This summary covers Game 3 of my best of 7 series against the Words With Friends® AI. I won Game 2 by a score of 391-333 to take a 2-0 series lead. Most of the explanation of how this series works can be found in the Game 1 summary. Other than that, let's get into the moves for Game 3. Note that for games 3 and 4, the Words With Friends® AI opponent will be making the first move of the game.
25 points. Winning 25-7.
Our opponent opened the game with GONE for 7 points.
This is a pretty modest way to start the game, so let's assume for now that their other letters were a bit funky, or something, precluding a better word.
We have a decent distribution of letters, but we also have the J and K, which is a mixed blessing right now. It doesn't appear that we can make a very long word, or take advantage of the premium squares, so this is where we can use the fact that JO is a valid word.
We can make 25 points by playing JO in two directions, and drawing with EIKTT should hopefully yield a decent word next turn.
32 points. Winning 57-29.
Our opponent has played PYE for 22 points. So far, this is looking like it might be a tighter, more closed game than Game 2, and more in line with Game 1.
The first place I'm looking here is at the DW square preceding GONE. We drew an A, which allows us to make AGONE. Anything else we can make with the A will also be a double word, so we really have carte blanche to play a strong word here. The only thing that might point us elsewhere, would be if we could play a long word that started with P, ON, or JO, but a quick glance confirms that is not possible.
It would be better if we could make something longer than four letters to go with A, but we can't. Our choices (4-letter words with a K) are as follows: IKAT, KETO, KITE, TAKE, TEAK, and TOKE.
The exact word we play should be dictated by leaving the best three letters on our rack, since the point value is identical. Our choices of 3-letter stems are EOT, AIT, AOT, and IOT. Of these, EOT is the best, followed closely by AIT. So we play IKAT, for 32 points, to hopefully set up our next turn.
20 points. Winning 77-44.
Our AI opponent plays JOW for 15 points. This game is definitely shaping up to be a far cry from Game 2.
We don't have a 7-letter word we can make, and we can only play outwards in a few directions. Unfortunately, our opponent can probably make an overwhelming response to whatever modest word we play. Expanding the board up or down from IKAT will cut right into the TW/TL danger zone, which could lead to disaster. Playing to the right from the P in PYE will do the same for the TW/TL/DW column in that direction, and we can't make a word that will hit the DW.
Given these problems, our opponent's move begins to make sense. They don't want to play out into open space any more than we do, unless there is a very good score to compensate. The fact that they played one W plays into this hypothesis. W can be an awkward letter to integrate into a long word.
If it is true that our opponent played W to hold their ammunition for a long (7-letter?) word, where would they be likely to play such a word? Downwards towards the IK in IKAT leaves room for only six letters, and does not hit any high value squares. Downwards from the T in IKAT could work, but only A, E, I, and U make a 2-letter word with T, and the word would need to start with one of those letters.
If our opponent has an S, they could add it to PYE and play rightward, or add it to AGONE and play downward. They could add an S or L to JOW and play rightward. Or, they could play anything that starts with P.
If we played JOYED by adding ED, it would block a couple of these opportunities, but it would also use our best 2-letter stem and leave us consonant-heavy. Another option would be to play GOT or GOD from the G in AGONE. But this would only anticipate one of the plays from our opponent, and would still leave us consonant-heavy.
After considering these options and ruling them out, I start searching for 4-letter words with an I in them, to offload some consonants and use the DW to the left of the I in IKAT. A decent word to play here is COIF. It makes 20 points. It has no prefix that can be exploited to play the TW. And it offloads two unwieldy consonants so we can draw at a 7-letter word next turn, to hopefully break this game open a little bit.
20 points. Winning 97-60.
Our opponent has played POLLED for 16 points.
This is exactly the kind of move we were hoping to avoid last turn, and I believe it is a mistake. We've already covered a couple of other moves like this earlier in the series, so I won't go into tremendous detail on why it is a mistake just now.
However, it is only a mistake if we can exploit it. Let's see what we can do.
There's no word we can make that is long enough to use the TW square above the E in POLLED. We could play REDDLE from the DW down to L, or we could play REEDED across the E, to hit another DW. Since this word is less than seven letters, and our letters are all low value, this doesn't really hit our opponent as hard as could have been the case.
REDDLE is more dangerous for us to play. T is a valid prefix and our opponent could use it to play across the top row, hit the TW, and rack up a pretty solid score. Therefore, let's go with REEDED, which nets us 20 points.
39 points. Winning 136-71.
Our opponent has played DUAD for 11 points. The AI played a decent game in Games 1 & 2. This time, though, there have been three moves already which have really raised my eyebrows.
