tile matching strategy

Options
Unknown
edited February 2014 in MPQ General Discussion
I've been thinking about the matching and tile drops, and while I don't do this for every move, sometimes you're in a bad spot and need to find the best move even when there seems to be nothing. That is, there are no match 4/5 and you don't even see anything that can cascade into a second chain, so what should you do?

You should assume that the unknown tiles are favorable, but favorable tiles are only useful if you allow it to. For example imagine you're matching RRR at the top row which looks like this:

RRRGBGBG
YPYBBGGB

No matter what dropped in place of the RRR, it's not going to helpful unless it's something like XGG, or 3 of the same color. That's obviously asking way too much luck.

Now let's assume we have the top row looking like this:

RXR

You're making a vertical match on the X column. There's a 1/7 chance you'll match the two Rs on the side, and while there's an equal 1/7 chance that a R will drop a row below, that's probably okay because the computer has to match something so it's an okay risk to take. Most of the time they'll match something you're not expecting to anyway. So in this move you have 1/7 chance of getting an additional cascade at no real risk.

But what if the board looks like this:

???????????
RRXRR
???????????

Again you're matching at the 'X' spot. Now, if the R drops (1/7 chance), 2 out of 3 times it'll drop above or below the row in question, which results the computer immediately turning that around for a match 5. 1 out of 3 times it'll drop into the right row and get you a match 5. This is a bad move and should be avoided, unless you're desperate. If you're ahead by a comfortable margin, you should consider just match the RRR so that the computer couldn't even try to take a chance and turn the game around with a match 5.

In fact, from this analysis we can conclude that when no obviously better moves than a match 3 are available, the best moves are matching near the top on something that looks like RXR (where you're matching the X column), or something that looks like:

RRRGGYYB (you're matching the RRR)

For the similar reason that if you get lucky, the new tile will drop a G and get you a match, and if the G happens to land in the wrong spot, the computer has to match something on their turn anyway. On the other hand, things that can form a match 4 or 5 based on unknown tiles are BAD. There's at best 50% chance you'll get the drop (if it's the top row), and it can be as low as 33%, and you can be sure if the tile you need dropped on the wrong row, the computer will always flip that around in a match 4/5 (unless there's another match 4/5 on board). Pretty much anything that could form a match 4/5 based on new drops can be broken up, and should be done so if you're firmly in control, because you do not need that extra match 4/5 if you're already firmly in control (so you don't want the computer to do that match), and you certainly don't want to be the one doing the matching (gives computer better odds to get a match 4/5). The logic is reversed when you're losing, as even 33% chance for a match 5 (and 66% chance for computer) should be taken if the situation looks hopeless already.

Comments

  • mischiefmaker
    Options
    Nice writeup, solid advice.

    Also worth noting is that if your team is weaker/you are losing, you should prefer higher variance matches, while if you are winning/have a much stronger team, you should prefer matches that lead to lower variance for your opponent. What this means in practice is that if you're losing (and you don't have abilities close to being charged that you are confident can turn it around), you want to prefer matches that are near the bottom of the board, because they shake up more of the board than ones at the top. If you're winning, you should prefer matches that leave the opponent only matches near the top, since they're less likely to lead to big cascades.

    Of course, if you're winning/stronger, but you prefer faster fights over the risk of having to eat an enemy ability or three, then high variance is better.
  • Someone's been reading Malcolm Gladwell's new book I see! icon_e_biggrin.gif
  • Whenever you're ahead you want to minimize risk. In theory, you could just figure out the outcome of every possible move, but I doubt anyone has the time to figure that out. Therefore, you want to minimize the amount of unnecessary board movement. For example, making a match 4 always causes a large board shakeup, but this is worth it since you get a boatload of AP for doing that. On the other hand, making a horizontal match 3 on the bottom row also moves the board considerably, but you only get 3 AP for that. If you're not sure all those movement are totally safe, it's best to avoid it.

    Likewise, when you see a large clump of similar color tiles, it is best for you to break that up with a match 3 if you can't figure out how to turn it into something useful, because again you don't want the computer to accidentally stumble into something and turn the game around. Yes, you might accidentally stumble into an awesome cascade as well, but if you're already winning, you've no need to take that risk.

    The logic is reversed when you're behind. If you're pretty sure you're going to lose playing safe, you might as well take extra risks. Earlier today I had a game against Doom/The Hood/Magento where I am down to my last guy with 200 HP, and he has about 1000 points of attack tiles on board. I figure I need at least a 5 chain to have a chance of winning, and obviously didn't see any, so I just went with the lowest horizontal match 3 and it somehow cascaded 5 times and won. Since I'm literally dead next turn anyway, there's no additional risk for that move even though I have no idea what the outcome is.