PVP revamp: defense wins championships

mischiefmaker

tl;dr: Make it easier to play defense.

I'd like to address three main concerns with the current PvP system:

1. Ratings fluctuate too much.
2. The clear best strategy for winning is to game the system by tanking.
3. Not very much team diversity (lots of OBW/Wolverine/Thor).

I should note here that shields have definitely helped #1, and had a nice side effect of giving people something to strategize about (when to shield), but they don't address #3 at all and they actually make #2 worse, because now the best strategy for winning is a combination of tanking and shielding at the right times; i.e., more metagaming, less actual gaming.

Phantron has eloquently laid out the fundamental issue: the PvP system uses an ELO-like rating, but ELO assumes that players of higher skill should win more frequently. This clearly isn't true: the single biggest determinant of who wins a match is simply who is the attacker. I propose changes that would make it easier, or at least possible, to play defense:

1. At the start of the tournament, before you enter, you choose a team to put on defense. This team can be changed every 8 hours or 10 losses, whichever comes first. The characters you put on defense are locked for offense. (If it's easier to implement, you could just have the team set for the duration of the tournament or be able to change it whenever, and not lock the characters.)

2. Any boosts the attacker brings are also given to the defense.

3. First turn is determined randomly (or, as in previous puzzle quests, could be tied to a character trait).

4. One chance to win. You may try multiple times but each loss results in a change to your rating.

5. Retaliations are unchanged. (It would be cool to be able to tailor your defense team for each retaliation, but that seems like way too much work.)

Matchmaking rating is also changed:

1. Each tournament has its own MMR, not shared with any other tournament.

2. You have a rating on offense, and a rating on defense. Each time you win on offense, your rating goes up. It goes up by less if one of your characters is downed during the fight. Each time you lose or skip, your rating goes down. On defense, your rating goes up if you win, your team is skipped, or you down one or more characters.

3. Your rating for the tournament is the sum of both ratings. Ties are broken by number of matches won.

4. MMR matches your offensive rating with others of similar defensive rating.

How does this address the issues we laid out originally?

1. Ratings should be more stable. It'll still be easier to play offense, but this should make more teams actually scary.

2. Eliminates tanking and correct shielding strategy as necessary for winning tournaments.

3. Vastly improves team diversity. This is especially true if your defense and offense require different characters. In turn, this encourages more diversity on offense, as players deal with different teams, which in turn rewards players with deeper rosters.

Additionally, this system offers some nice side benefits:

1. No need for grinding, playing at off-hours, and being present in the last hour. If ratings are stable, you can play when you want, your rating will stabilize at the right level, and you can go to bed or work or whatever. If you like grinding, you still can -- you still get iso plus the tiebreaker advantage -- but it's not necessary for doing well.

2. Increases and rewards strategic thinking. Right now it's all about who can build a team that wins fights the fastest, and who's willing to grind at off-hours and then throw up a shield at the right time. This system makes it so that the people who come up with the best teams win.

3. Encourages larger rosters, potentially with two copies of the same character with different builds. Maybe you'd rather play a 3/5/5 Patch on offense, but a 5/5/3 on defense. Now you can! Want to play Thor on offense and defense? Get two.

4. Improves character playability and expands the design space. Currently we all evaluate a character based on offensive capabilities. It would be cool to think about characters that make defense hard, or annoying, or time-consuming — I bet Daken, Bullseye, Juggernaut, and a few others would get a lot more playtime with no changes required.

5. Shields don't feel like as much of a pay-to-win mechanic (I realize they're not one, but to a player who just got knocked around for -200 points, it sure feels icky to be asked for hero points for protection).

6. It's really fun to come back to the game and find that you gained a whole bunch of rating points. Way better than coming back to find out how much you lost.

Thoughts? I'm particularly interested in hearing whether the changes to defense would be enough to make attacking difficult.

Unknown

Right now the game still overwhelmingly favors flawless victories so defense is still irrelevent. You can't exactly make it so that you'd lose on attack because then people would feel like they've no reason to continue playing if you actually consistently lose.

What people don't understand is that it's actually pretty trivial for the AI to beat you in this game if it is not artifically throttled (one ability a turn, always go last, cannot choose target). If you're given the AI's handicap to play against a slightly tweaked AI you would end up losing most of the time too, and that wouldn't be fun to play.

