You guys really need to lower the amount of chains the computer can get

they started happening way before x4 speed though, I can look it up for you… there was a huge thread on it…

Here is the thread: So how is everyone experiencing the console cascades?

@Mithran did a ton of research, and unearthed some unwanted behavior, which got fixed as posted by @Alpheon.

Unfortunately, the “balancing the adjustment further” never happened, and the game remained in it’s current state.

@Delinquent I was being sarcastic. Try laughing some time.

This was actually post Unity port for PC, so we did have 4x speed. However, the effect was not associated with 4x speed (I did plenty of testing to determine that). What was happening was they had some new code for spawners that was intended to start making them “miss” after a bunch of consecutive casts had a variable inverted. This code was always present for Adobe, but was rewritten and introduced for the first time on Unity when the PC port happened. Basically, if you took an extra turn of any kind before casting a spawner, they would cluster ridiculously. It happened for both sides, not just the AI.

The devs have gone on record multiple times that the spawner manipulation code is only supposed to take effect after the fifth spawner cast on the single turn, but I am still not convinced this is the case. What we seem to have now is a direct inversion of the “new” spawner manipulation code - where taking extra turns seems to lower spawner reliability, and I’m getting much more connected casts out of the first cast on a turn. This one is tricky to prove, though, since logic dictates you would get more first casts to successfully land because you can’t get a second cast in the same turn without the first one getting an extra turn. However, having used things like Wight extensively when I did my second account, connecting an extra turn with a 7 spawner used to be a very rare experience regardless of board state, but is a lot more common now. Its not broken reliable like when it was, well, broken, and the phenomenon does function in a similar manner for both the player and the AI, so I don’t see any AI bias here, but despite the huge public perception that gem spawning troops were nerfed into the ground when the auto-linking was fixed, they are still much stronger in my eyes than they were on Adobe.

Big cascade are still going on, also on both sides, which is a seperate issue I can only assume is related to the pRNG on Unity. However, it was also confirmed at one point that the AI cascade combo breaker that was supposed to be on “low” for PvP, thus not limiting AI cascades that much, was actually always on “high” (meaning Adobe cascades for AI were much more limited). Essentially, the game has been “cheating” more heavily the player’s favor for years on Adobe, and they essentially removed a big chunk of this extra anti-AI cascade manipulation. This was actually only brought up recently, but it explains a whole lot. Basically, the AI does get “increased luck” since Unity, but only because its luck is not as artificially limited as it was on Adobe - they never get luck boosted above what is actually “fair random”. Combined with Unity’s “streakier cascades on both sides” and the insane mana ramp from blob matches and that is how we got to where we are today. Unfortunately, it looks like both blob matches mana calculations and streaky cascades are here to stay, since they have taken to attempting to balance troops around them existing rather than fixing them directly (match trait nerf, troll/kraken nerf, etc).

In summary:

  • Gem spawners were and still are manipulated (both sides, no apparent AI bias) and first casts seem to still be more likely to connect than they ever were on Adobe

  • Cascades are bigger on both sides for Unity than on Adobe.

  • The AI combo breaker that limits AI cascades (and only AI cascades) exists for both Adobe and Unity, but does much less limiting on PvP and Guild Wars in Unity

  • Mana gathering is higher than it has ever been, limiting the opportunities for turn overs when not having mana to continue a spell chain.

  • Balance changes are already underway for existing troops (trolls, kraken) as well as the future design of some spoiler troops (35 mana cost mythic in the files, but this may change) in an attempt to work around the engine issues, so mana ramp/cascades are not likely to be further addressed directly.

17 Likes

OK, hmm, I wanted to say some things but Mithran’s post makes me not want to go over them. I’ll try to be more summarized than my plan was. (I am a dev too, and write long forum posts during long builds.)

I was going to explain how a “fair” RNG allows all the things we hate, and the linked videos show most experienced game developers realize players want an RNG slightly rigged in their favor. But if you screw this up, the player knows you’re “cheating” and it ruins the effect. So it’s a balance.

From there I’d argue the best GoW RNG would be crafty.

  • It would try to set up mo4s for me on occasion, and when generating gems for me perhaps it would bias the first roll towards “favorable” cascades with high probability.
  • If I haven’t had a free turn for some period, it would try to rig one for me.
  • It would with somewhat high probability try to “screw” the defender over if they’ve had “too many” free turns over some period.
  • It would not give the “help” to the gem generators on the defending team.
  • It might choose to fiddle with the rates of things like Devour either way, too.

What it really needs to do is make sure if anyone does a widespread data-gathering mission, it’s most likely even “small” sample sets will reveal a bias in my favor, not the CPU’s.

