That game was ridiculous our mid got shit on and then we turtled it out to victory
Dota 2 Beta Thread V: Real Talk Strikes Back [Tutorials]
- Thread starter Artanisix
- Start date
- Status
That game was ridiculous our mid got shit on and then we turtled it out to victory
rhophiehalul78
Member
i'm happy too that there's no concede option. Just had a game an hour ago where we were being stomped and through combination of warding/dewarding/smoke gank and proper positioning in clashes, we defeated a Faceless Void, Huskar, Zeus, Jakiro and Bristleback rape team.
GregLombardi
Member
The problem is that if you look statistically at Dota 2 matchmaking public matches, a very very high % of them are stomps with there being a MASSIVE divide in kill scores.
It's either the matchmaking or the balance, and I've played Dota since its inception -- it isn't the balance.
ah, that's fine. I don't want to play with them anyways!
I think there are way more factors at play in determining match quality than those two, though.
Hero selection, for example, is a big one. No one plays every hero equally well and the matchmaking system obviously can't take into account who you're going to play in a game.
GregLombardi
Member
Agreed. There are a lot of variables to consider when assessing the system and it complicates the math.
Still, the commonality of games ending 30 - 15 or greater on one side or the other is highly suspicious to me, and tells me there is likely bias in the system somewhere.
Gully State
Member
Score gaps happen, even in "evenly matched" games. The idea that every game should be close in score is fallacious. DOTA is snowbally, and the lower down the bracket you go, the less likely a group of 5 randoms will be able to recover from an early kill/farm deficit.
GregLombardi
Member
Close in score is one thing. But double the score on one side or another as often as it happens? That too seems fallacious.
What stats are we citing here?
The Technomancer
card-carrying scientician
Man, these keep getting better and better. This one is actually good.
"Our carry bought a quelling blade and more boots. Our carry bought a quelling blade and more boots!"
Yeah, I don't think it has much to do with the matchmaking algorithms. Most pub teams have little idea how to play from behind, which can cause an early disadvantage to snowball out of control very quickly. And even in pro games, you'll sometimes see completely one-sided stomps happen between two evenly-matched teams.
GregLombardi
Member
A fair point. Since I don't have access to all the data, I am simply citing the nearly 1,200 games I've played. I don't have the data processed, but I would like to someday.
But shouldn't player skill take precedence to team skill in a matchmaking system, especially the closer you get to teamless matchmaking (teamless being a solo queue)?
Dota recursively compounds leads in the game more than any other competative game ive played, id be more shocked if the average game wasnt one sided.
Do you think its matchmaking or just the nature of the game?
For example, what do pro-game scores look like?
GregLombardi
Member
I think there are a number of problems that Valve is not addressing for some reason. Valve has annoyed me a bit as of late -- I don't care about the competitive scene at all, and that's what they seem to be focusing on. I also don't care about game balance.
The game is generally balanced. Focusing all one's effort on balancing it seems a waste of time to me. I'd rather they focus on things like community, better team as well as solo matchmaking, and even graphical changes / enhancements to the heroes, as well as gameplay tweaks.
Edit: Also, there are still so many silly things going on with the game that should have been fixed by now. Chief among them: Why on earth are players still not forced to pick their heroes in all pick before the timer runs out?
Nope, not at all. Not all scores are even. Let's say you have two games, A and B, that ended in 20-10. In game A, there was a TA with 17 kills. In the game B, the kills were more or less evenly split among all the players.
What can you conclude about the two games?
Well, clearly, TA from game A must've been a really good player to dominate so hard, and so you might say that the matchmaking dun goofed. But is that the only option? Her opponent might've been the worst player on Dire, while their best player (the one matchmaking threw into DIre to counterbalance the TA player) might be playing support. Maybe it's just a matchup fail, and someone tried (stupidly) to go mid as Morphling vs TA. That is also a possibility. It's also possible that TA didn't win mid at all, but managed to snowball through lucky runes and ganks while the mid Morphling freefarmed. All three of these outcomes happen, in my experience, so it's just not enough to look at TA's score and go "ah, matchmaking fail".
What about the second game? Can we conclude that, because Radiant had a good overall score compared to DIre, that every player on Radiant was better? Still no. There's the matter of matchups and composition. Maybe Radiant had a team fight comp, and they forced a lot of early team fights. Maybe Radiant had a passive farming comp, and they just turtled until their Gyro came out with Rapier, but his teammates still mopped up the kills. Maybe Radiant had a good aggressive roamer like VS, Chen or Enchant, who, while not getting all the kills, still managed to dictate the flow of the game.
If you're looking at score deficits, you also need to look at the kill/death distribution among all the players in the game, the team compositions, and the lane matchups, to make any reasonable conclusion about how well the players were matched up, and that's too much effort for any single person.
I'm not saying matchmaking is perfect, but your metric for what determines an "unbalanced" game is very lacking.
