A Whopper of a Game
Kit Woosley, 2001
GammOnLine, December 2001
On a good day, an expert can hope to play through an entire game making at most two or three errors, hopefully none of them serious blunders. On a real good day, he might struggle through a full game without any noticeable mistakes. When we achieve this, we finally think that maybe we are learning something about this game of backgammon.

Reality soon sets in. There comes that game when we make one whopper after another, showing us once again that we really don't understand backgammon at all.

The following is such a game. I don't know what was so difficult about it. We weren't involved in a weird backgame or anything like that. The positions were all normal enough. Yet, move after move the proper play eluded both myself and my opponent.

I rolled out all the controversial plays on Snowie. The rollout parameters were as follows: 360 trials, 3-ply with variance reduction, no truncation until data base bearoff, 50% speed, medium search space, cubeless. These results aren't necessarily conclusive, but they are quite likely to be accurate. I will report the cubeless equity of the rollouts, since other numbers involve Snowie's strange biases when taking the cube into account and involve distorted pictures.

It should be noted that Snowie also had plenty of trouble with this game. There were several positions where there was a big discrepancy between Snowie's 3-ply opinion and the rollout results, resulting in Snowie's 3-ply evaluation choosing a whopper of an error. Maybe this game is just too tough for carbon and silicon beings alike.

This is a 7-point match. Woolsey leads 4–1.

Kit Woolsey (Black) Max Modesti (White)
1. 4-2: 8/4, 6/4
2. 4-2: 8/4, 6/4

White to play 3-2.

2. . . . 3-2: 13/11, 13/10

24/22, 13/10 and even 13/8 are all perfectly fine here, as is the actual play.

3. 6-1: 13/7, 8/7

White to play 6-2.

3. . . . 6-2: 24/16

With nothing constructive to do up front, running a back checker over Black's blockade seems to be the right idea despite leaving all the blots. Getting stuck with two checkers on the 24 point when your opponent has made his bar point often leads to losing a priming battle.

Black to play 3-1.

My first big decision of the game. Two constructive plays which follow two entirely different themes. The advantages and disadvantages of both plays are quite clear — it is weighing the priorities which is the difficult part.

4. 3-1: 13/9*

Making the five point looked nice, but that allows White to either run the back man home and play with one checker back or to make a new point of his own. It seemed to me that White just had too many good numbers. On the other hand I don't have too much ammunition up front, and making the five point puts these few checkers to optimal use. The hit stalls White for a roll, but I will have to follow it up quickly.

The rollout had the two plays very close. In Snowie's opinion 13/9* was a fairly serious error, although this was somewhat biased by the cube potential and the match score, as it thought the cubeless equities were fairly close. Apparently Snowie felt that the quieter play of making the five point would be less likely to lead to White getting one of those sharp cubes which I wouldn't like to see at this match score.

White to play 4-3.

4. . . . 4-3: bar/22, 13/9

This is a loose play, but since Black has only a two-point board it is quite reasonable. Snowie preferred bar/22, 10/6, but felt the two plays were very close.

Black to play 2-1.

I can buckle up and split, or I can keep White busy with the loose hit. White has a lot of good rolls if I let him be, but hitting loose is really loose. Which should it be?

5. 2-1: 6/3*

Snowie's opinion was that the loose hit was a blunder, and the rollouts backed this opinion up. The error was magnified due to the cube potential, I think, as it was reported as a whopper even though the cubeless equity difference isn't so monstrous.

So why is the loose hit so wrong? White has all those builders in position, and if I don't keep him busy just about anything he rolls will make a new point. After 6/3*, it is a free-for-all. The loose hit has the advantage of putting White on the bar, but White does have several shots from the bar and if I am hit I could be in trouble.

I guess that 24/22, 9/8 comes out better for several reasons. It makes a clear asset when I am relatively short of ammunition. The split of the back checkers is also a plus. Splitting to the 22 point isn't dangerous, and it keeps White from making full use of some of his good numbers. Maybe it simply comes down to doing two good things at once.

