Insane CPU cascades

geoffubazi2

2 games in a row, I decided to count the computer's AP gain in a single turn. Both times I lost track after 75 from a single match. All 3 characters dead. 2-3 rounds into the match. This happened 2 games in a row. I get the impression the CPU can either see 5-6 rows above the screen, or are set to occasionally just wipe you, to try to get $ from the player.

Unknown

One of these threads pops up every week or so. Yes, it is less than pleasant when you're hit by a cascade of doom. But we get them in our favor as well, but don't usually remember or document them all.

Here's a thread from a while back that has a good amount of them, some for the AI, some for the players.

viewtopic.php?f=7&t=18453

Unknown

Actually, we have some reasonable anecdotal evidence that the A.I. does in fact cheat and gives itself a higher portion of cascades, based on certain criteria.

viewtopic.php?f=7&t=30201

geoffubazi2

j12601 wrote:

One of these threads pops up every week or so. Yes, it is less than pleasant when you're hit by a cascade of doom. But we get them in our favor as well, but don't usually remember or document them all.

Here's a thread from a while back that has a good amount of them, some for the AI, some for the players.

viewtopic.php?f=7&t=18453

That sounds telling doesn't it

Vhailorx

geoffubazi2 wrote:

j12601 wrote:

One of these threads pops up every week or so. Yes, it is less than pleasant when you're hit by a cascade of doom. But we get them in our favor as well, but don't usually remember or document them all.

Here's a thread from a while back that has a good amount of them, some for the AI, some for the players.

viewtopic.php?f=7&t=18453

That sounds telling doesn't it

I hope you aren't suggesting that the existence of threads complaining about ai cascades is proof of ai cascades. I direct you, sir or madam, to UFOs, ghosts, bigfoot/sasquatch, the loch ness monster, and psychics. The existence of persistent popular theories and a persistent belief in something != proof that the something in question exists.

ai cascade threads on this forum are almost entirely anecdotal frustration-fueld complaints easily explained by confirmation bias.

There are some threads that rise above that level, including the one cited theHappyDance, are slightly better in that they attempt to gather actual data, measuring the cascade rate for the ai and a single player over a larger set of matches. Even this better data is not conclusive for the simple reason that it is obvious to anyone who plays this game for five minutes that humans and the ai play very differently. It is entirely possible that the ai produces more cascades because it plays like a dumbass: failing to prioritize high-value colors, always taking match-4s (even over match-5s), and never considering what moves will be available after taking a particular match. Humans do all of those things to different degrees, which could easily have a significant effect on the cascade rate and other secondary board outcomes. And while we only care about human versus ai matches, one particular human's play style is not very helpful. The only way to really measure this phenomenon would be to gather truly randomized data by sampling from a very large group of humans.

Also, there is the law of parsimony to consider. Why would the devs bother stacking the deck against the players by subtly tweaking the cascade rates? They can (and do) stack the deck in their favor openly with things like token pull rates, shield cooldowns, expiring covers, roster slot pricing, etc.

Unknown

I had an insane AI cascade today, 68 AP. The funny part was it was in the DPD on the one star node. I only have a level 50 iron man, but his protect tiles meant all the match damage was reduced to 1, even the crits.

Unknown

It's swings and roundabouts. I have spent some time in Sim and Marvel today. Sim I have had no issue at all, in PVP the exact opposite. The last four matches the ai has has double figures on at least one colur after two rounds, and it snowballs from there.

It's weird, it seems unfair at times for sure with the odd way it seems to go in waves, but it is just luck of the draw.

Unknown

theHappyDance wrote:

Actually, we have some reasonable anecdotal evidence that the A.I. does in fact cheat and gives itself a higher portion of cascades, based on certain criteria.

viewtopic.php?f=7&t=30201

We debunked that. The OP in that post plays badly and takes chances that give the AI a better chance to crit or line clear, thus cascade.

grunth13

daibar wrote:

theHappyDance wrote:

Actually, we have some reasonable anecdotal evidence that the A.I. does in fact cheat and gives itself a higher portion of cascades, based on certain criteria.

viewtopic.php?f=7&t=30201

We debunked that. The OP in that post plays badly and takes chances that give the AI a better chance to crit or line clear, thus cascade.

Nothing is debunked until proven. You are right that the OP in that thread may be playing differently therefore the AI gets more cascades because of it, but if everything is truly random, that shouldn't matter either because in the long run random is random and usually equals out close to the middle.

Unknown

grunth13 wrote:

daibar wrote:

theHappyDance wrote:

Actually, we have some reasonable anecdotal evidence that the A.I. does in fact cheat and gives itself a higher portion of cascades, based on certain criteria.

viewtopic.php?f=7&t=30201

We debunked that. The OP in that post plays badly and takes chances that give the AI a better chance to crit or line clear, thus cascade.