In this game, the primary limitation is supposed to be your health packs. Let us imagine healing abilities are simply removed from the game. Now, what happens if your character starts the next fight at half health? You can actually lose the fight and then you'd need to spend 3 health packs, and that's more than half of your normal reserves. Note that this happens a lot in PvE, as the super powered level 240 guys can often down a whole team even at full health let alone one that started out half dead, so we can see a pretty clear separation on how can most effectively grind out those missions consistently come out ahead overall.

Of course, on PvP it wouldn't make much sense to give the defense the same boosts as the level 240 PvE enemies get. There's really no way to make the AI more challenging without immediately getting a ton of complaints. If the AI can even choose its targets you'd pretty much have no use for tank type characters, and a character like OBW or Spiderman will be pretty much worthless (will always be killed first).

For the PvP to make sense, Spiderman have to be changed because currently he can pretty much bring anyone who isn't dead back to full health, on top of easily stunlocking 1 or even 2 opponents forever. OBW is hard for me to grasp. I do not think her heals can keep up with the stronger damaging abilities (would need 2 heals to just counter one Thunder Strike), and she's a huge liability on defense. It seems like that playing style is balanced because it's relatively safe but you also take a lot of extra beating on defense.

That said, this game is never going to be about not winning most of your attacks. That simply wouldn't be fun to play. The best player is the one that can win the most fights quickly and continously. If you have Spiderman's safety net removed, it'd definitely take a lot of skill (and luck and boosts) to keep on winning without running out of health packs. In the Smash Hit tournament I have won taking as few as 1000 damage (no boosts), and I've also taken more than 10K with a combination of bad playing + bad luck. Right now, these two are pretty much the same because as long as Spiderman didn't die your character still ends up with all HPs (Hulk might be missing some but not like it matters). The guy who can consistently win while taking minimal damage should be the best player, even if everyone still always wins on offense.

mischiefmaker

You bring up a good point; I considered that the developers purposely made it easier to win, because constantly losing isn't fun.

But I don't think this is actually an issue. If you come up against a team that is likely to beat you, you just skip it and get a new team. You still wouldn't consistently lose, you'd just skip more teams. Those teams would get rewarded for being scary enough to cause you to skip them.

Your PvE comparison is actually entirely apt. How do we determine who wins PvE? In the current event, near as I can tell, it's 1) ability to beat all the levels, and 2) willingness/ability to grind the hardest ones multiple times. I think this is a pretty fantastic system -- it's intuitive and it gives players the right motivations. I'm proposing we make PvP the same: rank people first by their ability to beat other teams, and second by grinding.

The guy who can consistently win while taking minimal damage should be the best player, even if everyone still always wins on offense.

The problem is that right now EVERYONE can win while taking minimal damage. This isn't going to change with Spider Man's eventual nerf unless they also change OBW -- if you don't think so, ask anyone who's beating the 240 fights with OBW. So the game becomes, as I said, primarily about understanding MMR, tanking, and grinding. I mean, hey, that's great for those of us who understand MMR, but it's not very intuitive and quite frustrating for new players, and I think we can do better.

Unknown

Phantron has said most of what I wanted to say but I will give my input as well.
First of all making it harder to win will make the game less fun, a counter to this would be to increase character heal speed and increase health pack regen limit to 10.
Increasing team diversity, I would like to see a "trait" system implamented, each character would have a trait, for example rags may get a 20% damage boost against thor, wolvie could do more damage when below 20% health. If this was implamented it would add to more tactics in defence thus more diversity in character choice.
MMR should be calculated from last 5 tournaments/sub events to make more concistant match making, reduce tanking.
Skipping oponents should be limited or removed and instead your have a choice of say 7 opponents and after each few fights or progression reward they refresh.
Points lost should be half points gained on attack and defence.
Spiderman is going to be nerfed, so wait and see what happens with him.

Unknown

OBW's power is tied to Wolverine. It's not unusual to have 1K worth of strike tiles up from a decent board (and trivial if you boost). With such an insane amount of strike tiles, you can basically say strike tiles determine your total damage, and OBW increases that by 50% if she gets 3 of your team's colors. So yes, such potent offense obviously makes her relatively weak heals quite potent since the enemy team isn't going to be around very long. I'm quite certain Wolverine will be nerfed hard, though, so I'm reserving judgment on OBW until Wolverine is balanced.

mischiefmaker

I used to think so too, but that's only half right. OBW's offensive power is tied not to Wolverine but to strike tiles. Since Patch and Punisher both produce very strong strike tiles relatively cheaply, OBW's offensive power doesn't go away with a Wolverine nerf, though it does get a tad slower.