In this case, in return for a mere 11 points, our opponent has opened the last column of the board to two different TW squares, with one of the most common letters of the alphabet. This decision makes absolutely no sense to me. But in any case, let's analyze our own options.
We have finally drawn a pretty decent rack this game, and can make the word SIXTIES outright. We can append an S to AGONE to play this word, and score 59 points. This is probably going to be our move, but let's quickly see if we can do anything with the second D in DUAD first.
The only other option which might have value is to play DIXIT. This would only be worth 39 points, but it would also block the TW in the lower-right, and if it's still open next turn, we would append the S for DIXITS while making some other, hopefully longish word.
Is it worth giving up 20 points, and the thrill of making a bonus, in order to block a potential counter-play through the TL/TW zone on the right? Having been victimized by a couple of monster plays, i.e. 80+ points, I believe it is. Let's say that 33% of the time, our opponent makes a monster play and scores 75-100 points or more. And that they make a normal word the rest of the time, for 20 points. And let's say we can basically forestall the monster word by playing DIXIT. This would save, on average, 25-35 points. Since we are also in the lead, and have more to lose than gain from volatility, let's go ahead and play DIXIT. It is a close call, but my gut tells me to do it here.
39 points. Winning 175-94.
Our opponent has played DATE for 23 points. This leaves the bottom of DIXIT open, and we also maintain the option of using the R in REEDED to play towards the TW there. Let's see what we can do.
If we tried to combine DIXITS with another word, there is nothing we can make that is longer than five letters, which means we cannot use the DW square. At the same time, we're opening a TW square for our opponent, which they would likely combine with a TL square. Based on this, I don't think it's advisable to play in the lower-right corner this turn.
We could try to use the C to the left of the board, and use the DW there. Our best word to do this is probably CLEVIS, but it makes a modest score and also has the problem of offering a lot of counter-play.
We could also try to use the first L in POLLED. Our best options here are VILELY or EVILLY, which both make 28 points and leave us with two S tiles. This isn't terrible, but let's see what we can do with the R in REEDED.
With an R, we can play VERILY into the TW, which blocks that square moving forward, while giving us 39 points. Let's go with that option right now.
47 points. Winning 222-130.
Our opponent has played YUTZ for 39 points. Most of this came from getting their Z on a TL square.
We now have two S tiles and a blank. However, we are just barely unable to make a 7/8-letter word, and our situation is a little more tricky than it first appears.
We would like to avoid using an S or a blank on a pedestrian, 25-point word. These tiles can have enormous value if played in the right way. However, we have a limited number of places where we can append an S on this board.
One option would be to play something like SUBPAR at the end of JOWS. This makes a decent 32 points, but it also uses an S and a blank, and seems a bit underwhelming for something that uses those.
Eventually, however, I realize that we can make a word with two S tiles, creating IKATS and JOWS in the process, and probably rack up a much better score. Sure enough, we can make PURSES for 49 points, as pictured. However, this opens up the first column to a strong response from our opponent. If we content ourselves with PUSES for 47 points, we protect ourselves much better.
Our opponent could have an O, making OPUSES, along with another word, and it would hurt. However, there are only four O tiles left in play, so let's hope that we can fade this possibility.
26 points. Winning 248-164.
Our opponent has played VINO for 34 points. This opens up the bottom row of the board, and could have been an impetuous move depending on the tiles that we hold.
However, we have managed to pull seven consonants. We do have an S, but there is not even a 2-letter word we can make to use the TW square, or to just deny it to our opponent. Thus, we have to think of other options.
I don't like the idea of exchanging here, in spite of our terrible letter distribution, because it is simply too risky to give our opponent a free shot at the TW or DW/DW rows on their next turn. Even if we can't make a great score, we need to muddle the lower-right corner of the board so that we can remain in the lead.
So, let's look for words that can use the O in VINO, on the second row from bottom. These will need to be words with the O as a lone vowel. And, if possible, they should use an S on DIXITS, and it would be even more fantastic if they could use the DW square. This is what we will be searching for.
Unfortunately, we can't make a word that uses the DW square. However, we can do something that plays DIXITS, while somewhat obstructing the TW. BLOCS is the only choice which does not use both of our S tiles. The B and L tiles don't make it impossible for our opponent to play the TW, but they should hopefully limit the options to something less damaging for us.
We are still likely to be consonant heavy next turn, but we might get lucky. And if not, it should be safer to swap some times next time.
28 points. Winning 276-228.
Our opponent has played MILT for 64 points, doing basically the exact thing that we were trying to prevent last turn. Oh well, that's how it goes sometimes.
