Constructive feedback from "new" player : we need strategy and lucky board algorithms back please!

Apparently there is a depth to replies and I cannot reply to your last message.

I was just gonna say, "lucky" board or a "oh just one more move sucker" board. 😁

Basically what likely happens is

Basic templates for all the level types, blockers, candies and such are stored on the device.
There is some code to get the template and all the other stuff needed for each level, this is not actually much data per level
There is a base algorithm stored on the device.
The rest of the algorithm (multiple) are also stored but there are much smaller than the base one.
They are applied to a level and the level is built, in the past when loading, now I suspect a few are built as you play in the background as that would be faster for loading.

So likely the whole process is

A template is grabbed
A manifest file is read and all static parts are put into place
The algorithm part is selected and added to the base one.
Any difficultly or unique algorithm info is then added.
The level is generated in reverse to line up with the algorithm
We play it.

There is a fair bit more to it I suspect but that would be the general gist of it. This way it would not take up too much memory and if done while playing a level, level would load faster.

I agree @PyroFunk, but I fear that all these recent changes are leading into exactly the opposite direction of the solution you are suggesting. It appears to me that the algorithms are being streamlined and simplified so that a "one method fits all" solution can be found in order to reduce computing power needed to run the game. I might be wrong here, but does the offline nature of the game not require the entirety of every single one of the 20,000+ levels and their algorithms to be downloaded onto our devices each time we open and load the app? That must take a huge amount of memory. I fear that all these recent changes have been made to reduce the memory required to play, since devices cannot handle it. In fact there have been fewer complaints of glitches and the game hanging up since September.

https://community.king.com/en/candy-crush-saga/discussion/comment/3434695#Comment_3434695

Oh I see now that you are right. That must be the winning move: The fish always nowadays cause the game to finish. I wish I had taken better care and time when I played this level! I wish I had noted down which last move the game actually recommended at that stage. But of course none of this theorising affects the validity of the initial argument. The game is rigged to leave 99% of players one frosting short at this point as 99% of players would not see that one perfect move. Let's not forget the game is made so that we rush through levels now. We are not meant to play slowly to analyse those algorithms carefully.

It may be that the overall problem is that the AI that writes these algorithms cannot factor in the way humans play.

My initial argument here was that there is a too stark yoyo effect in the game. Some levels and some entire episodes are too easy, making players bored, rely on intuition and/or hints and on speed, play to zoom through the game. The events and features with the timed boosters encourage this type of strategy. The whole of the first 200 levels which are essentially a tutorial, gear players towards this kind of game. There are some episodes in the 1000-2000 range like that too and in this particular account levels 200-260 were of this overly easy type. They were all easily passable with one colour bomb, which the game gave us as timed. From this kind of gameplay the newbie slithers into a totally different type of episode, for which those strategies the game taught them suddenly don't work. I don't think this is deliberate. I think this is because the game has become patchy: In between those easy episodes are some that do not follow this new format, often those are levels that were written by humans and that were left in the game during previous AI rewrites. Those episodes then rely on a totally different strategy to play, what I call the "best possible move" strategy. There is a solution to those levels and until August this kind of strategy used to work well for all levels in the game. I feel that since September the "lucky" end of play algorithms have become decoupled from the "lucky" early and mid play algorithms for each level:

We used to be able to expect luck to carry through and finish the level for us if we had luck at the beginning (because we played the correct starting boosters) and luck in the middle (because we played the best available matches on the board). On such boards we used to then reliably get rewarded with a lucky finish that we could actually see and take advantage of. I accept that in some legendary levels this still would require the use of a lollipop, switch, paint brush or UFO, the "use the correct booster at the right time" strategy, which can be immensely satisfying, but this too has now changed.

This last point is what angers me the most about the latest "end of play" algorithms: There is no longer the point in those final 3 or 5 moves where a player can see which booster to play to win. The above board of level 288 illustrates this and it's at the basis of all the complaints I have posted since September: The game may leave us one lollipop short, but it no longer allows us to use the lollipop hammer as this must be done before the last move, so it forces us to pay gold bars. This is unsatisfying, frustrating and infuriating. I cannot be the only person to find this is the case.

The end of play part of the algorithms are now controlled by the jelly fish. There must be a way to tweak them so that it creates boards where players get the opportunity to use a specific booster to pass a level, rather than pay 10 gold. This really needs to be looked into and changed in order to make the game enjoyable for new players.

https://community.king.com/en/candy-crush-saga/discussion/comment/3434073#Comment_3434073

Matching either the three greens (including the wrap) horizontally

This is a perfect match because the main point is:

You have also created a purple fish and this fish will be activated by the explosion. It will destroy the frosting.

I don't discover this perfect move, but if I play this board, I would move the top purple candy down to create a third wrapped candy and then activate the explosion as well. This time delay explosion can break the frosting and create higher scores. If score still exists 😆

Your structured feedback really helps highlight where things are working and where fine-tuning might still be needed, it’s very much appreciated

Objectively speaking 66 to 900, the whole thing is all over the place.

Also the game simply needs a new algorithm. Currently it is pretty clear it just checked to see if you made the move it expect regardless of how insane it might be. This is a terrible design.

