10 players. If you compare the match history of Serral vs Koreans and Dark vs Koreans, all of Serral's opponents are in the top10 ranking, whereas Dark and Rogue faced mostly Koreans below the 10th rank and many below even the 20th.
There should be a statistical category with the label: median rating of opponents.
|
On September 21 2019 23:38 tigon_ridge wrote:Show nested quote +On September 21 2019 22:04 MyLovelyLurker wrote:I was a bit surprised by this article's conclusion. Of course all models/stats are wrong but some are useful. Saying they depend on start date and timeframe is right, but doesn't exonerate from looking deeper into these. You can average/bootstrap over start dates or pick a sensible one (the first of Jan springs to mind). Similarly, timeframe is very important too - but 6 months or six years are equally inane, so I went ahead and picked a full year, safe in the knowledge you can average those to get 2 or 3 year estimates. But of course Aligulac's rampant ratings inflation issue makes comparing players across eras apples-to-oranges. So in order to objectively quantify domination, I've taken an hour that would otherwise have been spent posting, and coded up the pulling out of 1 year rolling, January-starting, offline winrates versus Koreans only, for twenty fantastic SC2 players. (Some players in the lists above don't show rates much above 55% and are therefore not good candidates for inclusion; in consequence I just cut them off the discussion, but I might have forgotten some, do let me know). First, the winrate in series, as it's more discriminative : Second, the winrate in games : Hopefully those should be legible via browser zooming. These # include this morning's Maru vs Trap series. I deliberately am not presenting my own interpretation of such Stats (hmm) for now, so that you can draw your own conclusions; but note for instance that 2/3 in series seems to be the 'magic' threshold above which players are able to win major tournaments repeatedly. I'll update these numbers after Blizzcon of course, and hope to do a writeup time allowing. I thought TL writers would do this It's also important to make note of the quality of opponents. 65% vs lower rated players, for example, may not be as impression as 60% vs top10 players. If you compare the match history of Serral vs Koreans and Dark vs Koreans, all of Serral's opponents are in the top10 ranking, whereas Dark and Rogue faced mostly Koreans below the 10th rank and many below even the 20th. There should be a statistical category with the label: median rating of opponents.
yeah that's a good idea. If anyone is willing to find that, that would be appreciated
|
There is a site for that, actually.
|
I could believe Cure being in the lower end of the top 10 at the moment. I would put him as 3rd best terran at the moment behind Maru and TY, pretty much equal with Inno.
Top 3 within your race should generally mean top 10 or close to it.
Cure has really good macro and was one of the few terrans punishing Protosses for all the corners protoss cut to get as much probes out before being ready for stim timing, with decent sucsess. He might have the best TvP despite having worse micro than Maru, at least I prefer his approach to the matchup.
I have seen people argue being active in online cups negetively effects a player rating too, I think Cure does well in the likes of Olimo. I haven't really looked into it.
|
On September 21 2019 22:04 MyLovelyLurker wrote:I was a bit surprised by this article's conclusion. Of course all models/stats are wrong but some are useful. Saying they depend on start date and timeframe is right, but doesn't exonerate from looking deeper into these. You can average/bootstrap over start dates or pick a sensible one (the first of Jan springs to mind). Similarly, timeframe is very important too - but 6 months or six years are equally inane, so I went ahead and picked a full year, safe in the knowledge you can average those to get 2 or 3 year estimates. But of course Aligulac's rampant ratings inflation issue makes comparing players across eras apples-to-oranges. So in order to objectively quantify domination, I've taken an hour that would otherwise have been spent posting, and coded up the pulling out of 1 year rolling, January-starting, offline winrates versus Koreans only, for twenty fantastic SC2 players. (Some players in the lists above don't show rates much above 55% and are therefore not good candidates for inclusion; in consequence I just cut them off the discussion, but I might have forgotten some, do let me know). First, the winrate in series, as it's more discriminative : Second, the winrate in games : Hopefully those should be legible via browser zooming. These # include this morning's Maru vs Trap series. I deliberately am not presenting my own interpretation of such Stats (hmm) for now, so that you can draw your own conclusions; but note for instance that 2/3 in series seems to be the 'magic' threshold above which players are able to win major tournaments repeatedly. I'll update these numbers after Blizzcon of course, and hope to do a writeup time allowing. I thought TL writers would do this
Great work, thanks for doing that. If you could find some way to represent the sample sizes it would be even better. Having a 65% win rate in 100 matches is, after all, a lot more impressive than having it in 25 matches.
