Sample Game 2

This summary covers Game 2 of my best of 7 series against the Words With Friends® AI. I won the first game by a score of 391-333. The overall explanation for this series can be found in the Game 1 summary. For now, let's get into the moves for Game 2.

Move 1: TOLANES

53 points. Winning 53-0.

We start with a tabula rasa and a pretty good rack of letters. Sure enough, a quick search tells us that we can play either ETALONS or TOLANES for a 7-letter bonus (Disclaimer: I would have never found either of these words without a lookup tool.).

The only question we have is which word to play and how to play it. We'll score the same no matter what -- we just want to minimize the opportunity for counter-play.

One thing we don't want to do is play it so that the S is on the DW square, aligned with the TW at top and bottom of the board. We cover this in detail in the section on the Opening, but in short the S leads to more 8-letter word opportunities by far than any other letter.

From the TW standpoint, it is much safer to play such that the O is on the DW square. There are far fewer 8-letter words that begin or end with O. However, if we put the O in ETALONS on the DW, this would also put an S between two DW squares. This offers our opponent the chance to quadruple their score, and is perhaps even worse.

However, if we play so that the O in TOLANES is on the DW square to the left, we get the same benefit in preventing a TW play, while preempting the chance of a 2x DW counter as well. This is how we will play our first move.

Move 2: JOY

34 points. Winning 87-18.

Our opponent has played ODYLE for 18 points, leaving us with a clear lead in the early going. However, there's no hope of making a 7-letter word with the rack we have drawn.

The good news is, we will have a lot of high-value letters to work through. The bad news is, they do not coordinate well together. Since there are no serious threats at this point, let's just focus on getting a good score this turn.

Let's start by putting the J on a TL square to make JOY. This will be a decent move while we see how the board develops.

Move 3: VIBIST

52 points. Winning 139-38.

Our AI opponent has played HIGH for 20 points, while opening play on the top row of the board.

There's no other way for me to put it -- I believe this a highly questionable choice. We can now easily play on a TW square, or play a word that hits two DW squares (for a 4X effect). Our opponent did not receive very much in return for giving us this chance. They could have likely scored 20 points almost anywhere.

If we use the I in HIGH, we can play VIBIST for 52 points. We also minimize the counter-play that our opponent has next turn. They will likely be able to make a 4-letter word at most, to hit the TW in the upper-right. Because of the V in VIBIST, which cannot make a two-letter word, there is no way for our opponent to play at the TW in the upper-left next turn.

Therefore, although we can expect our opponent to make a decent score next turn, they probably can't approach the 52 points we will score with VIBIST.

Move 4: NAME

40 points. Winning 179-100.

Our opponent has played ETERNIZE for 62 points. The only good thing about this is that they did not hit a DW or TW square, though they did get their Z on a DL square.

This move very much opens the right side of the board. It would not be surprising to see words leading right from the top E and the R in ETERNIZE over the next couple of turns. It would also not be a surprise to see a long word be played down from the bottom E in ETERNIZE, via to the DW square. All of this suggests a fairly open game over the next few moves.

Our best option, however, may be slightly different. We don't have the letters to play a long word down from E, and we only score around 30 playing for the TW squares (RAMEN scores 33, ENEMA scores 27). However, we can play parallel to ETERNIZE, through the DW square, and rack up more in that manner.

First, let's put an A down to make ZA. Then we can build around that to see how we can do. The only letter that goes with I is N. And, once we have NA, NAME becomes the only plausible way to hit the DW. This gives us 40 points, which is not bad.

Move 5: QUAKE

38 points. Winning 217-136.

Our opponent has played SETUP for 39 points. This opens up the right side of the board even more than it already was.

We are in reasonable condition to exploit this, although it would be better if we were less consonant-heavy. We could try to play something from the P in SETUP, or try for something through the U between the two DW squares, like we did earlier this game.

After experimenting for a bit, I have found two potential moves. One is to play BEAUT for 36 points. The other is to play QUAKE for 38. The points are about the same, so is one move better than the other?

