Collins Coalition

View Original

Adam Logan annotates a game from the World Blitz Scrabble Championship

(Editor's note: During our last major online tournament, the Virtual World Cup, we invited former world champion Adam Logan to annotate a game from the event on our blog. Adam did a fantastic job, and readers loved his analysis - so naturally, we told him he was welcome to return for a sequel! Below is that sequel. Adam chose to analyze a game between top Australian experts Chris May and Edward Okulicz, from Week 9 of the World Blitz Scrabble Championship. Enjoy! -Evans)

Chris: CDERTWZ

H6 CREW +18 18

There are three sensible choices here: CREW, DREW, or TREZ. DREW scores 2 points fewer than CREW; however, the leave is slightly better generically. TREZ scores 8 more points, but the leave is worse by almost the same amount and the defense is slightly worse. I would have expected that S-CREW and CREW-E make more difference than they do, but presumably a lot of plays using the latter give a strong comeback with CREWE-D. Simulation puts DREW 1.5 to 2 points ahead of the other two, which seems reasonable.

Edward: EIILMNN

10F MIEN +22 22

With no A or T available for MAINLINE or LINIMENT, there are no big plays here. MIEN is clearly the best move using the CREW-E hook. Plays such as 9H WIN keep better leaves, but at too high a cost. WIN increases Ed's average score next turn by about 9 points over MIEN, but even without taking the defensive disadvantages into account, this does not make up for the immediate loss of 15 points. The EN leave from 6C LIMNIC isn't even noticeably better than ILN.

Chris: DDKSTVZ

exchange DDKTV +0 18

There's no question that trading is correct here; what to keep is a bit less clear. Lower-level players who aren't as solid on seven and eight-letter word knowledge might do better to keep SZ instead of ST, from which the next play is more likely to be obvious (this is important in blitz); at Chris's level, that does not apply, and if anything the bias is reversed. So I will go with the simulation that says that keeping ST is about 1 point better than SZ. In turn, SZ and STZ seem about equal.

Edward: AILNOPS

J4 OILPANS +72 94

OILPANS is certainly the right word to play, but it would score 13 more points at 11B, and this easily overrides the dangers of B10. SALPICON at 6C offers no benefit for the 2-point sacrifice relative to this move.

Chris: AAPQSXZ

5I QIS +24 42

Plays like ZAX at K2 seem too dangerous, so the choices are QIS as played, ZAPS at K2, or SPAZA at 11B. The last seems clearly wrong; both the Q and the X need vowels, and the opening on row 12 is more likely to benefit Ed than Chris, so ZAPS must be better than SPAZA. ZAPS vs. QIS is more difficult: After ZAPS, drawing an E would give Chris AXE at L2 for 50 points. However, that would still retain the Q, and it is not especially difficult for Ed to overlap in the L-column after ZAPS. This move seems right.


Edward: ILRSSTT

11E LITS +25 119

If you know and see the best play, it's not hard to recognize it as such: H6 CREWELISTS. Otherwise, LITS is the next best; though it is easier to front-hook LITS than RITS or TITS, the superiority of RST to LST or LRS is a much larger issue.

Chris: AAHPTXZ

12D ZAX +52 94

ZAX is the most obvious play here; the only other reasonable possibility is HAPAX at 12B for 4 points fewer. However, the advantages of ZAX justify spending the Z. In particular, the opening on row 13 after ZAX is useful to Chris, who has at least 13C PAH or 13D APT to play there, whereas even if Ed does not have the right tiles to play on row 13 after HAPAX, it is not easy to use the Z there. Also, if Ed underlaps with ZA or ZE, Chris may have a strong reply involving ZA-P or ZE-P.

Edward: DDRRSTW

8G REWARD +12 131

RST is a better leave than DST, but not by nearly enough to prefer exchanging. REWARD is best.

Chris: ADFHPTU

13C FAH +35 129

The right play is certainly something on row 13, but there are many choices: not only FAH as played, but also FAP keeping the TH together, APT, FAD, and various others. To my surprise, it is APT that simulates best, by about 1.5 points, even though this scores 1 fewer than FAH and the leave is generically 2.5 points worse. The reason for this seems to be indirect defense; APT often induces Quackle to make a play on row 14 that creates a valuable opening for Chris. I suspect that the racks with which Quackle does this are less likely than random after REWARD, while a human would be less likely than Quackle to make such a play (perhaps especially in blitz where a calculated risk is something that is harder to take). So I'll call them equal.

Edward: DEEINST

M8 STEINED +79 210

It's STEINED or ENDITES. My instinct would also have been to play STEINED, in the hope of keeping the top-left quadrant of the board out of play, but ENDITES actually simulates 2.5 points better. Curiously, half a point of this comes from each of Chris's next move and Ed's reply; after that, Chris's next move after that scores the same either way, but Ed gains a point and a half on his second turn after ENDITES. Perhaps Quackle is expecting that Chris will be more successful at opening the board after ENDITES, which will favour Ed on average. Since the simulation results do not consider the long-term effects of STEINED, and it is often easier to see how to close the board than how to open it, I think STEINED is at least as good.

Chris: DGOPRTU

L11 PUDOR +32 161

PUDOR is very good, but GUP in the same place simulates almost as well. In fact, Quackle reports a slightly higher winning percentage after GUP. The problem with PUDOR is that it is easy for Ed to keep the board closed by playing on row 15, whereas after GUP, Ed may have no way to score except with a risky play on row 14 or column N. Also, PUDOR stops bingos starting with a blank S on row 14. Another issue is that in blitz, there is something to be said for not creating a position with a single obvious opening, because Ed would be likely to have an obvious best play that can be made quickly. Against this is the possibility that Ed could reply to GUP with something like VOR at 7F, killing the top left. However, there is no certainty that he has a V, and even if so, Chris might be able to create an S-hook for himself if he can draw a blank. I prefer GUP, but it is very close.

Edward: AKLOOTY

4J OAKY +24 234

The highest-scoring play of N13 YAK feels quite dangerous, although Chris is unlikely to have an immediate bingo using a blank as an S after playing 5 tiles. If he does not, then O14 LO on the next turn would kill that opportunity, though of course Chris might create another dangerous opening by playing on the left side of row 14. If Ed prudently avoids YAK, the best play would seem to be K12 TOYO, scoring well, blocking most bingos on row 15, and not opening anything new. (With random racks, Chris's bingo chance after TOYO is 7% as opposed to 9.5% after YAK, though the bingos after YAK score better.) Holding the K for next turn also makes it unlikely that Ed will be stuck for a play.

Under the assumption that Chris has a blank, the simulation favours TOYO over YAK. It is a common error to reason that Chris should be assumed to have a blank, because Ed is almost certain to win if he gets them both. The latter statement is true, but the cases also need to be considered in which Chris gets the blanks later. OAKY does have the practical advantage of setting up high-scoring plays on row 3 with -OO- above the -KY, but such plays often create useful openings for Chris, and the spot may not remain. So, difficult as it is to make such a play when it counts, I think YAK is best.

Chris: AEGNTUU

15H AUGUR +18 179

AUGUR is certainly the natural play here, and is best for minimizing the spread, but it's time to start thinking seriously about moves like N13 GAU, creating an opening that is very hard for Ed to block without a blank (he could have some crushing reply like O12 VULN, but this is not the expected result). In simulation, the plays are very close in winning percentage, even though AUGUR averages 4 points better.

Edward: AAGLOOT

3K GOA +28 262

GOA is best. This is not the time to score 2 more points with GOAL or GOAT. Last turn, though YAK was dangerous, it created a way to block the opening it made, while it might be difficult for Ed to block column O after GOAL or GOAT without creating new problems for himself on the top row.

Chris: BEGINRT

N11 GIBER +39 218

GIBER is a very good play but TRIBE, one space higher and scoring 5 more points, feels like a better one. The main threat to bingo is on row 7 in any case, and drawing five tiles with three T's unseen carries a risk of duplication. The opening after TRIBE is easier to block, but that also makes it easier to use, and increases the chance of row 7 remaining open.