Oh yeah, and thanks to Mizen for a great article. Here's hoping that the eternal polemicists of goat discussions will go read it once in a while for a sobering perspective.
|
On September 22 2019 00:59 sneakyfox wrote:Show nested quote +On September 21 2019 22:04 MyLovelyLurker wrote:I was a bit surprised by this article's conclusion. Of course all models/stats are wrong but some are useful. Saying they depend on start date and timeframe is right, but doesn't exonerate from looking deeper into these. You can average/bootstrap over start dates or pick a sensible one (the first of Jan springs to mind). Similarly, timeframe is very important too - but 6 months or six years are equally inane, so I went ahead and picked a full year, safe in the knowledge you can average those to get 2 or 3 year estimates. But of course Aligulac's rampant ratings inflation issue makes comparing players across eras apples-to-oranges. So in order to objectively quantify domination, I've taken an hour that would otherwise have been spent posting, and coded up the pulling out of 1 year rolling, January-starting, offline winrates versus Koreans only, for twenty fantastic SC2 players. (Some players in the lists above don't show rates much above 55% and are therefore not good candidates for inclusion; in consequence I just cut them off the discussion, but I might have forgotten some, do let me know). First, the winrate in series, as it's more discriminative : Second, the winrate in games : Hopefully those should be legible via browser zooming. These # include this morning's Maru vs Trap series. I deliberately am not presenting my own interpretation of such Stats (hmm) for now, so that you can draw your own conclusions; but note for instance that 2/3 in series seems to be the 'magic' threshold above which players are able to win major tournaments repeatedly. I'll update these numbers after Blizzcon of course, and hope to do a writeup time allowing. I thought TL writers would do this Great work, thanks for doing that. If you could find some way to represent the sample sizes it would be even better. Having a 65% win rate in 100 matches is, after all, a lot more impressive than having it in 25 matches. Oh yeah, and thanks to Mizen for a great article. Here's hoping that the eternal polemicists of goat discussions will go read it once in a while for a sobering perspective.
If out of those 100 players, 80 (edit: 50/100 is actually much more realistic) of them are no stronger than the average top foreigner, is it really so impressive?
|
On September 22 2019 01:35 tigon_ridge wrote:Show nested quote +On September 22 2019 00:59 sneakyfox wrote:On September 21 2019 22:04 MyLovelyLurker wrote:I was a bit surprised by this article's conclusion. Of course all models/stats are wrong but some are useful. Saying they depend on start date and timeframe is right, but doesn't exonerate from looking deeper into these. You can average/bootstrap over start dates or pick a sensible one (the first of Jan springs to mind). Similarly, timeframe is very important too - but 6 months or six years are equally inane, so I went ahead and picked a full year, safe in the knowledge you can average those to get 2 or 3 year estimates. But of course Aligulac's rampant ratings inflation issue makes comparing players across eras apples-to-oranges. So in order to objectively quantify domination, I've taken an hour that would otherwise have been spent posting, and coded up the pulling out of 1 year rolling, January-starting, offline winrates versus Koreans only, for twenty fantastic SC2 players. (Some players in the lists above don't show rates much above 55% and are therefore not good candidates for inclusion; in consequence I just cut them off the discussion, but I might have forgotten some, do let me know). First, the winrate in series, as it's more discriminative : Second, the winrate in games : Hopefully those should be legible via browser zooming. These # include this morning's Maru vs Trap series. I deliberately am not presenting my own interpretation of such Stats (hmm) for now, so that you can draw your own conclusions; but note for instance that 2/3 in series seems to be the 'magic' threshold above which players are able to win major tournaments repeatedly. I'll update these numbers after Blizzcon of course, and hope to do a writeup time allowing. I thought TL writers would do this Great work, thanks for doing that. If you could find some way to represent the sample sizes it would be even better. Having a 65% win rate in 100 matches is, after all, a lot more impressive than having it in 25 matches. Oh yeah, and thanks to Mizen for a great article. Here's hoping that the eternal polemicists of goat discussions will go read it once in a while for a sobering perspective. If out of those 100 players, 80 of them are no stronger than the average top foreigner, is it really so impressive?