I would make the case that QUAKE is better for two reasons. First of all, it gets rid of a Q and K from the rack and doubles their value. This is likely to make our future letter distribution a bit better, in what looks like a wide open game. In contrast, BEAUT would leave us with KQW on our rack, which is likely to lead to a somewhat ugly draw.

Secondly, QUAKE should give us better chances for counter-play on our next move. Let's suppose that we played BEAUT. Our opponent would be likely to have a word which would hit the TW space and end with ED or ER. Let's suppose they made a 7-letter word this way. If so, they would be playing right into the corner of the board, and we would not have a great chance to reply.

However, in the lower-right, if our opponent plays a word into the TW, we have plenty of chances to reply and hit the other TW square in return. It is possible they could play an 8-letter word with the X in the right place and really hurt us, but we would still have several places to come back with a long word of our own. And of course, no move is completely without risk.

With this in mind, let's play QUAKE and see what happens.

Move 6: BRR

28 points. Winning 245-171.

Our opponent has played PEAT for 35 points. Our letters wont allow for a long reply in return. In fact, there is nothing longer than four letters that we can play without using the board somehow. In this case, we must get creative.

One word we can make, which would use nothing but consonants, is BRR. This can be affixed to QUAKE to make QUAKER, and provides a respectable score of 28 points. We allow some room for a reply to the TW below B, but with INTW as our starting letters for a draw, we would hopefully be able to play a longer word in reply if this happened.

Let's play this, see what we draw, and hope that we aren't giving up too much in return for offloading some consonants.

Move 7: XI

23 points. Winning 268-198.

Our opponent has played HUED for 27 points. After the fireworks at the start of this game, things are settling down a bit.

There's not much we can do this turn to change that. We drew two I tiles and an X, and have a limited number of words we can make.

In this scenario, we have too many I tiles. We also have an X tile which doesn't combine well with our other letters. Since XI is a valid word, we might as well look at where we could play that for a decent score this round.

We could play the X alone, above the I in VIBIST for 25 points. However, by sacrificing slightly, we can play XI below TOLANES for 23 letters, and get rid of one of our I tiles in the process. This will probably be better for us in the long run.

Move 8: Exchange 6 letters (keep 1 N)

0 points. Winning 268-238.

Our AI adversary has played FINNY for 40 points. In terms of position, this play changes very little with the scope of the board.

Our letters are a little better this turn, but still not conducive to making a 7-letter word. In fact, there is little evidence of a strong word anywhere. For example, BIN into the TW makes only 21 points, and potentially opens up more space in the bottom row than such a score warrants.

Furthermore, if we look at the Tile Bag, there is reason to believe that the distribution could be favorable if we redrew. There are 22 consonants, 20 vowels, and 1 blank tile outstanding, which would suggest good chances of a favorable split. All of the high-value, yet awkward letters such as J, K, Q, X, Y, and Z are gone. And there just doesn't seem to be an imminent threat on the board that we need to defend against.

One thing we will do, however, is hang on to a single N. There are no other N tiles besides the two we have in our rack, and having one consonant should increase the chances of ending up with a 4-3 split. The hope here, of course, is that we draw something that can make a 7-letter word next turn, and that we're more than compensated for taking 0 points this round.

Move 9: ROUEN

27 points. Winning 295-249.

The WWF AI has played SCREW, for 11 points.

Unfortunately, things could have gone better on the tile draw. Even though there were 20 vowels and 23 non-vowels in the bag, we drew 5 vowels and 1 consonant. Let's use a hypergeometric calculator to explore these odds a little more.

If we begin with 20 "successes" (vowels) and 43 in the "population" (total tiles), and draw 6 times without replacement, this is how our distribution will look:

Number of Vowels% of Draws
60.6%
55.8%
420.1%
333.1%
227.6%
111.0%
01.8%

Specifically, there was only a 5.8% chance of us drawing 5 vowels, and an 80.8% chance of us drawing somewhere between 2-4 vowels, which would have been much more useful.