Edward: ABILOOT

12L UNIT +5 267

It feels too early for such a one-sidedly defensive move. Instead I recommend the highest-scoring play, BLOT at 2L. This draws four tiles instead of one and gives Ed the chance to reply to a bingo in column O hooking PIG-S with a high-scoring play using O1 (AJI would be devastatingly effective, but even AHI or AYE would be strong). UNIT is not as effective at preventing bingos as one would like; Ed's bingo chance is only 1% lower after UNIT than after BLOT, and 1.5% lower than after a neutral move such as WON at 9H, indicating that most of the bingos are on row 7 instead. Ed needs to score points so as to outrun Chris's bingos rather than try to prevent them entirely. Note that BOR at 7F is very effective at preventing bingos, reducing the chance by 6.5% compared to UNIT, but if Chris does bingo after BOR, it would be very difficult for Ed to win. It is also not impossible for Chris to catch up with a high-scoring play such as MARCH at 2J.

Chris: ILNNTUY

14F LUNY +16 234

This waiting move seems like the best choice. YIN would make it too easy for Ed to score points while cutting off the chances of a bingo at O5-12. Also, the leave of LNTU is considerably inferior to the INT kept by LUNY. Similarly, the rack leave from placing the Y on 9I prevents that move from being seriously considered, in spite of the 26 points it scores.

Edward: ABIILOO

exchange ABIILOO +0 267

Exchanging is the sort of panicked move all too easily made in a blitz game. With time pressure and frustration from being unable to block on the right, it's easy to understand this move, but BOOAI at 15B is much better: scoring a few points, keeping a reasonable leave, and shortening the game.

Chris: EIMNTTV

exchange MTV +0 234

Given the remaining tiles, VET at N1 seems too likely to give up a high-scoring play on row 1 that would also allow Ed to balance his rack. Plays such as 2M MET are hardly tempting either, so it's hard to see what to do other than exchange. Although the bingo probability from EIMNT is lower than from EINT, the bingos are higher-scoring, and exchanging just TV might be as good as MTV.

Edward: ?EEFIOT

7A FOETIdER +65 332

Ed draws well, and plays his only bingo, but...

Chris: AEIILNT

A1 ANTILIFE +158 392

Edward: ?BCENRT

2L BREN +27 359

I don't know whether Ed saw CaBERNET at C1 or BaCTERIN at E1, but this play seems to win just as many games as the first of these and a lot more than the second. If Ed is about to draw JVV, he loses either way, and if not, Chris is likely to have have a very difficult decision of whether to block the O column (where Ed can score more) or the left side (which is easier to use). In a game under normal time controls, Chris would have the time to make a proper decision, and Ed might be better off playing CaBERNET, leaving Chris in the difficult position of having two tiles in the bag. In blitz, the decision is more difficult, as it could be difficult for Chris to estimate the probability of a triple-triple or other high-scoring bingo through the N.

Chris: EEEIJOO

O1 ONIE +15 407

And indeed, Chris blocks the O column, where no bingos of any kind are available for Ed, leaving him free rein on the left side. The best play here is certainly D3 EEJIT, blocking as much of the left side of the board as possible.

Edward: ?ACHMOT

D2 OuTMATCH +84 443

Ed draws a bingo and plays it quickly, no doubt expecting to win the game at this point. He would have done better to play ACHrOMAT at 1A, however. One wonders whether he would have found this move had Chris played EEJIT.

Chris: ADEEJOV

C2 JOE +33 440

However, there is one final twist: Ed has only one place for the V (1A AVO), so by blocking it, Chris is assured of victory. Though JOE is by no means the best move in this endgame, it is certainly enough to win.

Edward: OV

[times out]

Ed runs out of time looking futilely for a spot for his V. Since he could not have won without finding one, this was the only thing for him to do.

Had Ed played ACHrOMAT, he would have lost only by drawing two consonants or EJ, a chance of 8/36 to lose and 28/36 to win. It turns out that CHAMElOT at C3 achieves the same result. With OuTMATCH, however, he also loses with EV or OV, reducing his winning chances to 20/36 based on a random bag. (Inferences from ONIE are unclear, since it is hard to construct a rack from which it is the best play, but it seems to suggest holding vowels, which means that the actual probability of winning is lower. In any case, no inference could justify OuTMATCH, since ACHrOMAT always wins when OuTMATCH does.) A one-tile fishing play would have been futile, as drawing a V or the J would make it impossible to bingo.

Final result: Chris wins by timeout. (With perfect endgame play, he would have won by 2 points, 459-457.)