League is probably more snowbally due to the nature of the cost efficiency of items, the weaker roaming power and the lack of gold loss or smoke
GregLombardi
Member
I agree the metric is lacking and I have no problem with any of what you've said here except for this: statistics is simple math. Statistically, a wide range of score outside of a certain standard deviation generally indicates some kind of inherent bias with either the population (defined perhaps as the players here) or the defining variables. This presumes that the wide range of score happens as consistently as I am stating, and I don't really have the data to prove that. What it is essentially saying is that EITHER the system is biased, OR there is ALWAYS a dominant player in most games, or ALWAYS a miss match, or ALWAYS an amazing player playing support, or ALWAYS one of the scenarios or any other scenario one could come up with from above happening.
Neither of those statistical possibilities sits well with me when it comes to assessing a good matchmaking system because it means there is always a flaw or a variable that the matchmaking didn't assess.
So yeah, basically what we're both saying is that its not perfect. You're saying that I'm specifying one variable and that that specification is flawed, but I'm saying that even if we specify all variables we can think of, that still indicates that the matchmaking may have severe bias problems, especially with the consistency of scoring as it relates to a certain normalized distribution and the standard deviation from the mean of that normalized distribution.
And one more thing: You're not differentiating between significant kills and mop-up kills. The former are kills that actually matter. The latter happens when a game is all but won, but one side is struggling anyway and the other side is just goofing around/farming kills. That can also warp the scoreboard at the end of the game, even though the game might've been "lost" much earlier, when the scores were more even.
This isn't a very reliable measure because a game could very well be pretty even for a good 30 minutes (i.e. 20-15 kill score). And then one team gets a complete wipe, and while they are raxxing, they well dive a couple times and score a few more wipes. Game goes from even to 35-15 in the last 5-10 minutes of the match. Haly explains it better than I did, the final scoreboard is almost always inflated with quite a few mop up GG kills.
Also I don't even know why I'm arguing this since you don't even have access to any stats to know if you're claim is true or not...
GregLombardi
Member
Agreed. This is another mathematical problem that perhaps only Valve themselves could address via data mining. I guess I just wish they'd address this publicly at this point. I've spent 1,400 hours in a game that I am becoming disenchanted with because there is consistency in its flaws.
Keylime
ÏÎ¯Î»Ï á¼Î¾ÎµÏÎγλοÏÏον καί ÏεÏδολÏγον οá½Îº εἰÏÏν
There are way too many variables to try and track.
This game has too much depth to try and programmatically determine a true ranking that can be depended on with a high degree of certainty.
For what it's worth, the match making system that exists seems to be generally working fine. The vast majority of people that I see have a near 50% win-rate...which is really the best thing you can say about a matchmaking system, right?
This game has too much depth to try and programmatically determine a true ranking that can be depended on with a high degree of certainty.
For what it's worth, the match making system that exists seems to be generally working fine. The vast majority of people that I see have a near 50% win-rate...which is really the best thing you can say about a matchmaking system, right?
GregLombardi
Member
Yes, this is the one saving grace of the system that I must say -- seems extremely consistent. We should all, for our skill levels, have about 50/50 win loss rate.
Perhaps if there was someone with a PHD in statistics, they could prove that in fact the numerical score of the games shouldn't matter. I find it hard to believe right now though.
I agree that Valve has been pretty damn slow on the features front, but as for game balance, Valve doesn't handle that at all. Icefrog does all the balance changes, Valve simply implements them. On the community front, the guild feature should launch next week so that should make you happy.
The one thing I'm pissed about is the lack of a ranking system still. It's inexcusable in a game like dota, ffs just copy the starcraft/league system if you hate visible mmr so much.
They're working on all of these. I don't understand why you think these are somehow mutually exclusive. Icefrog and his beta testers typically work on balancing and the rest of Valve works on everything else as far as I'm aware. We're getting graphical updates to heroes every other week, we're getting guilds and other community features, we're getting tutorials... There are a lot in the works right now.
GregLombardi
Member
Ah okay. I didn't know this. That makes me even more annoyed haha! The balance changes are basically all that happens with every patch, though that last balance change patch was absolutely massive and I give Icefrog tons of credit for that.
CarbonatedFalcon
Member
The thing about concede is that even in games that I think are lost (or would be very difficult to come back in), you can still have fun playing the game practicing various skills. It's only a very small percentage of games that are something like 30-0 at 15 minutes where I'd rather move on to the next game immediately. Less than 5% of the time for sure.
GregLombardi
Member
I guess they're working on the beginning game more than the end game then. Look, I've played DOTA for like 10 years, or whenever it was that the mod originally came out. I still think, after all these years, there are certain gameplay / spell interactions / FUN things about the genre that can be refreshed. I guess I just don't see Valve doing enough of that. I will keep a watchful eye though.
Seriously dude. There's a whole list every week of updates that include all these things you mentioned.
