|
Is there anyone doing a study of the ratio of gpm / xpm? It is universally understood pro would have a higher gpm and xpm compared to that of a nub, but I'm interested in the ratio.
The rationale is in a free farming lane the pro would be able to land more last hits consistently, while the nub only gets a few. Yet both get the same XPM if there is no denies. As a result the pro would have a better gpm / xpm ratio.
This is very relevant in guides because many times in guide people would advocate for a certain item choices, but unless those item choices hit at a certain timing it won't make sense. A good example is desolator, a deso when the entire enemy team is lvl 8 is very different from a deso when the enemy team is already lvl 16.
I don't have the means of analyzing the replays but if someone can run a study, just take some normal / high / very high games on dotabuff, even taking some tournament games by pros, and simply aggregate the gpm / xpm ratio of everyone in each match. It'll be interesting to see if there's any difference. The hypothesis is pro would have a significantly higher gpm/xpm ratio, meaning in a higher level game on average, one would have more powerful items while on a relatively lower level.
|
Northern Ireland22212 Posts
but the pros will have better farming patterns, so their xpm should be higher too as they are more efficient at killing creeps.
|
It also depends on the hero too. There is a huge difference between farming well on AM vs medusa or CK.
|
Northern Ireland22212 Posts
presumably he will also separate them by heroes. for example nagas tend to a pretty low xpm compared to their gpm, and even their gpm will be lower than other heroes with a comparable cs cuz nagas never really farm ancients
|
I would enjoy seeing this, but like others said, there's a lot of things to correct for.
On top of what's already been mentioned, high levels games may be different from nub games in simple ways like it takes longer for bad people to close games our, etc.
|
I always thought it was pretty clear that Pros have a higher GPM:XPM ratio than lower skilled players, but I just took that for granted I guess. We expect items to come out quicker, that much is sure, but they also seem to come out at a lot lower levels. To me the most notable are people that rush battle furies or BKBs on heroes like PA. They'll seem to do no damage at all, but part of that is they rush those items and get them at a much higher level compared to everyone else and then deal almost no damage because people aren't as squishy and there are more powerful spells online.
Supports especially end up with more I think because more of the map goes unused by carries and there's more gank opportunities from out of position people (never mind skipping tps, wards and dust).
I don't think you really need to correct for much. The sample sizes are so large factors like Naga Siren should just be washed away in the noise unless there's a reason to believe they have a really profound effect.
Though the jungle is heavily weighed towards XP so pros clearing it should get a low GPM:XPM ratio, but I think that's more than offset by supports making the effort to stack those camps and superior zoning. Later on more efficient farming routes ultimately give heroes more time in lane more so than more jungle camps (which many newbies can full clear eventually as a carry anyways) so that would be more about getting more GPM over your XPM anyways.
|
Its not that hard to get high gpm/xpm, if you are given a complete free farm lane that is similiar to the pros, if you are like 4.5k you should be able to farm efficiently enough.
The real difference between the pros and the 5-6k wannabes is the way in which pros build items in lane, beat the offlane out through good selection of items after given an adv, and the way in which they control the creep equilibrium and making it always in favor of them after having an adv, making it almost impossible for any offlaner to stay in lane against them given the right heroes.
|
|
|
The difference will be really massive!
2k:
- no stacking by supports
- no getting a jungle camp while the lane is pushed
- supports camp the carry's lane all early game
- no (or very little) pulling, which also costs some cs
- no idea about keeping the wave equilibrium
- simple mechanics (in an empty lane a pro will miss the occasional creep, 2k will miss some even uncontested)
- your support might try and last hit and ruin your cs
- you might not even have a support, cause they all want to pick the supermegauberstronk carry
There is so many little things, that generate small advantages/get you ahead, which are non existing in lower mmr.
|
I think most people here didn't understand that the OP was talking about the GPM/XPM ratio.
In my opinion, it is not a good measure for several reasons. Let's start with numbers.
Concrete example (of close to equivalent game length):
Winners of a 4.5k game : http://www.dotabuff.com/matches/1757694225
Drow: 602/575 = 1.047 g/xp
WR: 609/519 = 1.173 g/xp
Winners of a 2.xk : http://www.dotabuff.com/matches/1760674918
Jugg: 554/518 = 1.070 g/xp
Venomancer: 554/371 = 1.493 g/xp
I know that one data set comparing two heroes is not significant, but you get the idea.
This is just to illustrate that the ratio of GPM/XPM doesn't change that much between low level and high level games. GPM is smaller but XPM is also smaller so it evens it out.
It is not a good indicator of wether you should go for an item build or another because It does not account for the in-game time (notice m/m = minutes are not considered anymore, units left are g/xp).