Defensively, her power is due in part to the heal and the other part to Recon. If you're stealing all the opponent's AP, you only have to heal match damage.

Unknown

mischiefmaker wrote:

I used to think so too, but that's only half right. OBW's offensive power is tied not to Wolverine but to strike tiles. Since Patch and Punisher both produce very strong strike tiles relatively cheaply, OBW's offensive power doesn't go away with a Wolverine nerf, though it does get a tad slower.

Defensively, her power is due in part to the heal and the other part to Recon. If you're stealing all the opponent's AP, you only have to heal match damage.

Wolverine's strike tiles are way stronger in general. Patch gives strike tiles to both sides. Punisher strike tiles is much slower to generate, not to mention the target random 3X3 square means you might accidentally blow them away. It's also considerably harder to protect The Punisher's strike tiles since they're placed randomly. In the case of Feral Claws, you simply match all the red you could before you place them and they'll usually last a good while.

Aggressive Recon is no doubt very strong, but I don't think OBW will be that potent without Feral Claws easily putting down 300 or so worth of strike tiles before you know it (6 red + 6 green = 356 damage worth of strike tiles) so I need to see what happens after Wolverine is nerfed first.

The Ares tournament is a good example of what meaningful defense could look like. You're looking at potent, but not dominating offense. You can definitely win, but you're likely to take significant damage that is at least out of OBW's league to heal. Spiderman can still handle it, but he's obviously overpowered.

Unknown

If they put the progression reward back to 1400 level, I'd be happy to have a harder defence team.
At the moment, its serious grind already, no easy win for offence, the progression reward is not even obtainable.
In the Ares buffed up torment, its almost certain I need to use a health pack every 3-5 matches. Of course, I don't have a decent leveled Spiderman. But I think majority of playerbase does not have one either.

They already give the buffed up character 12k + hp and the ability to 1 shot any nontank 2 star characters, I am sure I don't want more handicaps.

Unknown

xinyucao wrote:

If they put the progression reward back to 1400 level, I'd be happy to have a harder defence team.
At the moment, its serious grind already, no easy win for offence, the progression reward is not even obtainable.
In the Ares buffed up torment, its almost certain I need to use a health pack every 3-5 matches. Of course, I don't have a decent leveled Spiderman. But I think majority of playerbase does not have one either.

They already give the buffed up character 12k + hp and the ability to 1 shot any nontank 2 star characters, I am sure I don't want more handicaps.

I don't think it's a bad thing that you need to use health packs. Keep in mind if people regularly do need to heal up, that also cuts down the amount of attacks on you and it'd also actually discourage people from just fearlessly attacking the highest point target. Obviously, none of that works while Spiderman is a one man stun/heal machine, but I assume that can't last forever.

Unknown

I like the idea of buying AP defensive boosts to offset the advantages of the player controlled team.

The AI is a lot harder to handle when they are spotted a +3 (or more) AP boost.

Unknown

Interesting thoughts, but it went to a completely fictional direction. IOW based on a similar flaw as the current system that claimed ELO system is proven to work for it.

You claim: " Increases and rewards strategic thinking. Right now it's all about who can build a team that wins fights the fastest, and who's willing to grind at off-hours and then throw up a shield at the right time. This system makes it so that the people who come up with the best teams win."

Now think again, suppose it is true. Suppose we had MtG tournaments redone to played by the AI for you. All you delegate is the deck. You let the AI play them, ignore effort and everything, just let the best composition win. The NETDECK bonanza will be ultimate: once the best team got revealed everyone will put that up. That one. Well, if you have it obviously. And here the variance is so low it would emerge just naturally, even by simple chance/elimination.

You can strip some of the numerous advantages of the attacker, the AI still plays your team crippled. You can put in good combination of heroes but the AI has no clue whatsoever how to play them. How to use one to support the other. How to apply strategy or tactics to the game.

