So, Whiteflash asked me to pull out some numbers for 1v1 automatch, similar to those that I presented a while ago here.
I just thought that more people might be interested to see them, so I'll post them here rather than in a PM put don't expect this to be as polished as this would be in a proper newspost, so.... If you want to know what the different graphs show or how they were derived, please check out the post linked above.
The data in this post is what I downloaded from the 9th of April to the 29th of May for a total of 315k 1v1 automatch games. The last bigger patch was applied on the 9th of April, so while there might be a few games from the period before, these should be dominated by post-patch games.
Total map picks:
Map picks per ladder rank:
Faction picks per ladder rank:
Win percentages per faction:
Win percentages per faction and ladder rank:
Win percentages per faction and map:
(error estimates as computed by bootstrapping are between 0.6 and 1.2 percent points (0.6 for the most played maps and factions, 1.2 for the less played maps and factions)).
Profile of Siphon X.
Post History of Siphon X.
Thread: 1v1 automatch stats29 May 2020, 14:56 PM
Thread: Map veto stats (v4)16 May 2020, 21:59 PM
16 May 2020, 17:34 PMLady Xenarra
No worries, the thread is title "Map vetoe stats", so it is normal to assume that the numbers show vetoes. "Map pick stats" could be a more accurate title, but then you have the disconnect to gameplay as players can't actually pick specific maps, so....
16 May 2020, 19:51 PMRosbone
Right, but in fairness you can see that Kharkov is significantly less popular with top players than with the rest. Sure, still is in the top 4, but not tied for first anymore like for pretty much the rest of the ladder...
Thread: Map veto stats (v4)15 May 2020, 18:33 PM
Updated original post (v4), now March 15th to May 15th...
Thread: double fuel point important or not ?30 Apr 2020, 10:55 AM
Well, depends, right?
I think the "Go double fuel" thing is basically the result of stage 2 in people's path to learn the game.
The obvious thing you realize instantly is that the goal of the game is to bring the opponent's VPs down to 0. You do that by holding more VPs than your opponent, so holding VPs is the most important thing to do: Strategy 1 therefore is to rush the VPs and defend them as best as you can.
Now, you do that, you hold 2 VPs, everything works great, and then you get rolled over by tanks. Why is that? Ok, they had both fuels, we didn't because we sat on the VPs, hence they were able to buy tons of tanks while we had none. So, update the strategy stage 2: "Rush the fuels. The more fuel the better, no matter the cost."
Eventually, stage 3 would then be that everything is situational, but at that point it becomes tricky because the best strategy depends on a ton of things.
So, yeah, I can understand that it frustrating if your teammate constantly and blindly pushes for the other fuel, but maybe it is still better than if a teammate would only secure and defend one VP at the start of the game and then flame you for not being able to hold "your" VP?
Edit: More to your initial question, if it is effective: Well, yeah, if you can secure both fuels easily, you will win quickly. The obvious problem is, that will only be able to pull this off if your team is significantly better than the other team; which may or may not be the case; if they are not, then you might loose to much in the attempts and it might cost you the game instead as adamírcz wrote.
In: Strategy Desk
Thread: Some automatch win percentages by map and faction4 Mar 2020, 13:33 PM
4 Mar 2020, 13:09 PMSiphon X.
That said, assuming that the vast majority of games actually is decided by matchmaking mishaps it IS suboptimal if the few last remaining games are actually decided by faction induced map imbalances. Like, if - say - 80 out of the 100 games are decided because of loopsided matchmaking, it kind of becomes relevant if 2 out of the 20 remaining games are actually decided by the map...
Thread: Some automatch win percentages by map and faction4 Mar 2020, 13:09 PM
So, while I was at it, I thought I also made some graphs of the win percentages that the factions achieve on the different maps in automatch in January and February of 2020.
Disclaimer(s): Deriving anything from automatch statistics is somewhat tricky; the matchmaker is supposed to level out player and faction power levels, while on the other hand being limited to the players that actually queued. It seems fair to assume that the matchmaker is only taking into account a player's performance with a single faction and not the map - which is why I think there is some point in looking at these at all - but previous studies showed that map vetoes do depend on the player level so, there might be some imprint of this.
The values below show the difference between a faction's win percentage on a specific map and its average win percentage. I also computed estimates for the standard deviations with a pretty crude boot-strapping technique. I didn't display them in the graphs to not clutter them further, but I'll add some sentence on the estimated accuracy to each section. Note, that generally the values will be more accurate, the more often that specific faction played on this specific map. Conversely, values for less popular maps with less popular factions might be severely less accurate. I put a gray rectangular on the plots to indicate the part that is roughly reflecting the typical uncertainties.
Note: The maps are ordered based on their popularity. The most recent numbers can be found here.
I have no means of checking which side each player started on. In turn this means that maps that have very advantageous starting points (like, "North side will always win!") will end up with very balanced stats.
Finally: Don't fret over the implications. Even the more extreme values below barely exceed a few percent. To put this in perspective: Assuming your faction shows a -2% win percentage on a certain map; this means that out of 100 games you play with this faction on this map, you loose only 2 because of the map; the other 98 games are decided by the players' skill gap and faction balance (so, matchmaking stuff), starting position, random elements and other things. So, practically this will hardly be noticable.