It needs a whole re-write to:

Generate each level:
1. and play it a number of times to teach the AI. Does not have to always been done just for "test users" for levels
2. take info from what moves people do and figure out the general idea of what is thought of as a good move
Weight all the items on the screen
Figure out what moves have the most weight, ie lead to clearing the board
Reward high weight moves and re calculate as the cascade happens (since the result is already fixed and not random).

Basically the boards would reward making what most would consider a good move not some random "should do nothing" move halfway across the board.

https://community.king.com/en/candy-crush-saga/discussion/513011/constructive-feedback-from-new-player-we-need-strategy-and-lucky-board-algorithms-back-please

I am not 100% convinced either move would not just invalidate the first explosion on the pink frosting. I am almost positive I have seen such nonsense.

QueenB

As always Christine, thank you so much for taking the time to play through and document these early levels so carefully… your detailed notes and screenshots are incredibly helpful.

It’s great to hear that some of the levels now feel more balanced and that the fish are behaving better in parts of your run. I’ve shared your observations (especially about level 288 and the inconsistent fish targeting) with the studio so they can take a closer look and make sure those logic improvements are applied consistently across all levels.

Your structured feedback really helps highlight where things are working and where fine-tuning might still be needed, it’s very much appreciated 🫡

https://community.king.com/en/candy-crush-saga/discussion/comment/3434073#Comment_3434073

When I read a complaint in Other Places about moving every best possible move but cannot pass a level, I always think that why he/she sure those are the best moves 🤔

🤔🤔

I sometimes found a player wants a lucky board but also does not want the algorithm controlling everything, not this player.

But a lucky board is a kind of manipulation too

Matching either the three greens (including the wrap) horizontally or the four purples (including the purple wrap) vertically would create a match hitting that frosting in the corner, and the green wrap explosions would hit it two more times.

I agree that your arguments are valid regardless.

Which one? I didn't see it. As a matter of fact this does not change my argument of course, since this was level 288! A genuine new player at level 288 cannot be expected to find that one hidden winning move.

I fully agree with your message @christinewupp, but I have to say that making a different last move would win this board.

Comments

Apparently there is a depth to replies and I cannot reply to your last message.

I was just gonna say, "lucky" board or a "oh just one more move sucker" board. 😁

Basically what likely happens is

Basic templates for all the level types, blockers, candies and such are stored on the device.
There is some code to get the template and all the other stuff needed for each level, this is not actually much data per level
There is a base algorithm stored on the device.
The rest of the algorithm (multiple) are also stored but there are much smaller than the base one.
They are applied to a level and the level is built, in the past when loading, now I suspect a few are built as you play in the background as that would be faster for loading.

So likely the whole process is

A template is grabbed
A manifest file is read and all static parts are put into place
The algorithm part is selected and added to the base one.
Any difficultly or unique algorithm info is then added.
The level is generated in reverse to line up with the algorithm
We play it.

There is a fair bit more to it I suspect but that would be the general gist of it. This way it would not take up too much memory and if done while playing a level, level would load faster.

https://community.king.com/en/candy-crush-saga/discussion/comment/3434695#Comment_3434695

It may be that the overall problem is that the AI that writes these algorithms cannot factor in the way humans play.

https://community.king.com/en/candy-crush-saga/discussion/comment/3434073#Comment_3434073

Matching either the three greens (including the wrap) horizontally

This is a perfect match because the main point is:

You have also created a purple fish and this fish will be activated by the explosion. It will destroy the frosting.

Your structured feedback really helps highlight where things are working and where fine-tuning might still be needed, it’s very much appreciated

Objectively speaking 66 to 900, the whole thing is all over the place.

Also the game simply needs a new algorithm. Currently it is pretty clear it just checked to see if you made the move it expect regardless of how insane it might be. This is a terrible design.

It needs a whole re-write to:

Generate each level:
1. and play it a number of times to teach the AI. Does not have to always been done just for "test users" for levels
2. take info from what moves people do and figure out the general idea of what is thought of as a good move
Weight all the items on the screen
Figure out what moves have the most weight, ie lead to clearing the board
Reward high weight moves and re calculate as the cascade happens (since the result is already fixed and not random).

Basically the boards would reward making what most would consider a good move not some random "should do nothing" move halfway across the board.

https://community.king.com/en/candy-crush-saga/discussion/513011/constructive-feedback-from-new-player-we-need-strategy-and-lucky-board-algorithms-back-please

I am not 100% convinced either move would not just invalidate the first explosion on the pink frosting. I am almost positive I have seen such nonsense.

QueenB

As always Christine, thank you so much for taking the time to play through and document these early levels so carefully… your detailed notes and screenshots are incredibly helpful.

Your structured feedback really helps highlight where things are working and where fine-tuning might still be needed, it’s very much appreciated 🫡

https://community.king.com/en/candy-crush-saga/discussion/comment/3434073#Comment_3434073

When I read a complaint in Other Places about moving every best possible move but cannot pass a level, I always think that why he/she sure those are the best moves 🤔

🤔🤔

I sometimes found a player wants a lucky board but also does not want the algorithm controlling everything, not this player.

But a lucky board is a kind of manipulation too

I agree that your arguments are valid regardless.