EPTOfficial 'what is Dota anymore' discussion - Page 9536
| Forum Index > Dota 2 General |
|
harodihg
Japan1344 Posts
| ||
|
Maverick_2009
Somewhere2002 Posts
| ||
|
harodihg
Japan1344 Posts
https://www.liquiddota.com/blogs/467325-proletarians-of-all-countries-unite#2 | ||
|
Faruko
Chile34173 Posts
| ||
|
Maverick_2009
Somewhere2002 Posts
| ||
|
Porcupinathy
235 Posts
If anyone knows of a resource doing this already please let me know, otherwise I'll probably maintain the sheet until the end of the season. | ||
|
harodihg
Japan1344 Posts
the broodwar chart is cool AF too, maybe someday there could be a dota one or something similar | ||
|
intotheheart
Canada33091 Posts
| ||
|
Fleetfeet
Canada2833 Posts
| ||
|
Porcupinathy
235 Posts
On May 08 2018 05:09 harodihg wrote: reminds me of this: https://www.youtube.com/watch?v=z2DHpW79w0Y the broodwar chart is cool AF too, maybe someday there could be a dota one or something similar I want to learn a visualization framework at some point so I can do cooler stuff like this. On May 08 2018 05:21 intotheheart wrote: I wonder how we would do the ratings for DotA players/teams. That's actually been a topic I've been wanting to figure out. A few people have done ELO charts, but every one I've seen seems to be pretty much just a representation of what teams did well in the last tournament. I think the tournament structure is too volatile to make stuff like that work, the teams change too often and the meta changes too often. There aren't enough games, so you have to make your rating calculator highly reactive to get anything reasonable. It just doesn't work very well. On May 08 2018 05:30 Fleetfeet wrote: Is there some statistical equivalent for traditional sports? There are tons of examples of Elo-type systems for 1v1 competitions, but I can't think of the equivalent for team sports. FiveThirtyEight does a bunch of stuff for team sports. It's a lot easier for national sports like basketball and baseball because the teams are generally a lot more consistent and closer in skill level. League structures also make sure you're getting a consistent stream of data for every team in the league, which isn't the case in Dota. | ||
|
harodihg
Japan1344 Posts
On May 08 2018 05:21 intotheheart wrote: I wonder how we would do the ratings for DotA players/teams. That's actually been a topic I've been wanting to figure out. MMR strictly based upon wins / losses I assume with some sort of weighting taking into account tournaments - but specifically idk | ||
|
harodihg
Japan1344 Posts
I say this because I watched a LoL tournament game and the score was 4-3 at like 30 minutes lmao | ||
|
HwangjaeTerran
Finland5967 Posts
like Nonelectronicsports don't compare themselves to baseball or poop like that i saw #tlwin lose a came because bm lucked out and got a creep that could root, that's pretty hecking lame | ||
|
intotheheart
Canada33091 Posts
Although that also only applies if we care about the extent to which a player can beat another in a set. This is useful in StarCraft, but I have no idea how useful it'd be for other settings like traditional sports, or chess. | ||
|
tehh4ck3r
Magrathea7084 Posts
| ||
|
Nevuk
United States16280 Posts
On May 08 2018 05:31 Porcupinathy wrote: I want to learn a visualization framework at some point so I can do cooler stuff like this. A few people have done ELO charts, but every one I've seen seems to be pretty much just a representation of what teams did well in the last tournament. I think the tournament structure is too volatile to make stuff like that work, the teams change too often and the meta changes too often. There aren't enough games, so you have to make your rating calculator highly reactive to get anything reasonable. It just doesn't work very well. This isn't terribly odd. Elo is a descriptive measurement, as it's based upon how well the team has done in the past. It's more valuable for historical purposes than it is for predictive behavior, generally speaking (though it's not awful at predictive behavior). Patches also make it really hard for a team to be super dominant for long periods of time, making elo even worse at predicting for dota than they are for say, basketball, on top of the league stuff. Gosugamers has a ranking as well, but I'm not sure what it's based on. I don't think it's elo, as it's not even close to the numbers or rankings that datdota has. https://www.gosugamers.net/dota2/rankings Why is TNC predator #2 on the elo k=32 chart for datdota lol? | ||
|
eieio
United States14512 Posts
| ||
|
Fleetfeet
Canada2833 Posts
...why? For the memes, I guess? | ||
|
goody153
44300 Posts
Also the games were good I prefer Kyle but bulldog did fine | ||
|
Yurie
12161 Posts
| ||
| ||
