First of all, I want to say that I am quite happy with the 1.5 update. It’s true that I feel that some of the changes have been detrimental to my way of playing. Losing chip damage, for example, has made my battles go longer and more difficult, but I believe it makes more sense and captures better the original puzzle quest feeling. All in all, I still enjoy the game and I hope I will keep doing it.
However, I wanted to voice my concern about how the autoplay feature feels much worse after the 1.5 update. I love having such a feature, and I understand that it wouldn’t make sense if the AI was 100% efficient. However, now that chip damage has dissapeared and only skulls and spells do damage, it becomes really frustrating when your spells are full, the board is full of skulls, and the AI keeps matching gems turn after turn doing… Nothing.
I would really be grateful if developers could look into autoplay and adapt the AI to the changes brought by the 1.5 update. Something as simple as increasing the frecuency at which the AI uses the spells, or even using them as soon as they are full as it happens with enemies, would help a lot.
Thanks a lot for all the work you do to make this a game to enjoy.
It’s absolutely tragic how bad the auto-play is now. The AI now takes turns for the enemies, and it generally makes logical moves. I’ve seen it do an occasional bad move, but not super often. Meanwhile our version of the AI seems to purposely ONLY make bad moves. It’s obviously possible to code the AI better (since it works for the enemies) so this needs fixed for auto-play.
I also honestly feel like the AI that plays for me on auto battle has gotten dumber. I sit and watch it now and I notice it match colors that have nothing to do with my build and make stupid moves matching let’s say three yellow when I am an obvious mono color green built character and there are 30 green matches to be had on the board that’s just ridiculous!
plus it never uses my spells when they are full and causes me to die when I could have destroyed the enemy five turns ago if it would have just used my spells.
it definitely needs some work that’s for dang sure!
After a very brief (totally unscientific) experiment…I have noticed that using AP, auto-play seems about as dumb as it has always been. Possibly, slightly less dumb. It was still making weird matches but it was casting my spells. It was casting them stupidly, but hey… it was casting them!
Meanwhile, using timer, omg. It was often avoiding any type of gem match that made any sense at all. I’m mono purple, enemy is mono red. Auto me matches blues, yellows and greens, ignores red, purple and skulls. It would match purple gems very infrequently, most mana was gotten via gem fall. When my spells were completely full, it would continue matching gems and avoiding skulls. I won the battles just because of random skull hits from gem fall. That was pretty much the same as pre-patch, heh. However, pre-patch, we also could count on random gem-fall chip damage. So it’s not really dumber, it is just much more important that it NOT be this dumb?
I agree. I believe that the problem is not the AI getting dumber, but the lose of chip damage making it much worse than it was. Before, as long as I was above enemy’s geascore I won most times using the autoplay. Now, unless I am significantly higher than the enemy, the AI makes me lose most times.
I hope you are right and the issue has been passed down. Something as small as making the AI use the spells when they are filled up (as enemy AI does) should help a lot.
I usually use timer. After reading what you have described, I have tried with AP and it may be true that the AI casts spells more frequently, if only because it does so between moves. However, it has also spent several turns without casting any of the full spells, so it does not seem to be much better. I will keep testing AP, since a small increase in spell casting frecuency can result in a significant improvement.
However, if there are indeed differences of AI behaviour between timer and AP I believe it should also be addressed. Having to change the mode depending on if I am playing myself or using the autorun does not seem confortable.
That’s pretty much my assessment as well, as someone who stares at a ton of AutoPlay frequently.
AutoPlay is more or less just a whole bunch of random dice rolls.
Roll dice to decide whether to cast a spell of match gems.
If casting a spell, randomly roll to decide which spell to cast. Go back to the top of the decision tree.
If moving gems, randomly roll to check to see if an activatable ultimate gem exists on the board. If so, activate it and go back to the top of the decision tree.
If there are no activatable gems, randomly roll to pick a gem that can be moved to make a match. If the gem can be moved in more than one direction validly, randomly roll to see which direction the gem will be moved. If playing on Timer Mode, repeat this step to keep moving gems randomly in valid directions until time runs out or the board is fully resolved. If playing on Action Point Mode, if there are still AP left to spend and the board is not fully resolved and there are no spells left or ready that can be cast by the player to modify the game board, go back to the top of the decision tree.
I think this is what both you and Higure are observing. In AP Mode, the AI has more opportunities to cast spells, where it cannot in Timer Mode once the first gem move is made that turn which significantly handicaps the AI algorithm for the rest of the turn.
Also, I think there are a couple of further confounding factors in 1.5 that have drawn a lot of additional attention to the situation:
The board is larger, which gives the AI more room to make obviously bad matches. Pre 1.5, the board was small enough that a decent end-game build had enough time to make matches (especially with the frontloading of match time on the first 3 matches per turn) for the AI to slowly fully resolve the board. So, even with the bad AI, through brute force the board would be resolved as reasonably complete as possible even if the matches weren’t optimal. That was acceptable for most players. In 1.5, the board is large enough that even with 3.5 or 4.0 base seconds match time, it isn’t enough to resolve the board. As result, players are observing a lot of bad matches being made by the AI, leaving reasonably expected matches to be made on the game board.
Opponent AI exists in 1.5, which brings a lot of attention to how well the Opponent’s AI plays the game board versus how poorly the same AI plays on the Player side of the game board.
^^^ This.
Obviously, the Player AutoPlay AI shouldn’t strong enough to be superior to manual play, else why play manually at all. I think most of us here just want it play as well as it did pre-1.5.
From personal experiences in 1.5, I’ve had to dial back what I can farm on Auto and/or have to keep half an eye on the Auto’d match just in case I have to manually step in and bail out the AI because it stubbornly refuses to cast spells almost entirely, (I’ve seen multiple 8+ turn streaks of full spells and ultimate just left to rot until the Player AI loses from making matches that do absolutely nothing).
Ya, just to clarify further, even with AP the auto-play is dumb as a box of hair. I’m just finding that it’s working significantly better with AP compared to timer. The combo of the loss of chip damage, poor matching, AND almost zero spell casting, makes timer auto-play a complete disaster. The fact that the AI will at least cast some spells using the AP system makes a big difference. But yes it’s still not where it needs to be, at all.
Thank you very much for such a detailed and useful analysis.
If autoplay decisions work as you say, it makes sense that using AP will improve the spell casting frecuency.
I agree that the autoplay AI shouldn’t be as good as enemy AI. Before 1.5 update it was still poorly efficient, but it was useful enough that we could rely on it if our gearscore was above the enemy one. We don’t ask for the same AI as the opponent uses. We just ask for a change that will make autoplay usefulness similar to how it was before the update.
As a follow-up, after several days of testing, AP Mode is currently vastly superior to Timer Mode for using AutoPlay.
This is because the when using AP Mode, the decision tree on what the AI will do is rechecked from the very top after every AP point spent in that turn, whereas in Timer Mode the checks to cast spells stop after the first gem movement for that turn.
Just remember to turn the game rules back to Timer Mode (if you use that) afterwards.