5. . . . 2-1: bar/22*

Black to play 4-1.

6. 4-1: bar/24, 13/9

Hitting loose and breaking the bar point here would be bananas. I have to take what I can.

Should White double?

White isn't too far from a cube, particularly at the match score. However, Black's board is as strong as White's, and White can't do everything at once. I agree with waiting, as does Snowie.

White to play 6-6.

White can dive in and make the two point, leaving no direct shots, or he can hold the eight point and make the positional slot of the five point. Slotting the five point will work great when it works, but is it worth the risk?

6. . . . 6-6: 13/7(2), 11/5, 10/4

According to Snowie's 3-ply evaluation, slotting the five point was a fairly serious error. The rollout, however, turned this around completely, saying that making the two point was a whopper of a blunder. What is going on here?

Making the two point is simply too advanced, as well as leaving too many blots. Slottng the five point does leave that direct shot, but there is some serious duplication since most of Black's fours can be used very effectively on the other side of the board also.

The key is that White has to play this as a priming game for now, and the eight point is better than the two point in conjunction with the made bar point and Black's three back checkers. Admittedly this isn't immediately obvious, and Max is to be complimented on finding the right play here. Snowie wasn't able to do so. The program can have problems in priming and timing battles, and this is a good example.

Black to play 1-1.

Again I face a similar problem. If I let White alone he will probably be able to cover the blot on his five point, and I have three back checkers stuck on his ace point. Still, making the five point is so pretty.

7. 1-1: 24/23, 9/8, 4/3*(2)

I felt that the combination of splitting the back checkers, knocking White off my three point, and keeping him busy so he couldn't cover the blot so easily was sufficient compensation for not making the five point. I was wrong, and by quite a margin. What I failed to see was that even if White does make his five point, which isn't guaranteed, I have those two outfield checkers to play with, so the timing battle could still go my way. I will have five parts of a prime, and White will have problems getting over my blockade before he crunches. After my actual play, my position is a ragged mess, and White still has several rolls which cover the blot on the five point.

This illustrates once again that when you are short of ammunition it is important to make full use of every checker. Making the five point is simply too much of an improvement to pass up. After my actual play, the position will be too difficult to put back together again.

Should White double?

Once again, White is on the verge of a match score cube. This time I believe that he should. If he rolls well and makes his five point, he could lose his market by quite a bit. This is exactly the sort of position he should be looking to cube. My recube vig will be virtually zero, and he can use a doubled gammon far more than I can. In the evaluation Snowie thought that doubling was a photo, but the rollout had it a clear double. Obviously I have a take, match score or no. I'm going to win this game quite often.

7. . . . 3-2: bar/23, 8/5

Black to play 6-4.

8. 6-4: 23/17*, 13/9

Locking up the nine point looks best, even though there are some return shots at the blot on the 17 point. I need that blockade. Snowie agrees.

8. . . . 4-2: bar/21, 23/21

Black to play 1-1.

Three entirely different possibilities. Advancing to the 22 point firms up my defense, but it leaves a shot instead of hitting one. 24/23, 24/22, 17/16* is an attempt to scramble out while White is on the bar. 24/23(2), 17/16*, 16/15 (a slightly better last ace than 13/12 which duplicates my sixes possibly) makes a stand on the 23 point. Which is the right idea?

9. 1-1: 24/22, 24/23, 17/16*

The gains from hitting appeared too great to pass up. I made what seemed to be the natural try of escaping. It was a whopper. I didn't appreciate the essence of the position. Timing was in my favor now that I had two outfield checkers. I didn't need to give myself a five to escape. Sitting on the 23 point ready to escape with sixes and watching White try to find a way to play his rolls was quite sufficient. The big danger for me wasn't the prime — it was the attack.

9. . . . 3-3: --

Black to play 4-3.

10. 4-3: 16/9

Clearly better than making the five point and leaving several shots. My strength is still the solid four-prime.

White to play 4-1.

10. . . . 4-1: bar/21, 4/3*

Obviously best. Now the pressure is on me to produce. This is what I should have been afraid of with my 1-1 play a couple of rolls ago.