The reason you can apply those kind of metrics to other team based competitive game, and not to DOTA2 (at least, not without sacrificing a lot of accuracy) is because DOTA is a game where much of the game is determined by the actions of the players AFTER they're already matched up. In your typical shooter, every player has more or less the same responsibility for the outcome of the game, with the same potential for agency, and the outcome is oftentimes numerically quantifiable. Thus, you can reduce every player to a number which can be used with other numbers to judge how likely they are to succeed vs another team.
In DOTA, most of that is not true. To judge whether a given game was balanced you have to:
1) Look at the individual skill of the players
2) Modifiy each player's skill by their role's responsibility
3) Look at the relative effectiveness of both sides of each lane
4) Look at the overall effectiveness of one comp vs another
5) Other outlier factors depending on hero choice/strategy
The problem is this: the only thing matchmaking can do is try to balance 1, when 2-5 are just as, if not more, important. The only solution to THIS problem is to force people to pick heroes/roles before they even enter matchmaking, and that's terrible.
Can you explain why you think evenly matched teams would have evenly matched scores? I see no reason why that should be the case. Dota is a game where kills beget more kills, and deaths beget more deaths. I've seen pro games where one team gets completely shut out, losing in 20 minutes with zero kills - but they still win the best two-out-of-three set. If the end score of every game is entirely reflective of skill of the players, how do you account for that?
Imagine if the rules of football were changed so that every time a team scores a touchdown, the yardage they require for a first down is reduced by one. Meanwhile, the yardage their opponents require for a first down is increased by one. That would lead to extremely lopsided scores between evenly-matched teams, and it's essentially the way Dota works.
GregLombardi
Member
Wow. That's a great idea though! What if instead of forcing people, this was a mode? That'd be super cool. Maybe not the picking of heroes thing -- but the role thing?
Eh. I've noticed a few of the massive changes in that patch not too long ago, but the other things fall on deaf eyes. I'm thinking more along the lines of what Riot Games has done with LoL. They really stepped up the game of changes to the genre, even if their game is fatally flawed in some ways.
I don't expect them to be perfectly matched, I just expect the scores to fall within a certain standard deviation over a massive data set. I feel 1,200 games ought to be enough to say -- hey wait, this BIG GAP thing is happening way too often. I guess I'm just looking for an explanation that justifies the big gap -- not necessarily saying that its the matchmaking that is explicitly causing it. You don't see such consistent, massive gaps in football do you?
GregLombardi
Member
Isn't CM not really in line with that mode idea though? You'd want people to be able to enter a queue as support, or carry, or whatever, and get paired in a team.
Exactly. Using kill-count as a metric makes zero sense when the last 5 minutes of a game can rack up 20 kills. This is why they provide a dozen metrics to give you a better picture of the game. Including networth/experience graphs. Which are by far the best measure since they have a timescale to them and not just a 1-dimensional snapshot that only tells you the outcome of 1 stat.
JakkuFurosu
Member
I can see it now.
Queues for support cause that would obviously be quicker then carry
Plays CM (either mandatory selection or voluntarily)
Plays as carry CM
GregLombardi
Member
Awww come on! This would be an amazing mode. You'd only have access to certain heroes based upon the role you choose. That's assuming they could come up with 5 roles or so, or perhaps 2 carries per team? I don't know.
I actually had a game long ago where we lost all our raxes around 55-60 min mark. The enemy had their 2nd tower mid tower in place and all of the 3rd tier intact.
We won the game around the 90 min mark. Madness.
Select role you want to play.
Select team strat you want to use.
When the game loads, you're all limited to the heroes that are most viable for your strat and you're all given pointers on who you should play and how you should play them.
Basically the game will captain for you.
OH GOD THIS IS TERRIBLE SOMEONE STOP ME
Select team strat you want to use.
When the game loads, you're all limited to the heroes that are most viable for your strat and you're all given pointers on who you should play and how you should play them.
Basically the game will captain for you.
OH GOD THIS IS TERRIBLE SOMEONE STOP ME
GregLombardi
Member
Would be the best learning tool ever too. I shall not stop you!
Heavy's Sandvich
Member
Were you playing solo?
I believe in learning through trial by fire. It worked for me in my day, there's no reason it shouldn't work for these whippersnapping upstarts.
This is already terrible. There's lots of fun in using a hero in a different role then just what he's normally played.
Terrible, yes. That would definitely stiffle creativity and innovation in the types of plays, combos, and builds we see.
No, you don't - and my second paragraph you quoted already stated why. In football, every time a team scores a touchdown, the game is reset with both teams back on even footing in terms of their ability to score future touchdowns. This is not how Dota works, at all. Dota is like football where every time a team scores a touchdown, the rules change to make future touchdowns easier for their team and harder for the opposing team. "Snowballing out of control" is impossible in football.
If you changed Dota so that both teams received the same gold and experience regardless of what happened during the game - no gold gain or loss for last hits, kills, towers, etc. - you'd see fairly even scores between evenly-matched teams. No one would want to play that game, though.
Maybe you should play LOL?
You are asking for things that would why DOTA is so awesome.
- Status