Now we could have a discussion as to what measure would be the best fit for figuring out how one's farm is doing overall in the game considering the game time, but this is definitely not what we are looking for.
|
One data point doesn't really prove anything one way or another though. We'd also want to probably know g/xp at different points in time. The end of game g/xp is going to be heavily influenced by the end of the game fighting and/or stomping while it's the early game g/xp that makes a serious distinction on hero level vs when items are obtained.
minutes intentionally drops out too. That's sort of the whole point, the idea of looking at gold xp ratio is whether or not high level players obtain major items at lower levels than lesser skilled players. Minutes is largely irrelevant here outside of largely irrelevant factors like additional creep spawning, day night cycle, and roshan's time mechanics. There's very little about dota that's based purely on what minute the game is. Often when people talk about what time it is in the game it's a shorthand for the expected XP level and item progression for a game of that duration. Like game where everyone is level 16 with 10k networth is in pretty much the same state whether that's at minute 20, 25, or 30.
|
I think that would come down to what the gpm/xpm ratio for killing creeps vs killing heroes is. More hero kills nad less cs in pubs, vice versa for pro games.
|
On September 04 2015 22:21 Logo wrote:
One data point doesn't really prove anything one way or another though. We'd also want to probably know g/xp at different points in time. The end of game g/xp is going to be heavily influenced by the end of the game fighting and/or stomping while it's the early game g/xp that makes a serious distinction on hero level vs when items are obtained.
minutes intentionally drops out too. That's sort of the whole point, the idea of looking at gold xp ratio is whether or not high level players obtain major items at lower levels than lesser skilled players. Minutes is largely irrelevant here outside of largely irrelevant factors like additional creep spawning, day night cycle, and roshan's time mechanics. There's very little about dota that's based purely on what minute the game is. Often when people talk about what time it is in the game it's a shorthand for the expected XP level and item progression for a game of that duration. Like game where everyone is level 16 with 10k networth is in pretty much the same state whether that's at minute 20, 25, or 30.
If you don't consider the in-game time, the ratio itself has no meaning.
Scnenario: Alchemist grabs a bounty rune at level 1 and has skilled greed. He suddenly goes to a ratio of about 5 g/xp. What does this mean?
http://dota2.gamepedia.com/Creeps#Lane_Creeps tells us that nearly every creep gives more experience than gold (exception of ranged creep). The bounty runes gives 50+5*minutes exp and 50+2*minutes gold, following the same logic of higher xp than gold overall.
This means that
- a player farming lane OR jungle by himself will have a low g/xp. (even when last hitting perfectly)
- a player dual or tri-laning will have high g/xp.
This doesn't account for hero kills xp or gold. (too complex to consider right now)
Gold is also received periodically by the sum of 100 gold per minute. This means that a player that just waits and does nothing will see his g/xp ratio increase.
Likewise, exp is received when in range of dying creeps. A player sitting in range and never landing a last hit will see his g/xp ratio decrease.
When we put all of this together: a carry farming safelane solo (& landing every last hit) will have a low g/xp ratio. Same thing if he grabs runes, same thing if he goes jungle or if he misses a last hit. It will however increase if he shares experience or if he wastes time (not in share range, not farming).
This is counter-intuitive but according to this analysis, having a low g/xp ratio is actually more efficient than having a high one. Perhaps I'm wrong somewhere or I misunderstood something. I am open to discussion.
EDIT: The previous version of my text had an error in this paragraph : When we put all of this together: a carry farming safelane solo (& landing every last hit) will have a low g/xp ratio. Same thing if he grabs runes, same thing if he goes jungle. It will however increase if he shares experience, if he wastes time or if he misses a last hit..
missing a last hit actually decreases a g/xp ratio. This adds ambiguousness to the accuracy of the low g/xp ratio meaning being efficient. I don't know how to interpret it yet.
|
This would be a similarity to the Spending Quotient (SQ) in Starcraft 2 that can be determined for each leagues from Bronze to Grandmaster while the values differ heavily from league to league.
If you do some thinking about excluding certain games or only compare certain heroes this can work great for Dota 2.
Idea behind SQ: www.teamliquid.net
SQ on GGTracker: ggtracker.com
SQ on Liquipedia: wiki.teamliquid.net
For people interested in developing such metrics I recommend taking a look how the metrics for Sc2 were developed.
If you need help feel free to pm me when it comes to values to look at and or developing a Software component to support calculations.
|
Gold>XP is somewhat unusual, though less so on individual heroes in high level matches. There are a few causes:
1) Split lane XP without splitting farm. In a trilane, the primary farmer gets extra gold per XP. This doesn't affect the team's GP/XP.
2) Fast pushing. Towers and barracks give more gold then experience. A team that clears the enemy's barracks and tier 4s before level ~14 gets a huge burst of gold that overwhelms their farming on non-core heroes. This affects pubs a bit more than esports games because esports games get surrendered before the tier 4s die.
3) Remote farming. A hero like Naga Siren that gets most of her last hits with no nearby heroes raises the team's gpm/xpm significantly.
4) Sucking. If a hero can't earn experience with any efficiency, they still receive tick gold. This effect is significant at <200 xpm.
5) Reliance on gold-gaining abilities over any normal economy. e.g. Alchemist and non-xp-optimized Midas.
6) Rubber-banding off a team that's already strongly in a gp>xp situation.
Here's an example match: http://www.dotabuff.com/matches/1769184052
The Naga is strongly GPM>XPM due to remote farming. Naga's supports are GPM>XPM due to base-crashing bounties. The Clockwerk and Skywrath are GPM>XPM due to Sucking, while the Morphling combines several factors.
|
United States2186 Posts
I don't know about the ratio, but you can guess very accurately the mmr of mid/carry players up to prolly mid 5k or so by simply looking at their average gpm on a bunch of carry heroes. Better players simply get more gold, up to a point at least.
|
|
|
|
|
|
EPT