11. 4-4: --

Should White double?

White has to be at least knocking on the door of a double here, particularly at the match score. True he has some escaping to do, but Black is in trouble and White has some very big threats. I believe he should double.

What is interesting is Snowie's opinion of the positions. On Snowie's 3-ply evaluation, Snowie thought this to be double and close PASS. That looks to be way off. White still has four checkers to get over a serious blockade, and he could have some problems. It is hard for me to imagine that this is a pass, match score or no match score. Snowie's cubeless equity for White is +.512, but that does seem quite high.

The rollout told an entirely diffierent story. Cubeless equity of +.375, and estimated as close double — huge take. For some reason, Snowie seems to have difficulty evaluating the power of Black's blockade here.

White to play 5-4.

11. . . . 5-4: 7/3, 7/2*

Either this play or 7/2*, 6/2 looks okay. This makes the better point, but making the two point leaves more diversification to make the other point.

12. 5-4: --

Should White double?

12. . . . Double

Now the double is clearly correct, with a four-point board and two of Black's checkers in the air. It is the take which has to be examined closely.

13. Take

It looked like a take to me. Sure there was gammon risk, and at the match score I wouldn't get too much value from the recube. But White has those four checkers stuck back there, and if he doesn't spring one of them quickly he will crunch. In addition, he hasn't covered the blot on his two point yet. There are so many good things that can happen for me this game that it appeared to be a take.

Snowie thought I had made a huge blunder on its 3-ply evaluation. It came up with a cubeless equity estimate of −.761! That sure would be quite a blunder. However, as we have seen, Snowie's evaluation of this sort of position may not be too accurate.

The rollout gave another story. Cubeless equity of −.425, a very clear take. This is a pretty big mis-estimation by Snowie, particularly since there will tend to be a bias in the rollouts toward its original estimations because play decisions will be made along lines which are consistent with the rollout estimates. I don't know why Snowie fell apart in its evaluation, since it certainly understands the value of primes and timing and one wouldn't think this was such an uncommon type of position. I guess Snowie can have its blind spots the same way we humans do.

White to play 3-2.

Taking two checkers off the six point is bad on some fives, but moving the back checker up to the 21 pont costs White his remaining flexibility on Black's side of the board. Which fate is worse?

13. . . . 3-2: 6/4, 6/3

The fives problem isn't as bad as it looks. Only on 5-2 and 5-3 does it matter that White has a spare on the six point. White can escape with 5-1 and 5-4 after playing 6/4, 6/3. The inability to play a four to the 20 point after 24/21, 6/4, however, is quite serious. The rollout had 6/4, 6/3 coming out a bit better. What was interesting, however, is that Snowie thought that 6/4, 6/3 was a whopper of a blunder. One would think that on the 3-ply evaluation where the program could see the results of the next move it would understand the reason for this play, but that doesn't seem to be the case.

14. 4-2: bar/23*

White to play 5-1.

14. . . . 5-1: bar/20, 3/2*

White definitely must continue to attack, particularly since he now has a checker on the 20 point which can spring with a five. If White allows Black to anchor, White's game can fall apart fast.

15. 6-4: --

White to play 3-1.

Two obvious choices. Making the ace point is impure, and 24/20 does bring another checker into position to escape and locks up a potentially important anchor. But that blot left on the two point may be too much to concede.

15. . . . 3-1: 24/20

This one is not close. White's forces up front are too thin to worry about super-purity. White's plan is to keep Black on the bar while he escapes. Making the ace point is clear, as shown by the rollouts. Max just plain mis-evaluated the priorities here.

16. 6-3: --

White to play 3-2.

White can gamble on remaking the fifth inner board point and leave the six point slotted, or he can play it safe.

16. . . . 3-2: 6/3, 4/2

White is trying to do too much at once. He has to accept that his plan to hold the best four points has failed, and he must try to make the most of what is left. Leaving the blot on the six point is risking too much to gain too little. Getting hit is a disaster, and it will be too hard for White to cover the point in time. The simple 5/3, 5/2, hoping to scramble out and rekindle the attack, is clearly best.