I can hear people shreiking in response. “What about guild wars? What about PvP win rates?”

I feel like making defense part of your score/rating is one of the biggest mistakes the devs have made. It means if there is a “broken” defense team, every competitive player’s priority is to make exactly that team and use it. In terms of normal PvP, this is a disaster. Week to week, even though there are goals like “get paid for killing purple troops!”, I face The Dragon Soul, Princess Fizzbang, and Kraken almost non-stop. There’s a big fat win/loss record for defense on every player’s page, and even though you technically get rewards for losing on defense then fighting revenge matches, I think players are NEVER going to want to have a “majority losses” record. So I’ve advocated as long as I’ve played for changes like:

  • The win/loss of your defense team should not be tracked or part of your PvP record.
  • Losses should result in rewards if and only if:
    • Your defense team is unique with respect to the last 7 days of teams.
    • Your defense team provides green gems for the player on offense if they win.
  • Rewards for losing should be increased, or at least “revenge” rewards should be increased.
  • Teams that win “too many” of their matches should be “retired” from the PvP rotation.
  • The “hard” team on the right of each choice should be exempt from all of these requirements and award increased ranking/glory points.

The first 4 choices make PvP “better” for people who aren’t looking for a challenge and aren’t expecting to race to the top 1000 every week. It means everyone will have a better time getting “snot gems” and there will be a wider variety of defense teams. If someone happens to sneak a Kraken or other “broken” team in, it’ll eventually fall out of the rotation.

The last point is for the people who complain that is “too easy”. It introduces a new “hardercore” tier of PvP where if you want to face “broken” teams over and over again, you’ll get what you want and be rewarded for not playing “easy mode”. People who want guild trophies, fighting for top 100, etc. are going to choose this option and ignore the others. If we create an “easier” PvP experience I think it’s important to preserve the old one as well, and reward the people who can beat the Kraken Team consistently with more stuff.

If we do this, I think looping teams on defense will naturally fall out for most people. If you want to get the best defense rewards, you’ll make a team that isn’t impossible to beat on defense. That means fewer people will constantly face a looping team, and I think that removes a lot of the impetus for “Something must be done about Kraken!”

I don’t know how to “fix” Guild Wars, but people seem happier there. “You are punished for losing on defense” is sort of crucial to how it operates.

So in short: while I think the RNG could certainly be tweaked, I think no matter what we do we’ll be plagued by some kind of “broken defense team” so long as for most players “I lose on defense” is worse than “I occasionally get rewards from revenge matches”.

5 Likes

After the 5th time this has happened JUST Today! I figured I record it and post here. BTW the amount of times I’ve had TDS revive and wipe out the entire opponents team when I’ve been playing has been Zero! I’d remember that as it’s rare he even revives when I play my dragon team let alone have the insane cascade luck you are going to see in this video which is the “norm” for my experience on PS4

I miss puzzle quest and the old days of actual strategy and you know “taking turns” lol

Enjoy the anarchy I’m off to play shadow of war again I can only take so much of this GoW “AI” fairness every day.

Edit. Just some additional info I was watching some netflix and chatting with my 3 year old while playing so that is why you see the pauses for my moves in most cases. My 3 year old is very much into the “why” everything phase atm.

Cheers
Calvin

9 Likes

The devs will just say that player and AI luck is the same :upside_down_face:

7 Likes

Tagging you both because each of your comments has made be better at understanding this game.

@Mithran I’m just gobsmacked at your atomic understanding of mechanisms I wasn’t even aware existed. Thank you for helping me understand why each of us see the same fundamental thing from so many different perspectives. I had no idea our individual play styles and inputs could bear so heavily on the peculiarities you’re discussing.

@Slypenslyde You’ve nailed so much of what is broken about Defense in how the incentive is forced on players to grief and play the most fundamentally unfair troops. You make it so clear how there is an imbalance in how much power is afforded competitive players in setting Defense teams, relative to how little that process rewards and how much it fouls the overall health of the game.

This is why I rub shoulders with you people on these forums. Like, wow.

1 Like

I don’t want quote the whole comment because most of it is germane to my question. Could this larger explanation partially account for why people think there is a big disparity between the luck the AI gets and the luck the human gets? In other words because I play like a smart human and the AI plays like my cat Ollie and therefore have totally different inputs, does this establish the platform for conditions which give “different luck” to human vs AI? Or am I stretching the concept you’ve outlined just a little too far?