Both of you make good points - finite sample corrections and stratified sampling both belong, as second order corrections. One is trivial to code and the other not, and it's sunny outside , so I'll do both together later. On balance, they are likely to mostly even out.
As per total numbers of games, they do vary, but not more than in a 1 to 2.5 ratio generally for active players. Remember those are games versus Koreans only, and confidence interval width can be tricky.
![[image loading]](https://i.imgur.com/JYGdhaV.png)
Once these are taken into account, the issue of valuing 1. consistency and 2. clutch factor remains, and those are both strongly subjective factors.
|
Canada8989 Posts
On September 21 2019 21:06 Xain0n wrote:Show nested quote +On September 21 2019 19:52 Anc13nt wrote:On September 21 2019 11:29 Xain0n wrote:On September 21 2019 09:02 Anc13nt wrote:On September 21 2019 08:36 Charoisaur wrote:On September 21 2019 08:28 Wombat_NI wrote:On September 21 2019 08:19 Charoisaur wrote:
1. INnoVation
2. Maru
3. sOs
4. Zest
5. Life
6. Mvp
7. Stats
8. soO
9. Classic
10. herO
11. Rain
12. Dark
13. MMA
14. MC
15. TaeJa
16. Polt
17. Rogue
18. Serral
19. NesTea
20. PartinG Serral at 18 cant see this having any kind of blowback The only one I can see Serral being ahead of is Rogue (even though I disagree) but there's no way he's above any of the others. TaeJa, Polt, MC and MMA have won around the same amount of tournaments as Serral who were on average harder to win so there's no argument there. Putting him above any of the top 10 is ridicolous. At least when we go by pure achievements without giving extra points for dominance which the Serral fans desperately want to do to justify him as the GOAT. I can't argue with a lot of your placements but I am awaiting Xain0n's response to this lol. I just think that Life and Mvp should be swapped with Zest and sOs) and Rain is a bit too high. I would personally put Serral above Polt, Rain and Rogue (but that may change if Rogue wins GSL and Serral doesn't do too well in Blizzcon) but I have hard time saying he is better than the other players you've mentioned. Edit: Yeah I agree that Classic and herO are a bit high (I'd replace them with Taeja and the WoL greats listed below them). Also, I think someone before mentioned that herO didn't actually have very difficult (relatively speaking) IEMs so I might see Serral being above him too but probably not Classic (although I think they are pretty close at this point). I'd put him somewhere between like 12th-15th place. Nakajin's GOAT contest voters ranked Serral 18th before he won two WCS, lost one in the finals, got another one GSL vs the World title and reached ro4 at Asus Rog; that was a pretty low placement already, there is no way Serral can be considered #18 in a GOAT list at the moment. Now, let's imagine an alternate reality in which WCS doesn't exist: Serral's placement seems reasonable at once. There would still be extreme anti-WoL and pro-KeSPa(Maru excluded) bias in Chairosaur's list, I would never agree with it, but Serral's being ranked 18th would not be a point in contention. Too bad that only in the world of Charoisaur(and of few other elects), WCS is worth nothing; six titles put Serral in a tier with Zest, Stats, MC, MMA, maybe Classic and Rain. And I'm not even considering the extra features that Wombat mentioned and that would easily place Serral at the top of this tier. I mean, I personally rate WCS win as a GSL semi (if it were worth almost nothing I would have put him around 18-20). But having 6 GSL semi + WESG final + Blizzcon + 2 GSL vs World + 2 HSC is probably enough to put him above Rogue, Polt and maybe Rain and herO but I don't think it puts him ahead of other players. i know you think the WCS are more important but I wonder what you think WCS is worth and why. I am assuming you value it like an ASUS ROG ot GSL final or Super Tournament win, etc. I agree, if WCS were worth nothing it would be reasonable to rank Serral being at #18-20; and, to someone, WCS really is worth nothing. I can understand your point of view and I think your ranking is consistent with your ideas. I think both ASUS Rog and Super Tournament are more valuable than WCS, which is roughly equivalent to a Code S final to me(absimally better in my opinion, but I would value them the same). While a second place is a very good result and, sometimes, extremely good money, winning a title is the big deal. I can't justify losing being more valuable than winning when we