I recorded 155284 1v1 automatch games (yeah, I think my server had some "disc-full" issues for a few days...), about 82k OH, 78k SOV, 73k OKW, 45k USF, 31k UKF. Win percentages are roughly 0.6% for the more played maps and factions but go up to slightly more than 1% for UKF on the lesser played maps:
I have data for about 190k games (AT and RT). I broke the numbers down to team compositions. Most common team was OH/OKW (108k), the least common was UKF/UKF and USF/USF (10k and 14k). For the other compositions I found between 34k to 51k games.
I have about 120k 3v3 games. The diagrams below show the axis win percentages (obviously, allied win percentages will be opposite of those). Uncertainties turned out to be 0.4% for the more commonly played maps to the 0.8% for the less commonly played ones.
I found the data for about 230k 4v4 games. The inaccuracies I found are about 0.3% for the commonly played maps which increases to 0.7% for Vielsalm.
Thread: Map veto stats (v4)3 Mar 2020, 19:04 PM
Updated the original post with map pick stats from January and February 2020.
Thread: Alliance of Defiance18 Jan 2020, 19:44 PM
18 Jan 2020, 17:20 PMRosbone
I just played around with the small barrel stack on palettes. Ahhhh classic Relic.
Yeah, I tried something similar today and got to the same conclusion. I spawned an ISU in the upper right field of the first photo. By mistake I first hit attack ground behind the car on the left. After a couple of shots the car was destroyed (actually leaving nothing) but the shots still got blocked.
Then shot across the barrels which blocked like 6 shots as well. Then I ran over everything with the ISU and after that none of the shots were blocked anymore... Funny.
Thread: Alliance of Defiance17 Jan 2020, 22:47 PM
17 Jan 2020, 21:22 PMRosbone
Ok, got you!
Well, I'm not bothered by the tree and the cars. There might be some justification for somebody ditching some wracks there. Also, it is not sticking out much as everything is just dark grey/black. What get's to me are the elliptical tracks around it and the elliptical fences. To me, that does not look natural/random/realistic at all.
Yeah, I realize that they should make it look like a battlefield. And of course the way they are laid out is supposed to lend itself to gameplay. But with the masses of circles to me they look artificial (not as in "man-made", but in "mapper made" )
I wouldn't say they don't make sense. The wall/fence combination on the right probably would also look good if it would have been wall all the way. But you can totally justify it looking this way: There are two different houses on the left so it is likely that they belong to two different properties. So maybe one of the owners wanted stone and the other didn't or couldn't afford it.
And for the stone wall/wooden wall on the left: Again seems plausible. The low stone wall is likely supposed to be the front side of that property. The stone wall is there to make the house look good and was probably built when the house itself was built. The wooden wall on the right is purely functional and needed to be higher so the people on the street to the right can't observe what is happening in the garden. Also, the wall doesn't need to be representative as it is to the side and likely was built later on anyways.
Yeah, this is something that I'm also not too happy with. Especially Kholodny, but there are other ones in particular on the vanilla maps... Not sure what this is supposed to be. The arrangement of buildings and streets doesn't make it look like a functional village. And the area is cluttered with just random stuff and this makes it look like the people living there left like 20 years ago.
The specific scene you picked out is not too bad, though. Ok, so there is the small shed with the wood and the cart on the bottom right. Seems plausible that that is where this was stored. And then we have the two carts, the dead horse and a couple of boxes: Well, maybe the people living there wanted to pack up their stuff and flee, but obviously something hit the house, and the splat seems to indicate another explosion, so maybe either an air or artillery strike hit the carts, scattered the boxes loaded on the carts and killed the horse?
That said, all the clutter is basically background noise; sure, you can look at it in detail but it is not that eye-catching, UNLIKE THE MASSIVE CIRCLES on AoD
Ok, I guess I think to much about this stuff....
Thread: Alliance of Defiance17 Jan 2020, 20:05 PM
17 Jan 2020, 17:45 PMRosbone
It is what I always say about COH2 maps. If you make it like a real world it does not look real in game. It feels more real when it has things like a suitcase or a cast iron tub in the middle of a field
Not sure how that relates to my point... The fence lines do not look unreal because of the objects that they are made out of, but because of the shape of the line. Unless the farmer is some sort of crazy or an artist or both you will not find fences curved like that.
On your point: Well, fences just made out of the same object would look sterile. It would look fake because these fences are made by hand. A guy hammers a thick piece of wood into the ground, takes four big steps and then plunks down the next post. And ultimately, he will nail some more or less random pieces of wood to those somewhat irregular post or wire them up, depending on what type of fence is used. There will be a random element to that (that's for farmland where a fence is about what it does and not how it looks; in a park or formal garden you would expect fence to be more regular and sterile...).
So, yeah, having some type of random variations of the same type of fence in a fenceline will look more natural. But for this purpose, you typically have different objects for the same type of fence. If you don't have that, putting some random scrub here and there helps breaking up the otherwise regular fence.
Btw, I always find it surprisingly hard to make something look random
Putting random types of fences in the same line, however, would look stupid.
Latest replays uploaded by Siphon X.
Ladders Top 10