New Updates PVP Algorithm


#21

Matchmaking does seem worse now.

I used to avoid certain team comps because it would put a 6k support with a 12k tank on my roster and matchmaking would always give me a big penalty for that.

I tried that same composition again this week and never got punished, many of my opponents had perfectly even teams that were very close in power and yet they’d get matched against my huge 6k disparity like it was nothing.



#22


he is punished as you can see here by getting 4k lower power than him hmmm fixed my ass


#23

They should put in a simple solution at higher levels. Once you get into the 50k range, team difference shouldn’t exceed 5k. At 60k it shouldn’t exceed 3, and at 70+ it shouldn’t exceed 1.5k.

This would help account for the fact there is no real min/maxing going on after 50k+ power, and people being punished for it aren’t even min/maxing, they are using lower heroes they have left at gold or silver but not even gaining benefits from it, they are just using weaker heroes AND facing higher power teams.


#24

The real problem with that is that hero power isn’t driven at all by how valuable their abilities are. Nightingale’s power is ridiculously low because her value is almost entirely due to her abilities. Look at Rygel’s heroes: Looking at Rygel’s heroes for a fair reference point… Nightingale, a top-3 viability hero, has a power of 13,716 while Baron, currently a complete waste of space, has a power of 17,565. That’s equal rank, equal star. It makes no sense.


#25

I have brought that up in other threads and totally agree. That’s a whole nother issue. Of course you also have cases like this where people still are trying to gain the system:

His Mandrake is almost 3k less power than mine since he kept him gold. Mandrake’s plat skill is garbage so the actual power difference isn’t nearly as big as it seems. Then you have his Nightingale who is 2 plat bars above mine (yet only 40 power higher? Didn’t level some skills there…), his Panzer who is a star stronger, and Min who is a star stronger than my Ifrit (yet again, 700 less power than my Ifrit. My 7* Min is 9800 power, yet his who is a * stronger is somehow 300 less power…). All the while his Razorback is pulling the same gold crap. 9 star yet 1100 power less than my 8* plat Heimlock.

His team is actually quite a bit stronger, but on paper the numbers look weaker. The small 3k power “advantage” I had wasn’t even really an advantage. His heroes are being artificially kept lower in power by not wasting points on crappy skills, his team would really be in the 66-70k power range if he had leveled them all.


#26

You cann not compare the heroes one by one. He got a 6,5k power difference in his team, you got a 5k power difference in your team, so I do not see any problem with this matchup.


#27


#28

the thing is, like @WalleWu said… you can’t compare heroes one by one… a Heimlock is a completely different role than Razorback, and same goes with Ifrit vs Min. Those two changes to the team makes the actual effectiveness of your teams very different. Since His team is much more disruption oriented (especially if you let Min live long enough to start getting her abilities going… because she will steal your cooldowns give them to her allies, she will silence your heroes and she will destroy your ammo… not to mention the fact that her passive brings up a team shield whenever one of your heroes uses an ability for the first time. While Razorback can taunt, deals decent damage for a front liner, and has some retaliation type passives.

Also speaking of the skills not being upgraded… the Devs have mentioned that you actually get punished in your matchmaking for not having your skills leveled up (I think Omnipotent or Huginn said in VIP chat that your skills are always counted as a certain minimum level, based on the hero’s level) when the matchmaker is doing it’s thing.


#29

Except you totally failed to touch on all my other points. I compared the heroes one by one because a few of our heroes are the same (Panzer, Mandrake, and Night). I was more pointing out, though, that his team is actually MUCH stronger. His power is just lower simply because he is not leveling up certain skills, but it’s not telling the whole story.