@KrudlerTheHorse
Not that much of a stretch, though I think it may be more of a perceptual issue. The AI will hazard extremely stupidly timed casts a sensible player never would unless backed into a corner, and occasionally still get rewarded for it. The times they do land stick out even more in players minds. The player, attempting to make the “smart” move, won’t do things like cast a troll with two gems of their color on the board, so they’ll also obviously never experience casting a troll with two gems on the board and then using subsequent cascades to team wipe the enemy team without dropping the turn (something I’ve been on the receiving end of). The player might also sensibly wait for alignment for a converter cast and avoid converting a color they need, while the AI won’t care and sometimes again get rewarded off it. “Go for the cascade” can actually be a valid strategy if recognized as such, but it generally isn’t perceived as that because the AI is pretty much casting blindly and a human would generally only do this as a last resort option.

There are a number of other factors I also believe play into the perception of AI bias, including the general composition of defense teams (focusing on big punishes off small mistakes/cascades, eg., Psion and Mab, whereas invade teams generally run with “make them dead, this is my plan to make them dead”) and the fact that sometimes people don’t recognize their mistakes as mistakes, or will make a “bad” move that pans out as a “good” decision retroactively rather than recognizing that they just got lucky.

Since I now run multiple teams that make liberal use of the Unity cascade factor, I can safely say that players still can and will get ridiculous cascades, and can still greatly benefit from “bad” or non-decisions. The key to making teams like this work on invade, where you actually have to finish your battles instead of annoying people into retreating, is having a backup plan for when things go awry (because they will) that doesn’t end up costing you more time in the long run (which it can).

@Calv1n - Mass exploders with storms up are borderline broken. What you’ve shown here isn’t even out of the ordinary for either side to get. I’ve run a team that uses Umberwolf in the back of dragons and I’ll occasionally just extra turn my way into victory with cascades from chaining Dragon Soul casts. I believe my record for chaining Infernus with a storm up is five Infernus casts in a row, and Infernus only explodes five gems and takes 22 mana to refill, and I might have been able to keep going if the entire enemy team wasn’t a smoldering crater by that point. They can, however, also fail pretty spectacularly for the explosion user. The question is how much of the enemy team is left standing if and when that happens to get a reversal. If you want a similar team that is still powerhouse without sacrificing your mono-red standing, try swapping one of the trolls for a Dark Maiden. The extra bit of control is generally enough to let you avoid early-game damage trades, leaving you in a much better position if and when you do happen to miss a cast.

5 Likes

As always… awesome posts.

Respect!

1 Like

I understand very much what you’re saying about perception but that isn’t really what my question was about. I’d like to hone my question down a little further if I may though I’m not sure I have the skill to express this properly.

From what we’re discussing here, we are in agreement that the AI is far more likely to make “stupid” choices vs what most humans would do. So is there some bizarre set of conditions that allow the AI to get more “luck” because over the long term, it tends to give the specific inputs off which the broken scripts function?

I am having a hard time not speaking in absolutes, as well as constraining my question to the main question to what is essential. I am having some trouble, can you see where I’m going with this?

Yeah, I got what you were saying but went a bit off track with my answer. I think that the robotic inputs that the AI does could set up conditions where its spawner cast is more likely to connect, but I don’t have enough evidence to really definitively say one way or the other, hence I still believe the larger issue is perception. A human using spawners will usually attempt to increase the balance of that gem on the board, possibly taking several extra turns in the process, before firing their spawner. An AI will generally just go in swinging with a spawner. Since I still believe that first spawner casts are more likely to connect (again, contrary to the official dev word, and something that is really hard to prove one way or the other), this may or may not be one instance where the “dumb” AI inputs can set them up for more actual luck.

Everything else I believe that the AI inputs set them up for more perceived luck. In addition to what I have said before, it has also been officially stated that the AI combo breaker kicks in for PvP has gotten a certain amount more “lucky” as the player, with luck being defined in a scoring system that I’m unable to find the original dev post at the moment but includes pre-matched sky drops and extra turns. With the teams that are generally seen on defense these days and the fact that the player goes first and may accrue some of this “luck points” to start, and the overall effectiveness a single cascade can have with the way that defense teams are constructed, I believe that a lot of the problem is that the combo breaker simply won’t activate in time for a player to not get straight annihilated by the events following a single cascade chain. The cascades that the AI gets still very well may be “fair random”, but it is enough to basically close out the game and is also not something that would have happen often on Adobe.

There still remains some possibility that the “combo breaker” is broken in a different way that is allowing the AI more “actual luck” at present, but it is not what I’m personally experiencing. The devs did do a sweep for AI bias a while back and posted the results, but I believe this was before the most recent tweaks to the combo breaker. Its hard to pull back player perception from actual data, but I also don’t think it is something that should just be ignored. I can explain it away all I want, but this doesn’t mean the devs shouldn’t be doing the due diligence. The very perception that it is unfair being so widely held at the moment (theres always a few, but I’m seeing now more than ever) is something that I would generally see as a problem.

