D3 better step up & fix AI cheating ASAP

Unknown

I'd love to see the end of giant cascades on both sides. For instance, after four combos the board resets. Still allows for setting up combos while removing game ending bad luck.

Unknown

Trisul wrote:

"Better cascades" is too difficult to quantify. Is it just total AP generated? Is it "important" AP generated?

That thread is full of anecdotal evidence. Unless a dev shows up and tells us directly that the AI has improved, I'm going to chalk all the complaints up to a relatively high-difficulty PvE event (especially with a better 2* Daken) combined with mass amounts of Juggernaut crashes, which have always cascaded ridiculously.

Yes, better cascades would be more AP. I do not care about the quality of the AP, which I think/hope is random. The real question is does the AI more frequently get better cascades that result in more matches and AP than the players? Not AP obtained on average, not number of cascades. I"m talking does the AI get "super" cascades more frequently than players get "super" cascades.

Unknown

Trisul wrote:

daveomite wrote:

Actually, I think that thread is full of various comments from longer term players who have played quite often, in a lot of events, and notice the difference in the system.

- Recent PvE's have been fairly easy, especially for players with pre-nerf Magneto (who basically ended a match after a certain point).
- Juggernaut has always been a pain, with frequent cascades. I still remember the trouble I had with Juggernaut nodes in the prologue when I was just starting the game.
- This is the first event with upgraded 2* Daken nodes.
- Limited heroic rosters mean over-reliance on buffed characters. Heroic Venom had Patch, where he could basically regen through any AI cascades. Here, Thor being slow overall with no regen and limited combo potential with any other character meant that games lasted much longer, giving more opportunities for cascades overall, endless Juggernaut combos, etc.

Considering these points, it's not far-fetched to suggest that the devs are telling the truth. Besides the fact that they are, you know, devs.

Actually, the changes are more pronounced in PVP than PVE. PVE difficulty is more scaling based (although not entirely).

Also, ever notice how your Rags green or Hulk Green does not cause the same frequency or degree of cascades as the AI? I would seriously questions any statistics that dispute this. The difference between player and AI results are too obvious too ingore or be marginalized as confirmation bias.

Trisul

stephen43084 wrote:

Actually, the changes are more pronounced in PVP than PVE. PVE difficulty is more scaling based (although not entirely).

Also, ever notice how your Rags green or Hulk Green does not cause the same frequency or degree of cascades as the AI? I would seriously questions any statistics that dispute this. The difference between player and AI results are too obvious too ingore or be marginalized as confirmation bias.

I find it peculiar that you give a clear hypothetical statement ("your Rags green or Hulk Green does not cause the same frequency or degree of cascades as the AI"), and then ask for contrary statistics to disprove it. Dude, the burden of proof is not on us, it's on you.

NorthernPolarity

stephen43084 wrote:

Trisul wrote:

daveomite wrote:

Actually, I think that thread is full of various comments from longer term players who have played quite often, in a lot of events, and notice the difference in the system.

- Recent PvE's have been fairly easy, especially for players with pre-nerf Magneto (who basically ended a match after a certain point).
- Juggernaut has always been a pain, with frequent cascades. I still remember the trouble I had with Juggernaut nodes in the prologue when I was just starting the game.
- This is the first event with upgraded 2* Daken nodes.
- Limited heroic rosters mean over-reliance on buffed characters. Heroic Venom had Patch, where he could basically regen through any AI cascades. Here, Thor being slow overall with no regen and limited combo potential with any other character meant that games lasted much longer, giving more opportunities for cascades overall, endless Juggernaut combos, etc.

Considering these points, it's not far-fetched to suggest that the devs are telling the truth. Besides the fact that they are, you know, devs.

Actually, the changes are more pronounced in PVP than PVE. PVE difficulty is more scaling based (although not entirely).

Also, ever notice how your Rags green or Hulk Green does not cause the same frequency or degree of cascades as the AI? I would seriously questions any statistics that dispute this. The difference between player and AI results are too obvious too ingore or be marginalized as confirmation bias.