When you dissect it you will see his Nightingale has 2 plat bars on mine but is only 36 points higher? If he had all his skills leveled, she should be roughly 800 points higher. His Mandrake is probably the worst offender. Both his and mine are 10*, except his is +1 gold while mine is +1 plat. Because of this, and the fact he probably didn’t level his plat skill at all since it sucks so bad, my Mandrake is 3k points higher than his, which basically the negates the small team power advantage I had right there. His Min is doing the same thing. My Min is currently 7* and +0 plat. Her power level is 9813. His is the same plat and 1* over mine, yet 300 power lower? I am sure he didn’t level her skills to artificially keep her power lower.

I’m not seeing it at all in that match I took a screenshot of. My team is 3200 points “higher”, but as WalleWu points out his gap in hero power is bigger than mine. His gap is 6.5k compared to my 5k. That should account for me having a higher team power. It doesn’t though.

That small increase in team power I had over him was actually totally negated by him scamming the system and not leveling all his skills. As I said, his Night should be at least +800 in power, his Min +1k, and his Mandrake +3k. That alone is almost 5k more power he should have tacked no to his team, but doesn’t because he skipped leveling weaker skills that hardly bring anything to the table anyways. Not having Mandrake plat and skipping skills hardly makes up for all that power rating it would have added to his team.


#30

Yup, this is fine. Totally fine…


#31

This would be cool but you would basically be playing 95% bots if that was the case. Some of the reasons why players like @RTTHINGS have screenshots like the one above is because of the lack of players who are actually playing at that range. Limiting matchmaking at higher levels would just further compound frustrations or just result in giving those players another PvE game mode to play.

From seeing a lot of these screenshots on this thread, it seems that there is a correlation between what star a hero should be at each rank. One team I saw there was a 10 and 9 star gold that got matched up against 7 star plats. This may equate to a trade off of more stars/low rank vs les stars/high rank even though inherently the platinum hero will have more power. Screenshots like @Sogui may help that observation. His opponent decided to make a 5 and a 6 star platinum while he is playing with a 7 and 8 star gold. He wins that matchup because the extra raw stats provided from the stars is much more valuable then the skill that is gained at platnium.

In short, Lots of stars + Low rank = Punish. Low amount of stars + High rank = Punish. It may be beneficial to rank up heroes relative to their stars. Would be interesting too see some tests with people running like 3 star plat teams and seeing the types of matchups they get.


#32

Just felt the urge to share this :joy:


#33

I think the new equipment changes made this much more difficult to do. I can level a bronze hero to plat in around 2 weeks if I play a decent amount, get the daily log in sta, and don’t spending any gold. I have almost 20 plat heroes. On the other hand it still takes months for me to get a low star hero to 10 star. I only have 3 that are 10* after playing fairly regularly for over 250 days.

Now granted I haven’t really been focusing on any heroes, I have 3 more that are almost 10*, and I have been mostly sinking heronium into the faction of the month heroes. All the same though, two weeks to rank to plat vs 2 months for 10* is a huge difference and makes it difficult to do them in conjunction.

This is but another problem with HH system. Even if you want to level heroes up simultaneously, or their stars and bars at the same rate, the game doesn’t lend itself to doing so. Combined with some skills being worthless so players skip them, heroes having widely varying power levels even at the same level/bars/stars (Gale vs Beck), and it just makes for a mess of a match making system.


#34

We provide reasonable leeway for normal progression. Any penalties for stars, grades, and levels are applied separately; Having everyone the same grade and levels but minor differences in stars applies a much lower penalty, than if all 3 metrics show considerable deviation.

Running a perfectly balanced team will not guarrantee an easy win. Even in an inequal power match, we still want to provide both players a chance at winning. We are not trying to force a loss for either player.


#35

What do you mean by grades?


#36

Why was this changed from a power disparity metric?

In the feedback threads on the topic I threw out the straightforward math that showed how you could stack a 9* plat Panzer at lvl 75 with a team of 7* golds and there wouldn’t be a “significant deviation” in stars or grades but Panzer would be sitting at 80% more power than the rest of the team. Said team would be sitting at low 40k power but the average opponent would be seeing their heroes die in the opening seconds to a 9* plat panzer.