For reference, here is one of the many posts on the subject of combo breakers and how they were “fixed” going into Unity:

Sirrian’s explanation of the gem spawner streak breaker (calling it that for clarity, since they also call this a “combo breaker”, but it is a different combo breaker):

Salty’s confirmation that spawn streak breaking still works as described by Sirrian:

(Note that my original findings when the streak breaker was inverted contradict this explanation, as taking any extra turn was very clearly influencing how gems spawned, and it was very clearly active after one extra turn taken. This may have very well been “fixed”, but the fact that the fix came mid patch cycle and leaves me very suspicious as if the underlying code is actually working as described.)

For reference, when it was broken, it looked like this:

4 Likes

Here’s my game where I held onto a spell instead of casting. Watch how I get battered, then the tables get turned.
Its quite amazing sometimes how this game decides that you will not win. Our guild has had a terrible start to the weekly gw. Many times today the game was impossible to play against, many times I simply didn’t get a single cast off.

1 Like

GoW has become both RNG gem drops combined with overpowered RNG troop abilities (RNG devour, RNG target enemies/allies, RNG remove/explode/create gems etc.) The devs have largely RNG’ed the skill out of the actual matches.

At this point choosing your team is probably the most important aspect of GoW. Before starting the match, look at the enemy troops and select what you hope will be the best counters. For me, the devs have managed to RNG much of the fun out of this game.

6 Likes

While I agree with the larger point and have said the same myself, I think there are still room for making crucial plays that the AI would not make that would influence the outcome of your match. The teams I build almost always use one deterministic board modder, so I can make at least one non-robotic choice that influences the outcome of my game on a semi-infrequent basis (sometimes not even every game do I get to make a “not completely obvious choice”, but its there). Its not like the careful control teams I used to play almost exclusively, where there was one or two “good” moves, and a dozen or more “bad” ones at pretty much every step of every game (which still work, they are just comparatively slow without really being that much “safer”), but its a compromise. The fact that a bunch of autopilot teams are currently just as effective as teams where you actually have to stop and make a decision is something that I’ve (mostly) come to terms with. I just try not to build full autopilot teams myself, because after 15 minutes using them it feels like I’m just wasting my time.

3 Likes

TL;DR

This game used to be fun. Now it is much less fun.

5 Likes

Its funny watching the echo chamber lol…

Disagree with the crowd: get ridiculed and insulted

Defend yourself in any way: get banned

Nice forums…

1 Like

These are the comments we need more of imo. However the thing you are touching on for me I don’t think I will be able to come to terms with until the effect is attenuated heavily. I want unskilled play to be devalued slightly in favor of skilled play. I don’t mind unskilled play because there are times when it is essential to good strategy, and emotionally satisfying it its own way.

The polar opposite of the bingo-hopper feeling I get when watching (former) Trolls in action is found when I play Joker’s Wild (Mercy/IK/Alch/GR). This is is a (minimum) 97% win deck that takes 3 Reds. But it isn’t fun to play where I know that I can win every time through perfect strategy and never relinquishing board control. It makes the game boring as hell. Likewise I never used Entangling Bones once I honed it down to the essentials, and that goes for every other team I’ve hyper-optimized. Because they wreck the experience.

So it’s a strange feeling to hate the ebb and flow of RNG and the huge swings it can produce, and at the same time to know that without sufficient RNG the game experience can be very emotionally unsatisfying.

2 Likes

Magic: the Gathering players know this well. “Mana Flood” and “Mana Screw” will lose your most important games and leave you with nothing to do. Depending on how you look at the numbers, your deck is configured in a way that can’t win 20%-30% of the time. That’s why in tournament play, everyone in the top 8 is considered “a winner”, because it’s hard to get that high just on luck but easy to lose 1st to luck. Hall of Fame players tend to have win rates around 60%. Some haven’t ever finished 1st.

But every solution to the problem’s been met with the same frustrating conclusion: when you take that aspect out of Magic, it’s never as fun.

GoW is like that. I want to get 10 free turns sometimes, and I understand that means the CPU should get 10 free turns sometimes too. But it feels like:

  • I can’t win unless I find a way to statistically guarantee 10 free turns more often.
  • I face CPU teams that consistently get 3-4 turns to my 1.
  • At 4x speed I say a lot of “WTF just happened???”
1 Like

I may need to drop out here. I was introduced to MtG 2 years ago by a friend who insisted I play. After $200 to get up and running I realized that I’d need to spend about another $40,000 to have a mathematical chance before I could even start to consider strategy, so I gave up. I’m exaggerating for effect… but seriously I don’t think I can discuss this issue in that context because it’s so far out of my area of knowledge.