However, none of that changes the face that we're stuck with the tiles we actually have. One good thing, at least, is that we drew an S. Furthermore, our opponent was not able to play a very good word last turn, for whatever reason. SCREW was a low-scoring word, and does not create any immediate threats as far as opening new TW squares, etc.

So we have some options here. We could swap again. We could play a short word, just to unload a couple of vowels while taking a few point. Or we could look for a real play, where the goal is to score significant points this round.

I don't think we're in bad enough shape to consider a full swap again. But one thing we could try is to play OUR or OUT with the OU on our rack. AEINS would remain on our rack. This isn't one of the very best 5-letter stems, but it is not a bad start. So let's keep that play in mind.

One good place to look for a real play might be the DW square next to the W in screw. Let's start with AO, OA, ON, etc. and look for words we can make from those fragments.

The thing is, however, that there's not much we can make without jutting into two "tripwires" that we could definitely avoid. Suppose we played NOISE, as pictured. We would be putting an S between two DW squares, and an E on the bottom row between two TL and TW squares, and all in return for 25 points. For this modest score, we would be asking our opponent to reply with something absolutely devastating. So let's table this idea, and see if we can't do something with the R in ETERNIZES.

We would like to make a 5-letter word, beginning with R. Ideally, we would also like to avoid playing our S, since it will be much more valuable if we append it to something in a later turn. Fortunately, there is one word that meets this criteria -- ROUEN. This will score 27 points for us, get rid of some unwanted letters, and preserve the S. Let's go with this.

Move 10: FLIES

49 points. Winning 344-267.

Our opponent has played DAL, combined with AT and LOX, for 18 points.

This move officially divides the board into a top half and a bottom half. There is no prefix to DAL, no prefix to TOLANES, and no one-letter prefix to XI. Essentially, there is no word that will be played through from above DAL to below XI, which leaves us with two different areas with very different dynamics.

The top-left of the board is very much closed off. The DAL, the V in VIBIST, and the J in JOY, and even the OH (few words end in OH), all combine to limit the ways that a word can be played into the top-left. Further, the first player who plays into this area is likely setting their opponent up for a much better score on one of the TW squares. Therefore, it is likely that our action will be centered on the bottom half of the board for the next few turns.

Here, the situation is much more open. There are places to play long words starting with N, or containing an E or W. It is to this area that we will now focus.

It is after trying a few words with the S on SCREW, that I notice QUAKER. Putting an S down to make QUAKERS, we can then work backwards, making RE and BI. From here we can prepend FL to IES to make FLIES and hit the DW square. This does open up the bottom row a bit, but 49 points should be worth enough to risk it.

Move 11: AHA

25 points. Winning 369-324.

Our opponent has played PUBIC for 57 points, making good use of their C, along with the TL and TW square. Even though we scored 49 points with our move, our opponent actually gained 8 points on us last turn. We begin this turn with a narrow, 20-point lead.

We can't make a 7-letter word with our letters, and the longer words we can make end with ING, which is hard to play right now. We're also getting close enough to the endgame that we might no longer be dealing with long words, so much as looking for ways to make good scores and preserve our lead.

One way we can do this, without breaking up our ING, is to return to the idea of playing something adjacent to SCREW. This time, we can play AHA for a decent score of 25 points. Let's do that.

Move 12: IGNITES

36 points. Winning 405-342.

Our AI opponent has played WORT for 18 points. This seems like a questionable decision to me, much like when it played HIGH for 20 points earlier, and gave us a huge opportunity in return.

R and T are both excellent letters to attach a word to, and they both align with premium squares (R with a DW, T with a TW). The first thing we'll be doing is searching for words we can play with those two letters.

We barely miss out on being able to play SEIGNIOR, needing one more square on the left side. There are a few 7-letter words we can play on the R, like NOISIER or SIGNORI. So let's keep that in mind while we look at the T.

