In a balanced game, how soon can you predict the winner with 75% accuracy on average?
moultano
Creator of ns_shiva. Join Date: 2002-12-14 Member: 10806Members, NS1 Playtester, Contributor, Constellation, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Gold, NS2 Community Developer, Pistachionauts
Creator of ns_shiva. Join Date: 2002-12-14 Member: 10806Members, NS1 Playtester, Contributor, Constellation, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Gold, NS2 Community Developer, Pistachionauts
One of the parameters of hive controls how much the beginning of the game counts more than the end. We're having some debate on how to set that parameter, so I thought I'd ask the community.
In a game that's more or less balanced (everyone is pretty happy with the teams at the outset, and you can't predict who will win just looking at the teams) how soon into the game can you predict the outcome with 75% accuracy on average?
In a game that's more or less balanced (everyone is pretty happy with the teams at the outset, and you can't predict who will win just looking at the teams) how soon into the game can you predict the outcome with 75% accuracy on average?
Comments
Essentially the balance of ns2 is like a seesaw, however if you can maintain enough of an economic advantage over the other team then the balance essentially stays in your favor.
Of course it is still possible to throw the game by under performing in team fights which may give the losing team the chance to turn things around.
Given that a somewhat balanced public game I would say it is typically very clear by the 4 minute mark.
This is also very different in a competitive setting.
I wouldn't quite agree with this. It's easier to predict when marines are doing badly, but if marines are doing well there's still that risk of them kindof fizzling out in midgame, allowing aliens to take over with Oni late game.
Accuracy of your prediction. You guess which team will win, and you are right 75% of the time.
What would you have chosen instead?
If the question was just "on average how soon can you predict" I'd say despite a huge variance dependent on said balance, it'd be mid game just after fades come out.
But the issue is two fold.
Current balance dictates an incredible advantage for aliens if they last until the onos explosion, so the situation just before that will dictate the chances of surviving it. Or conversely, the weak alien early game. (We have to account for this imbalance even when discussing perfectly balanced teams)
The second thing is that even if there weren't inherently imbalanced portions of the round, balanced teams are just *incredibly rare* in pubs and when it occurs ... Typically, the tell tale signs are a healthy amount of back and forth / countering, until the very end where the victor is decided by a mistake that the other team makes, hence my answer to this poll that appears to be discussing the topic in a vacuum instead of practicality.
Typically though, teams are imbalanced, and therefore the outcome is easily predictable. From as early as the first 2 minutes sometimes
This is a tricky epistemological question, but the way I'm defining "teams are balanced" here is, "teams are balanced to the best of anyone's ability knowing the players and how good they are before the game." It may turn out after a few minutes that those teams aren't actually balanced and it's apparent to everyone, but for the sake of the question, limit yourself to all the knowledge you can have before the game. So knowing the players, their skills, whatever else, is available information to you, but not anything about how they will actually play during this particular game.
The background for this question is that I'm trying to eliminate any incentive for players to leave early if they know the game is lost. If hive skills are accurate, then no matter how imbalanced the teams are, it's in your interest to keep playing to maximize your chance of winning, because the upside of how much your skill will increase balances out its low probability. The problem comes from how hive treats players that are leaving and joining. The player's impact on hive's prediction gets "locked in" at an exponential rate, but if hive locks it in slower than the players realize it, they still have an incentive to quit. I'd like to eliminate that, so I need some way of estimating how confident hive should be in the outcome of the game at any given time in the best case where its predictions are accurate.
I think that the incredible shift in chances of winning between the teams depending on the round time is going to make the answer you're looking for very difficult.
In other words, I think we'd have to address the weak early alien game and the weak late marine game before you could find a time that is reliable.
Because as it is now, there's the strong likelihood that even with the most balanced matchup assembled through an imaginary committee, the amount of potential failure points in the round where one team inherently has an advantage over the other, is going to upset your prediction. Meaning said theoretical dream matchup could result in a predictable game by the 2 minute mark just as easily as it could by the 10 minute mark - the variance is just too great. (Team win % by round time graphs highlight this well!)
That all being said, from a psychological point of view, I see players "hanging in there" until around fades (~7 min mark) where it's clear whether marines can sustain through it until the next power shift (Onos) or whether fades can bring aliens back from losing territory. Good fades are the first real power shift for aliens that can quickly change the tide of a round, not having one means you're holding on and praying for the Onos wave instead. Power shift after power shift..
So given your clarification, I'd say the 8 minute mark option. (i dont think I can re vote?)
But I just don't think any time is going to be reliable enough right now..
Why, what did you expect?
What data is legit for that prediction? Me, watching a game for x minutes, and then making a call who's going to win based upon my observation & what has happened on the field so far? That's how I voted.
Balanced game - still the same, as if its a balanced pub game, is it balanced based on 1 good player on 1 team with a high elo and a bunch of rookies.
on the other team a few players around the same elo not as high as the other player, alot better than the rookies. ?
It really all depends, even the most balanced looking games with the same ELO for both teams, most games can be decided in the 1st minute - aliens lose 1 rt / marines lose 1 rt. May not be then and there, but you can pretty much get it right 60-80% of the time.
Only things you can know from before the game, so assume you know how good all the players are, and at the start of the game, the teams look OK to you.
Like, there's an inherant bias in how people are going to answer this question, and I'd argue that it's quite a significant bias.
That would indeed be interesting, but considering the length of ns2 games (up to 40 minutes, sometimes even longer) and the variety of games there are (long slogs, the ones that end quickly, etc) it would require a great amount of work to collect and evaluate a representative amount of games.
Couple of scenarios that I usually look at...
To win marines must hold Aliens to 2 (or less) res nodes, by 5 minutes I usually know if aliens will be able to hold 3 or more.
I wonder how far you could get by just using simple statistics at a given time. Look at resources collected so far, kills so far, active RTs. If a dumb regression model based on things like that can work with the desired accuracy at the n minute mark, that's sort of a lower bound on how well a human could do it.
In unbalanced games then you can get an idea early on,usually with the first few minutes, but still all it takes is a sneaky player to drop a pg/tunnel and rush a tech point. Often overconfidence by the stacked team means not bothering with multiple cc/hive.
Ok, that aligns with my vote. So, in this fictional scenario, I think the teams are balanced, which means I expect to see a swing of momentum back and forth between Marines and Aliens as tech and lifeforms progress. If exactly this happens - then it's a super fair game and I couldn't possibly predict the outcome; but if it does *not* happen (Aliens being able to pressure Marines before lerks, or Marines not being challenged at all after lerks) then I think after 4 minutes the snowball is real enough that - with a 75% confidence - I can make an educated guess on how this will end.
One big disturbing factor there however is 'average skill' level. The higher it is, the more intense the overall metagame will be. On the other hand - spectate a match after a steam sale where players aren't aware of the metagame and play passively - it's random skirmish while both side tech up to end-game and from that moment on it's pure luck and chance who runs into whom.
Anyone voting below 3min is facetious... *looks at @SantaClaws accusingly* x)
In a balanced game, how soon can you predict the length of the game, with 75% accuracy on average?
A longer game implies more back-and-forth to the gameplay; which in turn, makes it harder to predict who will ultimately win.
@moultano it might be easier to make a time bounded prediction. Instead of asking "Who's going to win, after n minutes?', perhaps ask "Who's going to be winning at 10 minutes?".