Now I’m confused, would I actually be rewarded for increasing the power disparity in my team if it involved adding 20 levels to my Mandrake in this screenshot? Are teams with a 10k plat panzer and a 6k gold Operator really given a smaller matchmaking penalty than a 8k Mandrake and 6k Heimlock because the Mandrake has less levels?


#37

I was referring to penalties based on differences in upgrades.

Differences in total power between heroes affects matchmaking differently and applies its own separate penalties.


#38

I feel like the matchmaking should be decided by HOW meta both team is:

For example, 35k power team: Panzer, Nightingale, Heimlock, Mandrakes, Ifrit
should be matched with a team with:

  1. Team who has Panzer, that has same power as opponent’s Panzer,
  2. 45k power team consisting of meta poor heroes: Fischer, Oro, Francoise, Surge, Matador etc.
  3. Team who has 1 dps 4 support/healer (like Dogface, Nightingale, Mandrake, Flatline, Ifrit)

#39

This would be even more restrictive then my above suggestion of making smaller power penalties as you get into the 50/60/70k ranges. Finding near mirror matches would be very difficult. This would also require HH to go back and label every hero as good or bad meta heroes so they can be tagged when matching.

It would be better for HH to just go back and go through every hero with a fine tooth comb. The skills that suck like Mandrake’s plat should only increase power by 1-2 points each. Skillls like Nightingale’s heal should have a much bigger impact. This would help make the meta heroes have higher power AND make it so people would stop trying to cheat the system by not leveling up certain skills or ranks that are crappy to keep their power artificially lower.


#40

Perfect response. (Commented because a heart wouldn’t suffice)

One thing that I will note, however, is that while I also feel that the best solution is for devs to go through the heroes and re-calibrate how much they value each hero/skill in their calculations, there is a small issue with that. Aside from the time required, there is the fact that the game is ‘living’. This is an issue because, as I see it, the relative value of a given hero changes with the addition of each new hero. Typically this is minor if anything, but a good example of how the impact can be significant is Anvil The introduction of Anvil means there is now a strong counter to heroes who use a shield skill. Because there is now an additional hero to counter shield-heroes (and a very strong counter at that) that didn’t exist before, shield-based heroes are now less of an asset, than they were pre-Anvil. To take it a step further, let’s say, next month, a shield hero is added, Anvil’s ‘real’ value will go up since he is a strong counter to that type of hero. And so it continues.

If we agree that additional heroes change the relative value of existing heroes, there are two solutions that I can think of to ensure hero values stay calibrated after a hypothetical re-examination:
Option 1 (manual): schedule a manual recalibration (ie fine-tooth-combing the value of each skill) every X heroes or Y months.
Option 2 (automated) radical proposal but, bear with me, hero values are determined via a market-based solution (not market as in ‘store’ but market as in ‘stock market’). At it’s core, the solution would revolve around the belief that people use strong heroes and ignore weak heroes. If a given hero is used by a lot of players, that signifies that the hero is valued by the player base and, as such, the power value ascribed to that hero increases. If a given hero is never used by players, then its value is seen as low and, you guessed it, the nominal power value of that hero goes down.
Now, I will poke a few holes in this. (1) if this is a truly ‘live’ market, it’s confusing to users why the ‘value’ of their hero is fluctuating when they aren’t doing anything. To that, I say, don’t change values dynamically, do it once a month based on the previous two weeks of use in non-faction events (2) because upgrading a hero isn’t free (in the $, items, or frags sense) it’s easier said than done to just switch to less utilized heroes–they have probably been neglected and can’t be used off the bat. Fair, I don’t have a good response to that. (3) it’s too much work to implement. I actually don’t think it is that outsized of a project given it’s either this or endless recalibration but I also acknowledge that it’s a fair point.

So, now that this post has turned into a proposal of sorts, what other issues exist and/or how could part of this proposal be incorporated into an alternative solution? :slight_smile: