Preface
In terms of word study, I was better prepared for this season than pretty much any other season I’ve played. I went through all the 4s and learned around a thousand new bingoes between last season and this season.
Game 1
I’m up just 2 points in this position with 3 tiles in the bag (ACDEEILNV?).
There are two different things to worry about here. There are many bingoes unseen through the open letters through AnitaCH’s last play of TIARA (technically CELErIA(C) is also possible as the only bingo to the C). I also can’t take too low of a score here to block potential bingoes or else I risk getting outrun if AnitaCH has a strong out-in-2 sequence. The play I eventually settled on was M2 OH for 22 points, leaving AENN? and 1 in the bag. My reasoning was that it blocks some bingoes through the I and A of TIARA and words ending in -TE through the T. I also felt that with the blank still on my rack, a bingo to the T would still give me a decent chance of bingoing out by overlapping their play. This was a big miscalculation. I didn’t realize how many bingoes there were unseen to the T using the V or C in early position. With the V/C in early position, there is no room for me to try to fit a counterbingo underneath their bingo and I just automatically lose. Also, even if AnitaCH doesn’t have these tiles, if I draw them, then I’m more likely to have trouble fitting my counterbingo. I’m really just not as likely as I thought to bingo in response, especially with more duplicates of the tiles I’m keeping unseen such as duplicated As and Es or even triplicated Ns.
Macondo’s favorite play in this position is M1 ANE for 14 points, emptying the bag and sending us to a 7v7 endgame. The most obvious benefit of this play over my play of OH is that it blocks many bingoes through the T. Overlapping TIARA all the way down also ensures stray bingoes with a blank S hooking SOH don’t fit (such as DEVIAN(T)s etc). The only bingoes that play after ANE are the aforementioned CELErIA(C) and the 9s CANDLEp(IN) and ErVALEN(TA)/rEVALEN(TA). AnitaCH is just 2.8% to bingo compared to 32.1% after OH. OH being a problem can actually be partially demonstrated just by the natural bingo DIVALEN(T) being unseen. With the blank also unseen, this word alone means that ~6.67% of the time I lose on the spot after OH.
There is also the fact that Macondo doesn’t solve solve non endgames/non 1-in-the-bag situations exhaustively. It assumes that both players will make their top equity plays (just points - leftover tiles - penalty for not going out) in the endgame, so I’m probably less likely than it thinks to win the endgame after ANE as long as AnitaCH makes a reasonable out-in-2 sequence. Despite this, I think that ANE is solidly better, and while I don’t think it quite wins the 84.3% that Macondo thinks it does, it’s still likely a much better play than OH which Macondo rates at 53.7%.
In the actual game, I drew C? to AENN? but it wasn’t good enough because AnitaCH actually did have DIVALEN(T), leaving me no playable bingoes, eventually losing by 49.
Game 2 - Don’t always trust the engine
This is actually Irfan’s turn in our game. He’s down 91 points with 2 in the bag (EEIIILNNZ).
Two plays immediately jump out from this position, being the only two available bingoes, C1 (R)IOTRIES 68 and B6 RIOTISE 63. Both of these plays are not likely going to allow Irfan to win. Any time I have the Z, I’ll easily be scoring 20-60 points with it, easily outrunning any outplay he has. If I don’t have the Z, even though I’ll have all 1 point tiles with a lot of duplicates, I can block his highest scoring Z plays if necessary or underlap RHAPSODE to outrun Z plays. Just like in the earlier example, Macondo, because it’s not solving the endgames and is instead making my top equity play, thinks I will just leave big Z plays open if I have a better score elsewhere on the board. This is likely the reason why it rates RIOTISE as even 6.1% to win. If Irfan draws EZ out of the bag, I’ll have EIIILNN. Macondo assumes I’ll play something like D1 (H)INNIE for 18 to score and offload tiles, which would cause me to easily lose to ZE for 68 on the triple. In reality, I can just play something like A8 NIL 19 which easily beats Irfan’s best outplay of 15C ZE for 30 afterwards. This is likely the reason why (R)IOTRIES doesn’t even show up on the top 5 best plays, it doesn’t set up the threat of the Z bomb.
Looking at the Macondo analyzer, there is a surprising play that Macondo thinks is best.
As always, it goes back to simulation assumptions made by the 5 ply sim. Let’s take the rack of EEIIILN. Macondo, making my top equity play, will play 9H (NE)INEI for 7 leaving the somewhat balanced bingo tiles EIL. A reasonable player would never make that play in this position. With the unseen tiles from their perspective (EINORSTZ), if they draw the Z, then they lose by 5 after (H)ORNIEST 76 and being caught with the Z. One winning pathway out of that rack is to just pass. The only unseen bingoes are TRIZONE and many playable bingoes if the Z is in the bag. After (H)ORNIEST, something like 8A LIN(T)IE 18 easily outruns his outplay of C13 (BE)Z for 14 (similarly, TRIZONE also doesn’t take the lead and any reasonable scoring play wins).
Macondo also may think the opponent may open a triple-triple lane which would never happen against real play.
Game 3
I’m down 59 points early into this game with zzz.
The most obvious way to make up that deficit is to play a bingo through any of the open letters. Instead of bingoing here, I actually decided to sacrifice 40 points and play 5J ETA(T) for only 24 points here. The first reason why I did this was because my bingoes barely score anything. My top scoring bingo is (S)EAbLITE for just 66 points, which will only put me up 7 points. An average turn is around 30-35 points, so I’ll probably still be trailing after his next turn. The second reason why I passed up all of my bingoes was because of the massive counterplay. All my bingoes are from the C/L/S of GRACILES, so I have to put a tile into the triple-triple lane. I also have to put one of two of the most dangerous letters into the triple lane with any of my bingoes, the E and the S. His average score is just so high on such an open board after my bingoes. After ETA(T), on a relatively tight board, zzz averages just 33.6 points a turn. According to Macondo, my after (S)EAbLITE, my 2nd best play, he averages 51.7 points a turn, an 18.1 point increase. I average 72.4 points a turn on my next turn after ETA(T) holding the powerful EIL? while only averaging 46.9 points a turn from a completely random rack after playing one of my bingoes. Adding these together, just in the next 2 turns, ETA(T) does 43.6 better than (S)EAbLITE, making up for the 42 point sacrifice.
Game 4
In this position against MUSICBOX, the game is almost over. There are 5 tiles in the bag and I’m up 44 points. The unseen pool is AADEFMNRRSTU.
My bingo of QUINTAR does not fit on the board and MUSICBOX just set up a big S hook on the triple lane. There are no reasonably scoring plays available that effectively block the triple. Something I like to do when the game is almost over is see if I have the last of any tile, and if so, see if I can set up that tile. Here, I have the last (well, only) Q and the last I. The way the board is constructed, there isn’t really a way I can try to set up my Q on a letter multiplier square. However, paired with the Q, there is an extremely powerful setup available in this position. In this position I played N8 QUA(D) for 16 points.
With the last I in hand, leaving INRT, at a minimum I’m threatening TRIN on the right side of QUAD for a massive 47 points, likely to outrun any bingo MUSICBOX can have. It’s very difficult to block the setup as well, as many 2 tile plays can be extended to still hit the triple for a similar score. QUA(D) also blocks almost all bingoes MUSICBOX could have on the right side hooking TOADS. The only bingo available that still fits there in the unseen pool is RASTRUM for 90, which is extremely unlikely.
In the actual game, MUSICBOX played 4C R(E)D for 9 points to balance their rack, emptying the bag. I drew EMU to my leave of INRT. Although I have the massive play of O4 MUREIN for 68 points available, I can simply bingo out with C8 MINUTER for 82, catching MUSICBOX with a full rack of tiles to win by 153.
Afterword
I made a lot of word knowledge errors this season, but I still had a pretty strong season. I had my best ever result in division 2, finishing 10-4 +811 in 1st place. I did tie my record for lowest MiAOW at 2.6, but the biggest contributing factor is the absurd number of blanks I drew this season. I averaged 1.43 blanks played per game this season, or 20/28 possible blanks played by me.
Stats
Lost Endgame Points: 25 (10/14 optimal endgames)
Missed Bingo Turns: 7
Suboptimal Bingo Turns: 7