On the T, there is also nothing that uses our entire rack, unfortunately. However, we can play IGNITES, which will accomplish three good things at the same time. First of all, it will give us a solid score of 36 points. It will also suck a lot of the oxygen out of the lower-left section of the board. We are occupying one TW square without exposing the other, and we don't even expose the DW square under the D in DAL. In theory, our opponent could end a word with L and make LIGNITES, but there are no L or blank tiles left in play. Therefore, our opponent will be hard pressed to make the kind of monster word that they need at this point, if they are to come back into this game.

Finally, there is another benefit to playing IGNITES. Even though it does not give us the 35-point bonus, it does take six letters off of our rack. The game ends when one player has played all of their letters. Once we draw six letters from the Tile Bag, there will be only four letters left. This is enough for each player to make one more "real move", or two at the outside limit, before we get to the part of the game where we will be playing 1-2 tiles per word, for minimal point values. Since we will be winning the game by 63 points after our turn, we want the game to end sooner rather than later.

If the situation were reversed, and we were trailing, our bias might be to play very short words, so that we could drag the game out as long as possible, hoping for a perfect play to pull back into it. But since we are winning, our bias should be to play as many tiles as feasible (without doing anything stupid, of course), to close this game out.

Move 13: DEV

29 points. Winning 434-369.

Our opponent has played the Big MHO for 27 points. However, we still have a solid lead and we can come close to ending this game with a long word on our turn.

However, we are back in the old situation of having one consonant and six vowels. For whatever reason, in this game in particular, we have repeatedly found ourselves in situations with a very lopsided distribution of vowels and consonants.

Our first goal here is just to offload letters for anything short of a terrible score. Our second goal is to avoid exposing any of the premium squares that are now covered. If we do that, it will basically be impossible for our opponent to come back.

After some searching, it looks like DEV is a decent choice. While it does make the TW square in upper-left playable, it also scores 29 points. Furthermore, without being able to use a second TL square, there is reason to believe that the damage our opponent can inflict will be somewhat limited. None of the letters they have is of particularly high value, and if we assume they can play the full 6-letter word, they would still only make around 30 points. Furthermore, it's not obvious to me where else we might play at the moment. I can only find a score of around 10 points playing in a safer location.

NOTE: The image above covers the moves from this turn, and from Move 14.

Move 14: GOOD

21 points. Winning 455-390.

Our opponent has played ATT for 21 points. We can just continue on with our strategy of making decent words and closing this game out.

Since the TW square is still open, there's no reason for us to not take advantage. Since we have a few four-letter words that can end with D, let's also try to plan for our final move. To start with, we could play AGED/EGAD, GOAD, or GOOD.

If we play GOOD, we could hope to play GAE on our final turn. If we play AGED, we could hope to play GOO on our final turn. However, if we play GOAD, we might have to plow through two more turns to end the game.

Let's play GOOD, because we also have a couple of other places near the bottom where we could unload AE.

Move 15: EAR

13 points. GAME OVER. I won 469-402.

Our opponent has played RID for 13 points. We can close the game out by playing ARID and AE across the top of the board. Or, we can close it out for a few more points by playing EAR across the bottom. Let's go with the second option and book this win.

Conclusion/Thoughts

I am now winning the series 2-0 against the Words With Friends® AI. This game was a lot more open than Game 1, which I won with 391 points.

One thing that jumps out to me about this game is how asymmetrical the final board is. A few words early on, such as TOLANES, JOY, and VIBIST, really blocked out the top-left quadrant of the board. Later on, playing XI helped to block the lower-left quadrant for a significant amount of time. I wasn't supportive of, or opposed to, these developments. But they do show how a few well-placed letters can really alter the flow of the game. If one's goal is defensive play, this game definitely provides some examples.

Secondly, we both got in some nice, long words. There were several turns where the board was wide open, and both players could have probably booked even bigger scores with the right combination of letters. As it was, we made it into the high 400s with our score, and averaged over 30 points per move.

It is nice to be winning this series 2-0. However, I have had the first move in both games so far, which does provide an advantage. In the spirit of the best of 7 format, in line with the World Series and NBA Finals, I will let the AI opponent move first for the next two games. Let's see if that alters the balance in Game 3 (NOTE: I am letting the computer play first by simply passing on my opening turn.).


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.