compare tournaments belonging to the same tier(granted, not all the tournaments that share a tier are equal, but they should be close enough); for a second place to be more valuable than a title, the level of competition where the final is reached must be much higher(and I would question the fact that those two tournaments belong to the same tier). There were times in HoTS in which this would have been true, if WCS were a tournament devoid of top tier koreans; in 2018-2019, WCS players have improved consistently and I feel that the distance between GSL and WCS is getting thinner and thinner(for example, Reynor vs Serral is a better ZvZ matchup than Dark vs Rogue, and it could very well have happened at a superior level than a Code S semifinal), definitely not enough to think that a second place in Code S is better than a WCS title in my opinion.
On a relative side note, I think the 2013-14 Hots WCS is severely underrated, the NA version at least was I think stronger than a "Serral-less" modern WCS.
WCS players were in at least the semi-final of pretty much every single global event in Hots, with the only exception I can think off beeing WCS season 3 final in 2013 and Kespa cup 2014 (altought they were of course over represented in most of them).
I would take TaeJa, Polt, HerO, JD, Heart, Bomber, Hyun.... over Reynor, Heromarine, Elazer, Neeb... any day.
Plus altought the NA player were mostly shit (outside of Scarlett-Huk), the chinese player were surprisingly decent.
|
On September 22 2019 02:01 MyLovelyLurker wrote:Show nested quote +On September 22 2019 01:35 tigon_ridge wrote:On September 22 2019 00:59 sneakyfox wrote:On September 21 2019 22:04 MyLovelyLurker wrote:I was a bit surprised by this article's conclusion. Of course all models/stats are wrong but some are useful. Saying they depend on start date and timeframe is right, but doesn't exonerate from looking deeper into these. You can average/bootstrap over start dates or pick a sensible one (the first of Jan springs to mind). Similarly, timeframe is very important too - but 6 months or six years are equally inane, so I went ahead and picked a full year, safe in the knowledge you can average those to get 2 or 3 year estimates. But of course Aligulac's rampant ratings inflation issue makes comparing players across eras apples-to-oranges. So in order to objectively quantify domination, I've taken an hour that would otherwise have been spent posting, and coded up the pulling out of 1 year rolling, January-starting, offline winrates versus Koreans only, for twenty fantastic SC2 players. (Some players in the lists above don't show rates much above 55% and are therefore not good candidates for inclusion; in consequence I just cut them off the discussion, but I might have forgotten some, do let me know). First, the winrate in series, as it's more discriminative : Second, the winrate in games : Hopefully those should be legible via browser zooming. These # include this morning's Maru vs Trap series. I deliberately am not presenting my own interpretation of such Stats (hmm) for now, so that you can draw your own conclusions; but note for instance that 2/3 in series seems to be the 'magic' threshold above which players are able to win major tournaments repeatedly. I'll update these numbers after Blizzcon of course, and hope to do a writeup time allowing. I thought TL writers would do this Great work, thanks for doing that. If you could find some way to represent the sample sizes it would be even better. Having a 65% win rate in 100 matches is, after all, a lot more impressive than having it in 25 matches. Oh yeah, and thanks to Mizen for a great article. Here's hoping that the eternal polemicists of goat discussions will go read it once in a while for a sobering perspective. If out of those 100 players, 80 of them are no stronger than the average top foreigner, is it really so impressive? Both of you make good points - finite sample corrections and stratified sampling both belong, as second order corrections. One is trivial to code and the other not, and it's sunny outside , so I'll do both together later. On balance, they are likely to mostly even out. As per total numbers of games, they do vary, but not more than in a 1 to 2.5 ratio generally for active players. Remember those are games versus Koreans only, and confidence interval width can be tricky. Once these are taken into account, the issue of valuing 1. consistency and 2. clutch factor remains, and those are both strongly subjective factors.