Well, I've noticed that the frequency of good AI cascades is roughly on par to the frequency of good cascades on my end, and I would seriously question any statistics that dispute this. Obviously I'm right because I say so.

Look at the facts. We have a developer on here that specifically said that the AI has not been changed at all in 2014. Not even a "oh minor tweaks have been made" comment but straight up "the ai has not been changed for 9 months". If he were outright lying to us (which you think so), and someone found out, then there would be huge damage to the games reputation. This would cause a lot more damage than the profits that they would have gained from increased health pack sales caused by sneakily giving the AI more cascades. There's no reason to believe that some ridiculous conspiracy is going on with the developers meticulously modifying the AI code to screw the player over. If you think there is, then give us some concrete evidence. Otherwise, you're no better than the people who claim that Sentry is balanced.

CrookedKnight

I have a confession: I was a member of the same tinfoil hat club when I played the original Puzzle Quest. It wasn't until I got my hands on the actual source code as part of this gig and looked for myself that I believed that the AI wasn't any luckier than I was.

Really? Because it would regularly, demonstrably break its usual preferences for matching damage tiles over color tiles if the color tiles were going to give a "random" extra turn. (The equivalent would be MPQ AI matching a color it has no power for if that color is going to start a cascade, which it doesn't do so good job there)

Unknown

NorthernPolarity wrote:

stephen43084 wrote:

Trisul wrote:

daveomite wrote:

Actually, I think that thread is full of various comments from longer term players who have played quite often, in a lot of events, and notice the difference in the system.

- Recent PvE's have been fairly easy, especially for players with pre-nerf Magneto (who basically ended a match after a certain point).
- Juggernaut has always been a pain, with frequent cascades. I still remember the trouble I had with Juggernaut nodes in the prologue when I was just starting the game.
- This is the first event with upgraded 2* Daken nodes.
- Limited heroic rosters mean over-reliance on buffed characters. Heroic Venom had Patch, where he could basically regen through any AI cascades. Here, Thor being slow overall with no regen and limited combo potential with any other character meant that games lasted much longer, giving more opportunities for cascades overall, endless Juggernaut combos, etc.

Considering these points, it's not far-fetched to suggest that the devs are telling the truth. Besides the fact that they are, you know, devs.

Actually, the changes are more pronounced in PVP than PVE. PVE difficulty is more scaling based (although not entirely).

Also, ever notice how your Rags green or Hulk Green does not cause the same frequency or degree of cascades as the AI? I would seriously questions any statistics that dispute this. The difference between player and AI results are too obvious too ingore or be marginalized as confirmation bias.

Well, I've noticed that the frequency of good AI cascades is roughly on par to the frequency of good cascades on my end, and I would seriously question any statistics that dispute this. Obviously I'm right because I say so.

Look at the facts. We have a developer on here that specifically said that the AI has not been changed at all in 2014. Not even a "oh minor tweaks have been made" comment but straight up "the ai has not been changed for 9 months". If he were outright lying to us (which you think so), and someone found out, then there would be huge damage to the games reputation. This would cause a lot more damage than the profits that they would have gained from increased health pack sales caused by sneakily giving the AI more cascades. There's no reason to believe that some ridiculous conspiracy is going on with the developers meticulously modifying the AI code to screw the player over. If you think there is, then give us some concrete evidence. Otherwise, you're no better than the people who claim that Sentry is balanced.

You just did not go further back in the thread to see what else I wrote. I already conceded that the number of cascades was on par and the statistics backed that up. I was questioning whether the quality of the cascades were the same. This was not addressed in the developer's previous statement. I did in know call him a liar or question the information he provided. Also, I can question whether certain character abilities work better at producing cascades for the AI than players without questioning the preceding statement that the overall number of cascades is even between the AI and players.

