AI now making match 5?

Phantron wrote:

I still don't get why people always think match 5s non row/column are better than match 4. It's about equal so if you're not going to sit around to calculate the board you might as well make do it, but there are a lot of time I ended up with a crit tile in the wrong place that I can't get rid of easily that later came back to haunt me.

It's definitely not equal. Given a random board, you can calculate the likelihood of a cascade chance by evaluating how many tiles meet new partners after the match.

4 match row - On average 15 new tile meetings.
If on rows 2-7:
-8 tiles above meeting 8 tiles below
-8 new tiles falling on top of the old top row.
If on row 1(top row)
-only 8 new tiles falling on top of the old top row.

Average = (16*7+8*1)/8 = 15

4 match column - On average 14 new tiles meetings
If on columns 2-7:
-8 tiles left of column meet 8 new dropped tiles
-8 tiles right of column meet 8 new dropped tiles
If on column 1 or 8:
-Only 8 tiles left/right of column meet 8 new tiles

Average = (16*6+8*2)/8 = 14

5 match L shape - On average 20.8 new tile meetings
If L is on very bottom row:
-8 tiles left of the L meets 8 shifted new tiles.
-8 tiles right of the L meets 8 shifted new tiles.
-Because L is uneven, 7 tiles in the center column above the L meets 7 new tiles.
-3 tiles falling from the top onto the old top row.
If L is on rows 3-7:
-As the L is moved upwards in rows, you lose 3 new meetings (one left and one right of the L, one center above the L) per row from 8.
If L is on rows 4-7:
-3 tiles meetings as the tiles above the L drop down to meet the tiles below the L
If the L is on row 3:
-2 tiles meetings as the tiles above the L drop down to meet the tiles below the L

Adding this up you get:
L on row 8 (bottom row): 8+8+7+3 = 26
L on row 7: 8+8+7+3+3 - 3 = 26
L on row 6: 8+8+7+3+3 - 6 = 23
L on row 5: 8+8+7+3+3 - 9 = 20
L on row 4: 8+8+7+3+3 - 12 = 17
L on row 3: 8+8+7+3+2- 15 = 13

Assuming L shapes can appear equally on row 3-8, then the average new tiles met for matching an L shape is (26+26+23+20+17+13)/6=20.8.

So on average matching an L shape will produce 25% more new tile meetings than using it to match a 4 clear.

You can do the same analysis for the various rotated L shapes and T shapes. But my guess is that they're also advantageous, with the exception of maybe an upside down L or T exactly on Row 1.

In addition a 5 match gives you:
-Added Crit block wich deal additional damage and can cause another cascade.
-A free turn, which leads to at worst +2 AP (Leaving you only one less AP than a 4 match), but has the possibility of giving you another 4 or 5 match + cascades.

So if you're playing randomly without strategy, it is almost always beneficial to take the 5 match over the 4.

Naturally if see that the 4 match will guarantee a cascade, and the 5 match won't, then you should do what's optimal.

*Note-It's actually a little bit less for the L since I didn't include when the L is adjacent to the right edge, which is minus 8 meetings and occurs 1/7 of the time. So -1.14 tiles on average, giving 19.6 meetings on average. But the conclusion still holds.

JCTthe3rd81 wrote:

Seriously though, the Ai does seem to have gotten smarter. I swear the thing has started to deny me the colors I need, even if it doesn't use them.

It does actually do this. If you have an ability that requires three matches (7-9 AP) and make two matches of that color, the AI will try and match the same color to prevent you from getting that last match to power the ability. It doesn't happen 100% of the time, but it's pretty often. Try it right at the beginning of a match if you have the right board, and you'll see what I mean.