This does make the game a little more interesting. I was afraid this was going to be a yawner, and I was starting to doubt the Words With Friends® AI in general. Now we begin the turn with a narrow 20-point lead.
The draw went pretty well for us, and we have three vowels on our rack. However, we can't use them to make a 7/8-letter word. Since we still have an S, let's begin by seeing where we can play it.
Our choices this point are basically to make JOYS, AGONES, or VINOS. JOYS would be a promising start, but there's no way to build a long word out from there, so our final point value is limited. With AGONES, we'd need to make a second word which begins with S. It would be possible to do something like SHARN, but it would only make 18 points, and possibly do more good for our opponent than ourselves.
So let's look at making VINOS. We could then make BA, forming an ATS ending. If we prepend HE, we make HEATS and put the H on a TL square, while making two other words in the process. This is worth 28 points and gives up limited counter-play.
31 points. Winning 307-243.
Our opponent has played HIND for 15 points. Charitably, we can call this a holding move while they see how the rest of the board develops. Perhaps they were also consonant heavy and trying to fix their distribution.
We actually have an interesting situation here, positionally. There is one clump of letters to the center-left that was built early in the game. This was opened up when our opponent (unwisely, I believe) played POLLED towards the right of the board. Now the right side of the board is pretty close to saturated. So the question now is where the action will drift to next.
We would like to draw an O so that we could play OPUSES and run something down the left-most column of the board. Conversely, every turn that this area remains open, there is the risk of our opponent doing so as well. We could make a small word (AN, perhaps) by the P and U in PUSES, but it's unclear if this will improve the situation much defensively, and it would not make very many points for us.
The top third of the board is remarkably well blocked. The C in COIF and V in VERILY contribute to this. The word HIND has no one-letter prefix, and playing a word like BEHIND into the top row would be almost suicidal. The same could be said for playing a short word into the E in VERILY. There's no way to score more than 10-15 points there, and the response from our opponent could be worth 4-5 times that much. Playing a word into the O or F in COIF, or into PYE, is also unlikely to be worthwhile. In short, the rest of this game will be likely fought in the bottom half, and it could become a pretty closed affair.
Ideally we could get in a long word now, perhaps into the H in HEATS, or from the L in POLLED. But our letters don't line up for this. In fact, I'm willing to admit that we seem a bit stuck here, at first glance.
One possibility, though, is to play WHEATS by prefixing W to HEATS, and make a second word ending with W. If we could use the DW square in doing so, we might make our best score. But is is possible?
As it turns out, it is. BRAW is a valid word. It makes a good score, 31 points to be exact, and does not leave us too exposed. The TW on the left side of the board is far more accessible to a player with an O than to someone playing an 8-letter word ending in the B we have laid down. In fact, there are only 23 words in the entire Words With Friends® dictionary that have eight letters and end in B.
For these reasons, we play BRAW. Hopefully we can pull an O before our opponent does.
66 points. Winning 373-269.
Our opponent has played TRAVELER for 26 points, weaving between two established words.
Of course, I spent the previous turn writing a paragraph on why the top third of the board was likely to remain closed off, and the AI opponent has immediately proven the opposite. However, I again believe our opponent has made a big mistake (as with POLLED), and this time we are well-placed to exploit it.
Needless to say, the first place we'll be looking to play is on the T in TRAVELER. If we had space to the right, we could play TRIMARAN, which would pretty much put this game in the books. However, we have to settle for MANTRA, making 66 points.
Again, not to hammer on an example too hard, but our opponent would have been better off not even making a move last turn, than to play TRAVELER for 26 points. By setting up a 66 point move for us, our opponent actually cost themselves 40 points in this exchange! If they had literally passed and done nothing last turn, and we had played a 35 point word, our opponent would still be better off.
As it is, our opponent's chances of getting back into this game are almost destroyed. They were recently as close as 20 points behind. Now they are losing by over 100 as we approach the endgame. If you are in a similar situation, try to be more cautious about extending the board for moderate value.
26 points. Winning 399-317.
Our opponent has played HONE for 48 points. Just as we finally drew an O tile (two of them, actually), our opponent has grabbed one as well, and played on OPUSES, as I was explaining earlier.
48 points is not bad for a move, but our opponent really needed better to get back into this game. Perhaps they knew that when they played TRAVELER. If they kept the O while they played the other six tiles, it's possible they were hoping to draw the Q and make some kind of monster word. Or maybe not.
In any case, we're going to win this unless our opponent can get to some premium squares and rack up another two moves like the one they just made. We don't have a knockout, 7-letter word to play from our rack, so let's think about where our opponent has the best chance of playing.