For example, the AI may get 5 cascades from Rags and I get 1 cascade from Rags in the same number of attempts. However, I may get 5 cascades from MMN's purple), while the AI may only get 1 cascade from MMN. Therefore, the AI is getting better cascade results from Rags, but I am getting better cascade results from MMN. So while the AI does get more cascades from a character than the player, the overall number of cascades is even. Therefore, my supposition does not dispute the statistics or claims made by the developer (who was nice enough to take time to post on the forum).

NorthernPolarity

stephen43084 wrote:

You just did not go further back in the thread to see what else I wrote. I already conceded that the number of cascades was on par and the statistics backed that up. I was questioning whether the quality of the cascades were the same. This was not addressed in the developer's previous statement. I did in know call him a liar or question the information he provided. Also, I can question whether certain character abilities work better at producing cascades for the AI than players without questioning the preceding statement that the overall number of cascades is even between the AI and players.

For example, the AI may get 5 cascades from Rags and I get 1 cascade from Rags in the same number of attempts. However, I may get 5 cascades from MMN's purple), while the AI may only get 1 cascade from MMN. Therefore, the AI is getting better cascade results from Rags, but I am getting better cascade results from MMN. So while the AI does get more cascades from a character than the player, the overall number of cascades is even. Therefore, my supposition does not dispute the statistics or claims made by the developer (who was nice enough to take time to post on the forum).

Fair enough, my bad. DemiurgeWill / IceIX: does the quality of cascades differ between the player and the AI? I suppose I could have been too naive in assuming that if the number of cascades is the same between the AI/player that the quality of cascades is the same as well, but the exact same logic applies to both situations so it seems highly unlikely that they would respond any differently. It's still nice to know so that either all of the conspiracy theorists can be satiated or we uncover some huge conspiracy of d3 doing this to sell more health packs.

Unknown

NorthernPolarity wrote:

stephen43084 wrote:

You just did not go further back in the thread to see what else I wrote. I already conceded that the number of cascades was on par and the statistics backed that up. I was questioning whether the quality of the cascades were the same. This was not addressed in the developer's previous statement. I did in know call him a liar or question the information he provided. Also, I can question whether certain character abilities work better at producing cascades for the AI than players without questioning the preceding statement that the overall number of cascades is even between the AI and players.

For example, the AI may get 5 cascades from Rags and I get 1 cascade from Rags in the same number of attempts. However, I may get 5 cascades from MMN's purple), while the AI may only get 1 cascade from MMN. Therefore, the AI is getting better cascade results from Rags, but I am getting better cascade results from MMN. So while the AI does get more cascades from a character than the player, the overall number of cascades is even. Therefore, my supposition does not dispute the statistics or claims made by the developer (who was nice enough to take time to post on the forum).

Fair enough, my bad. DemiurgeWill / IceIX: does the quality of cascades differ between the player and the AI? I suppose I could have been too naive in assuming that if the number of cascades is the same between the AI/player that the quality of cascades is the same as well, but the exact same logic applies to both situations so it seems highly unlikely that they would respond any differently. It's still nice to know so that either all of the conspiracy theorists can be satiated or we uncover some huge conspiracy of d3 doing this to sell more health packs.

I don't buy into the whole sell more health packs thing. First, I'm not sure how MPQ makes money, but I do not think health packs are paying their bills. Secondly, whatever they have to do to make money to improve the game and keep it going so I can play it (as long as I'm still having fun), I actually do not have a problem that. Caveat, as long as, other players in my alliance are also having fun, I have no problem with it.

Trisul

NorthernPolarity wrote:

Fair enough, my bad. DemiurgeWill / IceIX: does the quality of cascades differ between the player and the AI? I suppose I could have been too naive in assuming that if the number of cascades is the same between the AI/player that the quality of cascades is the same as well, but the exact same logic applies to both situations so it seems highly unlikely that they would respond any differently. It's still nice to know so that either all of the conspiracy theorists can be satiated or we uncover some huge conspiracy of d3 doing this to sell more health packs.