Nothing is debunked until proven. You are right that the OP in that thread may be playing differently therefore the AI gets more cascades because of it, but if everything is truly random, that shouldn't matter either because in the long run random is random and usually equals out close to the middle.

So on a road trip from New York to Los Angeles, a guy who goes generally at 100 mph and a guy who goes at 80 mph will arrive there at about the same time because random traffic is random and will balance out in the long run. ==> That doesn't make sense because traffic is only one of the factors.

Players' play is not random. That's why it doesn't balance out; that's why the AI gets more cascades then he does. It's not random in the way he picks his moves. It's not random in the characters he chooses. His play is long term and isn't "balanced out". Player cascades vs AI cascades depends hugely on play style. I'd like to see someone play with IF/HB/Cyclops and claim that they don't get as many cascades as the AI.

Zen808

If you look at Polarity's Ability Analysis, viewtopic.php?style=1&f=14&t=18598, you'll see a section where he compares using Loki's Illusions by itself, vs after Polarizing force. The result of decreasing the variety on a board by one color increases odds of cascades by 12%, which is statistically very significant.

Now, imagine the typical human playstyle: bumrush one or two colors. Compared to the AI, which chooses colors to match at random.

It's no coincidence that you had that one guy who claimed that he reduced AI cascades by "playing like the AI."

Buret0

808SpicyToro wrote:

Compared to the AI, which chooses colors to match at random.

Anyone else find that the AI was plundering yellows in Top Gun?

Even if Falcon had no special tiles to increase, I found that the AI was attacking yellow as a very high priority. I often had to leave a garbage match 4 to keep the AI from stealing my + 122 to three of my special tiles.

grunth13

daibar wrote:

grunth13 wrote:

daibar wrote:

theHappyDance wrote:

Actually, we have some reasonable anecdotal evidence that the A.I. does in fact cheat and gives itself a higher portion of cascades, based on certain criteria.

viewtopic.php?f=7&t=30201

We debunked that. The OP in that post plays badly and takes chances that give the AI a better chance to crit or line clear, thus cascade.

Nothing is debunked until proven. You are right that the OP in that thread may be playing differently therefore the AI gets more cascades because of it, but if everything is truly random, that shouldn't matter either because in the long run random is random and usually equals out close to the middle.

So on a road trip from New York to Los Angeles, a guy who goes generally at 100 mph and a guy who goes at 80 mph will arrive there at about the same time because random traffic is random and will balance out in the long run. ==> That doesn't make sense because traffic is only one of the factors.

Players' play is not random. That's why it doesn't balance out; that's why the AI gets more cascades then he does. It's not random in the way he picks his moves. It's not random in the characters he chooses. His play is long term and isn't "balanced out". Player cascades vs AI cascades depends hugely on play style. I'd like to see someone play with IF/HB/Cyclops and claim that they don't get as many cascades as the AI.

Actually, traffic is seldom random. Traffic happens when increased amount of people are on the road, such as when people have to go to work, people going to lunch, people going home, etc. Therefore, your traffic is random analogy doesn't equate in my opinion.

Malcrof

Unknown

Buret0 wrote:

808SpicyToro wrote:

Compared to the AI, which chooses colors to match at random.

Anyone else find that the AI was plundering yellows in Top Gun?

Even if Falcon had no special tiles to increase, I found that the AI was attacking yellow as a very high priority. I often had to leave a garbage match 4 to keep the AI from stealing my + 122 to three of my special tiles.

Yeah, I've seen that same behavior many times. Enemy characters have absolutely no use for a color, yet keep picking it over other colors they could actually use.

I'm sure it's not actually happening though, and someone will eventually be kind enough to inform me that I'm imagining it.

DBC

Katai

The problem with CPU cascades isn't their frequency. From a statistics point of view, players get far more amazing cascades than the CPU player does over the course of the game. The problem with CPU cascades is the devastating effect.

Typical PvP event requires about 30-40 matches to reach 700 or 800 points (at roughly 25 points per match, plus however much you need to make up after getting attacked). To get up that high requires you to lose as little HP as you can in every match, as well as finish the match as quickly as possible. A miracle cascade for you means you lose almost no HP in very little time, which is good, but it's also what you were trying to achieve anyway. That miracle cascade, at most, only did 1/30th of the work that needs to be done.

A miracle cascade for the CPU, however, is devastating. It's an instant wipe of your primary team. That's 3 healthpacks worth. If it happens once, you can keep moving, but you probably won't reach your goal of 700 or 800 points. If it happens twice, just give up on that goal and find a place to shield up.