Consistency isn't too complicated. It's just the statistical variance of a player's rating sampled over a period of time, which is a simple function. Interval of confidence is interesting, but I think when people bet on who will win a match against whom, the interval probably isn't too relevant. Given two players with equal MU ratings, you're not likely to bet on a player with a 49% vs a player with 50% just because the 49% guy has played twice as many relevant games, unless you have some additional information about him (recent practice or tournament results, for example).
Finding the median of a player's opponents' ratings is a level of complexity only the admin of Aligulac can resolve, since it requires reading through the entire database, to see each player's rating immediately prior to each match, which is something even Aligulac doesn't show (it only updates a player's rating after every 14 days, and doesn't show the change immediately after a match). This stat, combined with the win% stat would provide a more clear picture of how a player's rating is justified. EDIT: Nevermind—I guess inputting the rating value of each 14-day update can still yield a close enough approximation of the median.
Win% alone isn't enough, and we all know how frustrating it is when people misrepresent or misinterpret incomplete statistics (not saying you're doing this). Half-truths are just as bad as, if not worse than, lies.
|
On September 22 2019 02:01 MyLovelyLurker wrote:Show nested quote +On September 22 2019 01:35 tigon_ridge wrote:On September 22 2019 00:59 sneakyfox wrote:On September 21 2019 22:04 MyLovelyLurker wrote:I was a bit surprised by this article's conclusion. Of course all models/stats are wrong but some are useful. Saying they depend on start date and timeframe is right, but doesn't exonerate from looking deeper into these. You can average/bootstrap over start dates or pick a sensible one (the first of Jan springs to mind). Similarly, timeframe is very important too - but 6 months or six years are equally inane, so I went ahead and picked a full year, safe in the knowledge you can average those to get 2 or 3 year estimates. But of course Aligulac's rampant ratings inflation issue makes comparing players across eras apples-to-oranges. So in order to objectively quantify domination, I've taken an hour that would otherwise have been spent posting, and coded up the pulling out of 1 year rolling, January-starting, offline winrates versus Koreans only, for twenty fantastic SC2 players. (Some players in the lists above don't show rates much above 55% and are therefore not good candidates for inclusion; in consequence I just cut them off the discussion, but I might have forgotten some, do let me know). First, the winrate in series, as it's more discriminative : Second, the winrate in games : Hopefully those should be legible via browser zooming. These # include this morning's Maru vs Trap series. I deliberately am not presenting my own interpretation of such Stats (hmm) for now, so that you can draw your own conclusions; but note for instance that 2/3 in series seems to be the 'magic' threshold above which players are able to win major tournaments repeatedly. I'll update these numbers after Blizzcon of course, and hope to do a writeup time allowing. I thought TL writers would do this Great work, thanks for doing that. If you could find some way to represent the sample sizes it would be even better. Having a 65% win rate in 100 matches is, after all, a lot more impressive than having it in 25 matches. Oh yeah, and thanks to Mizen for a great article. Here's hoping that the eternal polemicists of goat discussions will go read it once in a while for a sobering perspective. If out of those 100 players, 80 of them are no stronger than the average top foreigner, is it really so impressive? Both of you make good points - finite sample corrections and stratified sampling both belong, as second order corrections. One is trivial to code and the other not, and it's sunny outside , so I'll do both together later. On balance, they are likely to mostly even out. As per total numbers of games, they do vary, but not more than in a 1 to 2.5 ratio generally for active players. Remember those are games versus Koreans only, and confidence interval width can be tricky. Once these are taken into account, the issue of valuing 1. consistency and 2. clutch factor remains, and those are both strongly subjective factors.
Great work. Already we see that ByuN's peak 2015 was based on only 21 games.
|
|
|
|
|
|
EPT
