EPT[ASL17] Ro8 Preview Pt2: Electrifying
| Forum Index > BW General |
|
prion_
79 Posts
| ||
|
bochs
116 Posts
On April 08 2024 23:51 prion_ wrote: Did you write this with ChatGPT? The writeups for each map in Mini vs. SK are almost identical to each other, and the Snow vs. Bisu one isn't much better. I was just about to say the same lol. Soulkey's ZvP win rate was cited as 65.5% in 3 places, and 54.5% in another place. A human writer wouldn't cite the same fact over and over again, nor make the obvious factual mistake. | ||
|
namkraft
534 Posts
 AfreecaTV/StarCraft League Remastered/17 | ||
BisuDagger
Bisutopia19315 Posts
On April 08 2024 23:51 prion_ wrote: Did you write this with ChatGPT? The writeups for each map in Mini vs. SK are almost identical to each other, and the Snow vs. Bisu one isn't much better. Lol, yeah. To be clear, I never take any payment for my previews and have hand written every article I’m in for over a decade now. Because this is just the ro8 and a preview which really just needs to includes how they got here, who they play, and what maps, I wanted to run an experiment where I fed it all the correct data from LP and tailored it with a specific format. I honestly wanted it to work out because reading Elo-board is so tough to do since they use their real Korean names/everything is in Korean, and gathering tournament stats takes many hours. But as you can see it didn’t nail it. If I try it again, I might just use it for formatting only, but I will write all the actual content myself. | ||
|
qet
Australia245 Posts
I don't understand the point of writing about a game before it happens, especially in such imaginary and fictitious detail that its very unlikely to match the actual game when played. These write-ups were so completely wrong, and so Bisu-biased. Edit: OK so it was just ChatGPT drivel, now I understand... | ||
BLinD-RawR
ALLEYCAT BLUES50643 Posts
I can only do better and from here on and ensure that its all human written previews that hype up upcoming matches properly. | ||
|
BisuDagger
Bisutopia19315 Posts
On April 09 2024 19:53 BLinD-RawR wrote: Its more on me for letting this slip past completely unchecked despite taking the editor credit, there are no excuses that would ever justify the complete oversight in letting this go through. I can only do better and from here on and ensure that its all human written previews that hype up upcoming matches properly. Dude, don't beat yourself up on this. There's nothing wrong with us blending tech for previews. There's only been like 4 writers for all BW content for years now. I think we can produce some seriously cool content using tech blended with writing skills. This preview was a great way for us to gather how many people are actually reading the content of the article versus just seeing who's play and when. Think about it thought, we've spent weeks (months) worth of hours in total writing all the content we have for TL! It's pretty awesome we've been around to do it for so long. | ||
|
BisuDagger
Bisutopia19315 Posts
On April 09 2024 19:35 qet wrote: "Expect Bisu to defend tenaciously, gradually seizing control of the map with superior macro management and precise engagements" (didn't happen...) I don't understand the point of writing about a game before it happens, especially in such imaginary and fictitious detail that its very unlikely to match the actual game when played. These write-ups were so completely wrong, and so Bisu-biased. Edit: OK so it was just ChatGPT drivel, now I understand... I let GPT predict the outcomes based on tons of stats. For the Bisu preview I fed all his PvP results from the Kespa and Sonic era too. Then I went to the elo-board and found all the results there. I also went into the KCM thread and got all the results from that thread over the past month. I actually still did an absurd amount of data research for all these players. What I learned is GPT has a terribly bad memory. I had to repeatedly re-enter data and correct its interpretations. It would properly spit back a page of wiki code in text format, only to give me a wrong answer 10 minutes later. I actually put in the results data for every ASL (17 seasons worth of players, maps, wins/loses, race stats, literally everything) cause I was super excited to have those stats at my disposal. We have to literally research these numbers and do the math every season for the semis/finals for each player which I take pretty seriously when I do this by hand. After spending all these hours I was left with a very defunctory stat machine that kept telling me Jaedong won the most ASLs out of any player.  | ||
|
TMNT
3123 Posts
On April 09 2024 19:23 BisuDagger wrote: I honestly wanted it to work out because reading Elo-board is so tough to do since they use their real Korean names/everything is in Korean, and gathering tournament stats takes many hours. Hey. I can help with that. Will paste some stats for the semi pairs in this thread later today. | ||