Let's just say I'm skeptical that legitimate statistical analysis will ever prove that the AI gets better or more cascades than players (anecdotal evidence is not enough, people!). The only way that can happen is if the AI sees the next few rows that are dropping, and Will specifically stated that this is not the case.

NorthernPolarity

Trisul wrote:

NorthernPolarity wrote:

Fair enough, my bad. DemiurgeWill / IceIX: does the quality of cascades differ between the player and the AI? I suppose I could have been too naive in assuming that if the number of cascades is the same between the AI/player that the quality of cascades is the same as well, but the exact same logic applies to both situations so it seems highly unlikely that they would respond any differently. It's still nice to know so that either all of the conspiracy theorists can be satiated or we uncover some huge conspiracy of d3 doing this to sell more health packs.

Let's just say I'm skeptical that legitimate statistical analysis will ever prove that the AI gets better or more cascades than players (anecdotal evidence is not enough, people!). The only way that can happen is if the AI sees the next few rows that are dropping, and Will specifically stated that this is not the case.

His answer still doesn't explicitly answer stephens question because even though the new tiles are dropped the same way, it's possible that the code that chooses which tiles will be destroyed will destroy tiles in such a way that the AI gets better cascades from the existing tiles on the board than the player does. Not that I think that this is how it works since it seems very convoluted, but there won't be a concrete thing to refer to when asked with questions like this until all possibilities are put to rest.

Unknown

I will say this, in pve and pvp, if the starting board sucks, I've practically also lost in 80% of cases.
Cascades seem fairly even to me. I will say that the ai still occasionally launches abilities it doesn't have the ap for. I'm guessing there's some sort of bug when checking ap costs of abilities.

Unknown

Every time I see an AI conspiracy post I just laugh. The AI is not out to get any of you. It's just there for you to play against and enjoy the game. Just because you lost a couple matches to cascades of doom doesn't mean the AI cheats.

IceIX

NorthernPolarity wrote:

His answer still doesn't explicitly answer stephens question because even though the new tiles are dropped the same way, it's possible that the code that chooses which tiles will be destroyed will destroy tiles in such a way that the AI gets better cascades from the existing tiles on the board than the player does. Not that I think that this is how it works since it seems very convoluted, but there won't be a concrete thing to refer to when asked with questions like this until all possibilities are put to rest.

It doesn't, for the reason you state. It would be pretty convoluted to create something like that and wouldn't actually accomplish much of anything other than annoy our player base. Not to mention that an asymmetric drop algorithm would be a nightmare for QA. We try not to make QA's heads explode.

Dayv

NorthernPolarity wrote:

His answer still doesn't explicitly answer stephens question because even though the new tiles are dropped the same way, it's possible that the code that chooses which tiles will be destroyed will destroy tiles in such a way that the AI gets better cascades from the existing tiles on the board than the player does. Not that I think that this is how it works since it seems very convoluted, but there won't be a concrete thing to refer to when asked with questions like this until all possibilities are put to rest.

It's a pseudo-religious belief. I'm losing hope it can be put to rest at all.

NorthernPolarity

IceIX wrote:

NorthernPolarity wrote:

His answer still doesn't explicitly answer stephens question because even though the new tiles are dropped the same way, it's possible that the code that chooses which tiles will be destroyed will destroy tiles in such a way that the AI gets better cascades from the existing tiles on the board than the player does. Not that I think that this is how it works since it seems very convoluted, but there won't be a concrete thing to refer to when asked with questions like this until all possibilities are put to rest.

It doesn't, for the reason you state. It would be pretty convoluted to create something like that and wouldn't actually accomplish much of anything other than annoy our player base. Not to mention that an asymmetric drop algorithm would be a nightmare for QA. We try not to make QA's heads explode.

Cool. For you conspiracy theorists, are there any other theories about other ways that the AI can be cheating? Speak up now or forever hold your peace.

