Yes, i mean that and yes, it is there, just hidden its the slope of the curve you plotted, easy way to do that would be to plot the difference between each data point and the next point (and scale that to minutes)
yes, but it would mean less caches. thats why im also interested in a comparison between low and higher level 4v4 to see the amount of caches built
Right, the first derivative should be something similar (it's kind of a weird derivative because it's technically looking at differences between different games ) But it should do, I agree.
some nice graphs, but if you have the time plotting the income is probably also very interesting.
You mean like the "+23" fuel per minute? That's not in there (it wouldn't even be in the replay files, really).
Not sure what to do about the AFKers. Yes, I could filter by ranks (although that is kind of awkward and adds the additional problem of deciding which players to pick (per AT and RT). Maybe reject all games where the CP difference is higher than - say - 2 from the best to the weakest?
That said, it shouldn't effect the fuel/mun income past a certain point as all points WILL be taken by either team.
So, this is resulting from a request by vipper and I thought this might be of general interest.
What I did here was to create some graphs which show the resource income for each game mode per time.
To explain a little bit about where the data comes from:
What the graphs show is the average fuel, manpower and CP income after a match ended in a specific game mode, binned in 3 minute windows. The statistic was derived from all automatch game that I found since the start of 2017. For a total of about 280k, 356k, 202k and 340k games for 1v1, 2v2, 3v3 and 4v4, respectively.
Let's start with command points:
As we can see it seems like CPs increase more quickly the less players are involved. The difference between 1v1 and 4v4 is about 3 minutes. Now, some notes on this:
The CPs should be capped at 32 so I would have expected that within 60 minutes each player would have reached that point. Now, seems like that that is not happening in 3v3 and 4v4 (at least on average). I looked into this a bit more, and found that occasionally the cpearn value is actually higher than the theoretical maximum 32. So, occasionally there seem to be a few hiccups either in the client or on the server side. That said, this happens rarely and probably hasn't much influence here.
What I guess is that in the higher gamemodes occasionally poor players that are near AFK and thus do not accumulate a lot of CPs are carried by stronger players (or you have one of those on either side). So, since the maximum is capped at 32 it means that the average is lower in this cases.
This does not happen as often in the smaller gamemodes because if e.g. in 1v1 a player is AFK, the game will surely end before reaching 60 minutes!
Furthermore, the average of CPs for 4v4 is 0 even after 9 minutes (likewise the average for all other gamemodes is unexpectedly low. That surely can't be correct, right? Well, I guess the "problem" here is that games that end before 9 minutes probably do so because one or more players are basically AFK. So the values at very early times are not expected to be very representative for how many CPs the average player will have by - say - 9 minutes.
Next comes fuel:
What we can see here is that in 3v3 and 4v4 the fuel income indeed is higher than for 1v1 and 2v2. This is expected to the typically slightly higher number of resource points on larger maps and the prevalence of caches in these game modes.
That said, the difference between 1v1 and 4v4 is smaller than I expected (around 200 fuel after 60 minutes). Also, it is noteworthy that before about 14 minutes the fuel incomes is slightly lower in the larger game modes.
This is partially probably due to the longer distances that need to be covered, but surely also subject to the "players are AFK effect" described in the CP discussion. That said, I expect the effect here to be smaller. The reason is that an AFK player does not only earn no CP but he also keeps the opposing players from earning many CPs. In terms of fuel, however, the AFK player might not actually cap for his team but he makes it easier for the other team to cap points.
And for the later stages of the game it shouldn't make a difference because all point will be captured, if not by the team with the near AFK player then by the other team.
Finally manpower (which is basically is the inverse measure of how many units the players have on the field):
Here, the higher the gamemode, the less manpower was earned. This would mean that on average there are higher casualties the smaller the gamemode. No idea what I expected here... This might have to do with players tending to have more units idling around, camping or simply not being very aggressive (see the points above on players being carried); that said, a player that is so AFK that he even fails to build units would push the curve up, so...
Edit:
Ok, after the discussions below, I added a criterion that only games are included in which all players are at least level 10. This restricted the number of games considerably (about a tenth is left). This did help a little bit, but the CP curves for the late games in 4v4 still looked a bit wonky. Then I realized that there are not that many games that last 60 minutes, so it's probably simply not worthwhile looking at the end there (I just happen to like asymptotical behavior as sanity check. Sigh...).
This is how many games there are left for each 3 minute bin:
Updated to include the formerly missing replays from Luvnest vs. Finndeed and Barton vs. TwistedTootsy. Also, pimped the game list to also include commander picks.
Oh look, once a bigger sample size and more even matchups come in the factions converge around a 50% winrate, who would have guessed that?
Well, the bigger sample size helps, but the matchups of day 3 and 4 weren't significantly more even than before; 5 out of the 8 rounds were 3:0 sweeps. So, it is probably more an indication of the diverse faction preferences of the top players (which in itself I find interesting...).
Oh look, its just on meta synergy from soviet, that only mean that thay are bad.
Not 100% sure what you mean. I guess with "meta" you mostly refer to commanders?
If you want to say that the Dshka commanders overperform: Ok, fair enough (not saying they do or do not, but I can see why you might want to argue that way). If you want to say that they are bad if they wouldn't have those doctrines: That would be a stretch based on this data.
Talking about "synergy" with commanders: I find that a tricky thing with Soviets. The faction was clearly designed around a somewhat weaker core that has all the basic tools but does not particularly excel at specific tasks, which then is augmented by the commander to add missing bits.
Like, all commanders have one form of elite infantry or conscript upgrades (guards, shocks, PPsh, partizans), upgraded versions of stock team-weapons (120mil, Dshka) and something that helps for the lategame (e.g. T34-85, Mark Target, IS-2...), unlike OH doctrines.
So, saying that SOV without doctrinal units suck: Well, it seems likely that that's intended (even though you might not like that design choice).
Yeah, pretty sure that's it. The "Don't work" replays seem to be all from the first two days (when GCS was not live so they were played with the mod), and the most others from day 3 & 4.
I think I noticed that indeed you and Aimstrong used the GSC mod when playing the game even though that would probably not be necessary, and this seems to be the only really weird one.
So, I'm speculating but my guess is:
PrussianPrince can't view any replays with the mod because he didn't have the mod and its now not available anymore.
You can't probably view the old replays anymore because I guess the program refuses to apply a mod to an old patch replay file (like, it says: GSC mod & WBP patch? Can't do!).
You can watch your series with Aimstrong because you have the mod installed and it is applied to the current live patch.
Germany
its the slope of the curve you plotted, easy way to do that would be to plot the difference between each data point and the next point (and scale that to minutes)
) But it should do, I agree.
ku2521
HelpingHans
HelpingHans
Thanatos
Thanatos
Relic Entertainment