Why MtG works even with the general presence of netdecks and all listings and evaluaitons going public? Because it is played by people, directly. To play a deck you need a lots of brains and stamina, must playtest it a lot and even know the metagame. Otherwise you may have the most broken cards and still will likely lose consustently. And the same very skilled people do not play with all kinds of decks from the winner pool, but chose one fitting their nature.

IOW the "best team" as a general concept does not exist -- just like the abstract best car and settings is close to a myth in race: best is what works with the driver. You can't separate them without ruining.

What eventually leads us back to the base: this game, with AI playing the opposition is really just good for PvE style, and the winners are supposed to be those playing with least consumption of time/health and applying more play. A rating based PvP will only work differently if it gets true PvP hooking up players against each other. For that what you write would mostly work though be irrelevant.

I just fail to see how you arrived to that your system would improve diversity.

This doesn't mean all the ideas are bad, or that the current ways could not be improved a big deal. Yes, being able to use a defender team would be better than the current system. Much. Due to the very fact that the AI does not play your offensive strategy on defense -- the different team helps by taking AI's strength and weakness into account.

I see no point to argue individual points when the whole concept was mis-based, I suggest to think put back the reality of AI's consequences and redraw the system based on that. (btw It's quite possible that a good/fair system for this just does not exist in the first place... only choice from differently broken ones) By all means work to think up something better, that must be possible.

My first improvement attempt would change just a few things: remove ability to chose target (so same jump-ahead works as for you), and rework the point system completely. How I'm not yet clear, I had an initial idea that counts wins and losses, and value of each is always dependent on the opponents' win/lose counts. (i.e wins over mostly winners worth a lot and wins on bagman teams worth 0 in the end.) But it has a plenty of unresolved issues. until then I think just cutting the loss points proportionally with the remaining time and scrapping the shield system would make a better play, and pairing is stiill open.

Unknown

Given the complexity of this game is next to nothing (there are usually like 8 or so moves you can match, and usually one or two choices as to which moves to use), you can simply do this by trial and error, as in the AI can simply run the game 100 times behind your back using current parameters using a random selection of abilities and just pick the one that's got the best outcome. For example if you have Thunderous Clap and Feral Claws, all it has to is quickly play the current game (against itself) 100 times using random combination of the abilities (e.g. 3 Feral Claws or 1 Thunderous Clap). It will almost certainly be able to trivially determine that it should never use Thunderous Clap over Feral Claws. It'll likely notice that games where it is dominant it tends to have a large amount of red APs. Of course, if you're lazy you can simply tell it that Feral Claws is good.

The whole point of having restriction like no repeat ability per turn is that while it feels pretty cool to combo stuff up, it will be considerably less cool when the computer does the same to you. They are quite capable of lighting you up with pre nerf Thunderclap -> Godlike, or more complicated stuff like Mistress of the Storm -> Thorned Rose -> Lightning Strike -> whatever, and there's nothing fun about being on the receiving end of such broken combos. I mean honestly, do you really think it's that hard to program something that just keeps on spam pre nref Thunderclap? It doesn't do that because it'd probably result in a lot of rage quits if you end up losing on your second turn because the AI just spams Thunderclap over and over like a human would.

Unknown

Sure the AI could go pretty far, though precise calculation of outcomes for several turns are limited because of the random tiles that arrive after a match.

I'm on the fence with an AI that would do the best within turn -- tend to agree to have some limits like 1 ability per use. However think the current damping is way on the bad side ruining the play. I'd make it mandatory at least to:

1. always take a possible opportunity that results in repeated turn (i.e 5-match). [the only motivation to not do that could be grabbing the last points of winner AP for the next turn where you don't care some critical damage...]

2. never pick the worse ability from alternatives, like Yelena black over Venom black doing less of the same for more AP.

Looking at these I might be relieved that I survived after all, but it also feels like it's not worth playing at all, the opponent is throwing the match.

And I'd seriously consider to take AP gained within turn to use -- currently it seem to decide what to do at start of turn and go ahead. AP gained from ability including rose gets used only next turn. Doesn't feel right either.

An example for game with limited action space and AI that can think way ahead through comprehensive play of all variants: Spectromancer. It is damn hard to win and you face a big deal of photo finish losses. Sure you can set AI do be dumber -- I never did that.

Here a really strong AI would change many factors and likely remove the chance to defeat the big evolved teams with * storm and widow and brains. But it could be compensated by different means. And until then these deliberately bad AI must be considered as factor in tournament design.