There is a TW and DW in the lower-left quadrant of the board. There's no chance of the opponent hitting the TW in one turn (note that they could play BUCKWHEATS, but the letters are no longer available). The DW is open, but there's probably not a great score to be made.
The TW near COIF is pretty well blocked right now. Unless we play into that area and give our opponent the chance to reply, we don't need to worry there. The TW by MANTRA is closer to the action, but still not immediately playable.
Our opponent's best chance probably lies in using one or both of the DW squares tucked by MANTRA and TRAVELER. If we can block these up and confuse the situation here, we can probably close the game out. So let's think about how to do that.
Our letters make good 2-letter words with the MAN in MANTRA, so we have a good opportunity here. A lot of this is just a matter of knowing the words, and arranging the letters in different combinations, but if we play GOOIER, we will make 26 points and use one of the DW squares.
This definitely lowers the value of the DW square below AI. There just aren't many 4-letter words starting with AI and none of them (AIDE, AILS) are valuable at all.
However, we should consider the fact that we're opening the TW square by the G in GOOIER. Is this something we should be worried about? Let's consider how our opponent might exploit this.
First of all, there is only one way our opponent can get to the TW -- they must play a word of five letters or less which ends with A. AG is the only 2-letter word ending in G, and there is no single-letter prefix to GOOIER that would allow our opponent to play vertically.
Now let's look at the Tile Bag and think about the worst-case scenario. First of all, the only letter that matters at this point is the Q. No other letter, even tripled, can significantly cut into the 82 point lead that we'll have after this turn. So we're really looking for 2-5 letter words that end in Q, that our opponent can plausibly play. The three candidates are AQUA, QUA, and QUOTA. AQUA and QUOTA (using a blank for O) would both score 46 points. Even if they played QUOTA, we could easily make the other TW in this corner unplayable, and our opponent will be left with just a couple of final letters.
The only other miracle which might save our opponent, would be for them to have a 7-letter word on their rack which ends in A, or AA/AE/AS/AU, such that they could hit the DW and play into BRAW. If this were the case, they could score a well-deserved victory. But they could more easily have a word ending in ER to hang under MANTRA, which we are blocking now with GOOIER.
We would probably win no matter what, but GOOIER gives us enough points, and blocks enough of the board, that it should close this out for us.
GAME OVER. I lost 384-378.
Well... Our opponent played URETHRAE for 46 points, using their last seven letters. This brought the game to 399-363, with me still in the lead. Then, however, the letters on my rack were subtracted from my score and added to my opponent's. Since I drew the Q and F, among other letters, and were stuck with them, I lost a massive 21 points, dropping me to 378. Our opponent then gained 21 points, bringing them to 384. Game over, they won.
After trailing for the entire game, and pretty badly towards the end, the Words With Friends® AI made up an 82 point deficit on the final move of the game. It would be tempting to chalk this up to bad luck and move on, but let's reexamine what I did in my last move, and whether it was the best choice.
I actually made a subtle, yet critical mistake last turn. When I played GOOIER and drew six letters, I left exactly zero letters in the Tile Bag. Had I settled for a 4-5 letter word instead, we would still be able to close the game out right now. Can you see how?
Let's suppose that we would have played GOER instead of GOOIER, making ME and AR in the process. We would have made only 19 points, but we would have achieved about the same thing defensively.
However, once the AI played URETHRAE, for the same 46 points, the game would not be over! There would still be two tiles in the bag, and we would have a chance to unload at least a couple letters from our rack. Getting rid of the Q alone would have comfortably assured a victory.
In short, there is a lesson to learn from this. If you are nearing the end of the game, try not to play a move which leaves 6-7 tiles in your rack, and zero tiles in the Tile Bag. If your opponent can find a way to lay down all of their tiles, you could be handing a lot of points to them. Even in this game, which I dominated from start to finish, this one mistake was enough to undo everything.
Hopefully this game has provided some object lessons on a couple of points. Most obviously, how to manage the endgame and number of tiles left. And beyond that, the perils of playing a word into open space on the board, without adequate compensation in scoring. I made the first mistake, our opponent made the second mistake on a couple of occasions.
I'm now winning the best of 7 series by a score of 2-1. The player who has played first has won all three games. I'm obviously disappointed to have squandered victory in this one, but on the other hand, this guarantees that there will be at least five games to write about in this series. I'm definitely looking forward to Game 4.
The words in our dictionary are meant to match, as closely as possible, the words used in Words With Friends®. We have no affiliation with Words With Friends, and offer this site for entertainment purposes only.