|
While many notable people have tried to make the connection between Dota 2 and statistics (such as Nahaz, all the dotabuff guys, datdota, and others), I think they have generally done of poor job of using statistics that can be used to describe games at the highest level. Most analysis has focused on relatively small things like win rates or pick rates of heroes, rather than using team focused statistics to tell a story of how teams are approaching the game. So, I wanted to take a stab at this myself!
The basics of the analysis are presented here: https://compdota2.wordpress.com
Examining the stats from TI (mostly the ratio of personal gold or experience per minute to team gold or experience per minute), I found a lot of interesting things that teams were trying to do. Some teams were incredibly focused on single players, some were more spread out, but I noticed that pretty much every team sacrificed at least one or two players to get some advantages in other places. What was more interesting about this was that most teams were sacrificing players or positions that you maybe wouldn't expect, like EG.
Since the people of this forum would likely care more about this type of analysis than the average Dota 2 fan, I would really appreciate if you guys could take a look at what I'm doing and see if the ideas within speak to you, or if you think I'm on the wrong track looking at the things I'm looking at (though the preliminary results feel interesting, so who knows).
|
Pretty good read and an interesting line of analysis. I would be interested in seeing the pentagons for all the teams; also I think looking at how hero choice influenced the distributions could be interesting (for example-did sumail get sacrificed because he played heroes who require less gold, or did he get less gold relative to other players even when playing more farmy heroes?)
|
For the pentagons you should put some legend for the axes. I'm not sure what 0-20 means in these cases.
For the "sacrifice" teams, I would expect that one of the players is the least farmed overall in some position (e.g., samh for tnc). I don't see this for ig or eg. I think that you are not taking into account some sort of overall efficiency measure that needs to be considered here. It may not be sacrifice if the team is overall getting a higher gpm than the other teams on average. This may not be the case, but I think it needs to be explained.
|
I assume in the following posts you'll look at the conclusions to be drawn from these measures. I think to make the arguments really convincing, you have to look at what was dominant previously (looking at previous TIs, this patch, last patch, etc.). If it's possible to identify broad trends there and link or contrast with the current state then this really can be an indication of what works or doesn't.
|
On August 19 2017 18:26 emperorchampion wrote: I assume in the following posts you'll look at the conclusions to be drawn from these measures. I think to make the arguments really convincing, you have to look at what was dominant previously (looking at previous TIs, this patch, last patch, etc.). If it's possible to identify broad trends there and link or contrast with the current state then this really can be an indication of what works or doesn't.
My goal, at least at the start, is not to necessarily say what works the best, but merely attempt to identify team types and begin to synthesize our understanding of what strategies teams are actually using (even if they aren't entirely on purpose) to win games. Eventually, using the upcoming major and minor's system, we'll hopefully be able to expand this analysis to look at what seems to work.
Unfortunately for previous TI's, the fact that they were played before 7.00 makes the analysis hard to extrapolate to the current environment! However, I plan on trying to account for that in future as I continue this project forward.
|
On August 19 2017 16:04 tehh4ck3r wrote: Pretty good read and an interesting line of analysis. I would be interested in seeing the pentagons for all the teams; also I think looking at how hero choice influenced the distributions could be interesting (for example-did sumail get sacrificed because he played heroes who require less gold, or did he get less gold relative to other players even when playing more farmy heroes?)
I'm actually going to look at both of these when I look at all the team types in depth!
|
this would have to assume that each player in their positions play the same, right? even amongst earthshaker or riki players, the amount of gold you get has to do with how good you are, first of all. there are a lot of things to consider when talking about GPM or even levels.
an example would be teams like OG that typically favor mid to late game. they'll pick Alch often, and suddenly their alch player (ana) gets a very lopsided amount of resources distributed to him. this does not change their observed teamstyle or playstyle per-say, but it really depends on the creativity of the team and what they ended up picking, or who they're playing for officials. i think the correlation between playsyle and picks is something only upper bracket players can observe.
some teams will even pick for a very specific kind of teamfight centered around a handful of timings throughout the game. it's called ease of execution in some cases. and i know percentages account for this, but game length also eventually plays a large role. depending on the meta, carries will top off fast and supports and offlaners who can farm will take a noticeably larger portion.
|
I like the idea for this kind of analysis a lot. It's difficult to have reflective analysis based on net worth when things like like Greevil's Greed and GPM talents exist though.
I think there might be a major flaw in how the gold priority hexagons were made. Are those based on the teams' wins and losses or only their wins? Looking at EG's hexagon it seems incredibly odd that Sumail gets prioritized the least. Other teams focus on shutting down Sumail early on is why he ends up getting below expected prioritization. Having 2 separate hexagons, one based solely on a team's wins and another solely on losses would lead to more insightful analysis, I think.
|
On August 23 2017 08:24 Kishin2 wrote: I like the idea for this kind of analysis a lot. It's difficult to have reflective analysis based on net worth when things like like Greevil's Greed and GPM talents exist though.
I think there might be a major flaw in how the gold priority hexagons were made. Are those based on the teams' wins and losses or only their wins? Looking at EG's hexagon it seems incredibly odd that Sumail gets prioritized the least. Other teams focus on shutting down Sumail early on is why he ends up getting below expected prioritization. Having 2 separate hexagons, one based solely on a team's wins and another solely on losses would lead to more insightful analysis, I think.
He isn't prioritized the least, but rather compared to the mid players in other teams he doesn't get that big of a pie of his own teams net worth. But note that the highest and lowest values for mid players are Ana with 32.1% and James (Xctn) with 24.2%, the differences are mostly not that massive. Not sure about the numbers in between but Ana's number is so high because he played Alch, AM, Invoker, Sven and TA for example. Meanwhile Sumail played 5 games on Earthshaker, some Mirana games, some Puck games, etc.
Seems to me that a lot of the difference is explained by hero choice alone, and it doesn't have to have anything to do with who the team prioritized in game otherwise. For example with Earthshaker mid EG may concentrate heavily on helping Sumail dominate early on, but then after getting blink it's Sumail who just stays off the map and threatens the enemy instead of maximizing his own farm.
|
|
|
|