17. 4-2: -- 5-5: 20/10(2)
18. 6-4: bar/19* 2-1: bar/23, 21/20
19. 5-1: bar/19 5-3: 20/12*
20. 4-4: -- 6-2: 23/15
21. 5-2: --

White to play 5-2.

21. . . . 5-2: 12/5

With Black camped on White's six point, White doesn't gain much by slotting the bar point with 15/13, 12/7, and 15/10, 12/10 is too inflexible. The spare on the five point is as good as White can hope for for future flexibility.

22. 5-3: -- 5-2: 15/8

Black to play 6-4.

23. 6-4: bar/19, 9/5

It looked to me as though I would have to make my five point sometime if I were ever going to make progress, so I thought this slot was worth the risk. Snowie felt that this play and the safe bar/19, 6/2 were very close.

23. . . . 2-1: 21/20*, 10/8
24. 3-2: --

White to play 4-3.

24. . . . 4-3: 10/3

8/5, 8/4 clears the eight point and brings a new builder in for the ace point, but it leaves a few indirect shots. I prefer 10/3; Snowie thinks the plays are very close.

25. 4-3: --

White to play 3-2.

Can this be the time to slot the ace point in order to smooth things out and hold the eight point, or is the simple but ugly 8/3 sufficient?

25. . . . 3-2: 8/3

It wouldn't have occurred to me to slot the ace point, nor did it occur to Max. According to Snowie's 3-ply evaluation, however, 8/3 was a serious error. The rollout corrected this concept — it is slotting the ace point which is the serious error. I have no idea why Snowie came up with this slotting concept in this position, since usually it tends to play safe more than humans. I guess it couldn't stomach the thought of five checkers on the three point. Considering that White's game plan is simply to bolt out and run home, a bunch of checkers on the three point don't make too much difference.

26. 4-4: --

White to play 3-1.

Is the higher priority to move up and try to escape, or to smooth out the builders for the ace point?

26. . . . 3-1: 21/20, 8/5

It would seem automatic to move the back checker up, since the game plan is to get those checkers out of there. On the 3-ply evaluation, Snowie had a small preference for 8/4. The rollout conclusively showed that this is the wrong idea — 21/20, 8/5 is clear.

27. 4-2: -- 6-6: 20/8(2)
28. 6-4: bar/15 6-4: 8/4, 8/2

Black to play 4-4.

29. 4-4: 19/11(2)

This is not the time to waste so much as a single pip. My distribution is good enough that I can aim to bring every checker exactly to the six point, perhaps dropping a checker on the five and/or four point at the last minute. Until then, the goal is to aim for the six point and no further beyond.

29. . . . 6-5: 5/off(2)
30. 5-1: 11/6, 7/6 6-3: 5/off, 3/off
31. 3-2: 9/6, 8/6 3-1: 3/off, 2/1
32. 3-3: 15/6, 9/6 6-6: 5/off, 4/off(3)

Black to play 4-4.

33. 4-4: 19/7, 8/4

Now it is necessary to look at crossovers. I can't afford to play 19/7, 11/7, since then I would need five crossovers to take a checker off (four to bring the checkers in and one to take one off), and I can't afford that. However, it is not necessary to bring two checkers in now with 19/11, 8/4, 7/3. That would involve wastage if I rolled two small numbers. The plan is to leave myself with three men outside, so I can bring two in next turn, and the last one in and one off the following turn. Hopefully I will roll a five or six next turn, so I won't have to waste. If I don't I will have to bring in both checkers from bar point and hope to roll well the following turn.

33. . . . 5-4: 3/off(2)
34. 6-1: 11/5, 7/6 6-5: 3/off, 2/off
35. 6-3: 7/4, 6/off 5-4: 2/off, 1/off

Single game
2 points

More articles by Kit Woolsey

More annotated games

Index of GammOnLine articles

Backgammon Galore