_RiO_

NorthernPolarity wrote:

IceIX wrote:

NorthernPolarity wrote:

His answer still doesn't explicitly answer stephens question because even though the new tiles are dropped the same way, it's possible that the code that chooses which tiles will be destroyed will destroy tiles in such a way that the AI gets better cascades from the existing tiles on the board than the player does. Not that I think that this is how it works since it seems very convoluted, but there won't be a concrete thing to refer to when asked with questions like this until all possibilities are put to rest.

It doesn't, for the reason you state. It would be pretty convoluted to create something like that and wouldn't actually accomplish much of anything other than annoy our player base. Not to mention that an asymmetric drop algorithm would be a nightmare for QA. We try not to make QA's heads explode.

Cool. For you conspiracy theorists, are there any other theories about other ways that the AI can be cheating? Speak up now or forever hold your peace.

There are multiple accounts of people seeing various AI controlled characters use skills before the AI had obtained their requisite AP. And some report that the AI had used a skill at no actual AP cost.

NorthernPolarity

_RiO_ wrote:

NorthernPolarity wrote:

IceIX wrote:

NorthernPolarity wrote:

His answer still doesn't explicitly answer stephens question because even though the new tiles are dropped the same way, it's possible that the code that chooses which tiles will be destroyed will destroy tiles in such a way that the AI gets better cascades from the existing tiles on the board than the player does. Not that I think that this is how it works since it seems very convoluted, but there won't be a concrete thing to refer to when asked with questions like this until all possibilities are put to rest.

It doesn't, for the reason you state. It would be pretty convoluted to create something like that and wouldn't actually accomplish much of anything other than annoy our player base. Not to mention that an asymmetric drop algorithm would be a nightmare for QA. We try not to make QA's heads explode.

Cool. For you conspiracy theorists, are there any other theories about other ways that the AI can be cheating? Speak up now or forever hold your peace.

There are multiple accounts of people seeing various AI controlled characters use skills before the AI had obtained their requisite AP. And some report that the AI had used a skill at no actual AP cost.

I think that the majority of those accounts can be explained by people not knowing that goons generate AP. I saw some posts like that of people wondering how Yelena was able to cast recon so quickly, without noticing that she had 2 purple goons generating AP for her. Feel free to post any examples that don't seem like this is the reason though: it'll be like Mythbusters MPQ Edition!

IceIX

NorthernPolarity wrote:

I think that the majority of those accounts can be explained by people not knowing that goons generate AP. I saw some posts like that of people wondering how Yelena was able to cast recon so quickly, without noticing that she had 2 purple goons generating AP for her. Feel free to post any examples that don't seem like this is the reason though: it'll be like Mythbusters MPQ Edition!

We've investigated quite a few of these reports internally and our results too are about the same. In each case, we discovered that it was a case of minions gathering AP and the user just not taking this into account. Admittedly, the mobile UI doesn't do the best job of constantly updating the player on opponent AP totals at current , but we haven't hit a certifiable situation where AP mysteriously gets generated after quite a lot of concerted testing on the matter.

Unknown

IceIX wrote:

NorthernPolarity wrote:

I think that the majority of those accounts can be explained by people not knowing that goons generate AP. I saw some posts like that of people wondering how Yelena was able to cast recon so quickly, without noticing that she had 2 purple goons generating AP for her. Feel free to post any examples that don't seem like this is the reason though: it'll be like Mythbusters MPQ Edition!

We've investigated quite a few of these reports internally and our results too are about the same. In each case, we discovered that it was a case of minions gathering AP and the user just not taking this into account. Admittedly, the mobile UI doesn't do the best job of constantly updating the player on opponent AP totals at current , but we haven't hit a certifiable situation where AP mysteriously gets generated after quite a lot of concerted testing on the matter.

Even on Steam the enemy's AP total is not updated until the end of a move. Not sure if it was fixed but one particularly obvious example is when you trigger an enemy's Captain Marvel's passive with a move (so you still have your turn) and you can see the AP total do not update to reflect that, but the appropriate abilities will light up to indicate they can be used even though the display would should insufficient AP. It doesn't affect gameplay but can easily lead people into thinking the AI had considerably less AP than they really have.