|
Flatularissa Fasole
2 Posts
| ||
|
BLinD-RawR
ALLEYCAT BLUES50643 Posts
On April 09 2024 22:00 TMNT wrote: Hey. I can help with that. Will paste some stats for the semi pairs in this thread later today. This, more often than not I use TMNT's compiled data because its just more convenient. | ||
|
Cricketer12
United States13992 Posts
Edit: Well I wasn't expecting that! | ||
|
WombaT
Northern Ireland26402 Posts
On April 09 2024 20:25 BisuDagger wrote: Dude, don't beat yourself up on this. There's nothing wrong with us blending tech for previews. There's only been like 4 writers for all BW content for years now. I think we can produce some seriously cool content using tech blended with writing skills. This preview was a great way for us to gather how many people are actually reading the content of the article versus just seeing who's play and when. Think about it thought, we've spent weeks (months) worth of hours in total writing all the content we have for TL! It's pretty awesome we've been around to do it for so long. Hey least you know people actually read your previews usually so noticed! | ||
|
TMNT
3123 Posts
Hope this helps. Do ask me if you need some specific stats: ![[image loading]](https://i.imgur.com/eAmC6gB.png) Note that the map win rates are for all time. But the players head to head scores are only counted for approx. 6 months, from October 2023, as I think it's more relevant to their recent form. Individual matchup win rates for each player in the last 6 months, all games (with Proleague only stats in brackets): Sharp's TvZ: 59.6% (63.2%) Snow's PvZ: 60.1% (55%) Hero's ZvT: 51.9% (48.2%) Soulkey's ZvP: 55.3% (57.5%) (Sharp's win rate is probably boosted a little bit by him playing in the k league) | ||
|
WindOw
Sweden407 Posts
SnOw > Bisu SoulKey > Mini R04 Sharp > herO Soulkey > snOw - this one honestly is though call, I think Mini had a better chance to stop Soulkey than SnOw does. I think Soulkey fares better of against players like snOw who are less volatile and more straight forward in their gameplay compared to Mini, kinda bummed out Mini lost even though I predicted it, woulda have been nice to see another PvP. What do you guys think about soulkey vs snow? Finals Soulkey > Sharp | ||
|
BLinD-RawR
ALLEYCAT BLUES50643 Posts
| ||
|
Nirli
Bulgaria381 Posts
| ||
|
SerpentFlame
415 Posts
On April 09 2024 19:23 BisuDagger wrote: Lol, yeah. To be clear, I never take any payment for my previews and have hand written every article I’m in for over a decade now. Because this is just the ro8 and a preview which really just needs to includes how they got here, who they play, and what maps, I wanted to run an experiment where I fed it all the correct data from LP and tailored it with a specific format. I honestly wanted it to work out because reading Elo-board is so tough to do since they use their real Korean names/everything is in Korean, and gathering tournament stats takes many hours. But as you can see it didn’t nail it. If I try it again, I might just use it for formatting only, but I will write all the actual content myself. Thanks for all your hard work! As you may know, a lot of folks are probably disappointed in reading GPT content.... I like coming to Teamliquid to talk with humans and read what my fellow humans say. I feel if there's some light GPT editing for grammar and polish, that would be fun to read, but when GPT is generating a lot of the writing, it makes me want to skip these (and I used to look forward to these every set of matches). I know it's a bunch of hard work to write so much -- but speaking for only myself, I prefer to read fewer human words, rather than more AI words. There's something that feels great about reading what a fellow human truly thinks, rather than what an AI could write for them. Just my two cents. It's a tough job. Thanks for your hard work over so many years. | ||
|
HOLYBATS
Turkey766 Posts
On April 10 2024 04:56 TMNT wrote: + Show Spoiler + Hope this helps. Do ask me if you need some specific stats: Note that the map win rates are for all time. But the players head to head scores are only counted for approx. 6 months, from October 2023, as I think it's more relevant to their recent form. Individual matchup win rates for each player in the last 6 months, all games (with Proleague only stats in brackets): Sharp's TvZ: 59.6% (63.2%) Snow's PvZ: 60.1% (55%) Hero's ZvT: 51.9% (48.2%) Soulkey's ZvP: 55.3% (57.5%) (Sharp's win rate is probably boosted a little bit by him playing in the k league) Damn this data really shows it is a coin flip. | ||
|
mtcn77
Turkey666 Posts
Now it is all the tougher who to root for next. It couldn't be more neck and neck... Honestly rooting for all contending Ro4 pros. Let's go! | ||
| ||
