Learning and Not Learning… Humans and Computers

by admin on April 12, 2020

Coronavirus check-in! It’s Easter Sunday of 2020, and here in the U.S. the coronavirus epidemic seems to be near its peak. That’s both good and bad. It’s good because “peak” means that the number of new cases and new deaths should start going down every day. In New York and California, the number of intensive care beds used already seems to be going down. On the other hand, peak also means that there is a whole lot of coronavirus out there. 800+ deaths in the US just yesterday. (And I suspect that reporting slows down on the weekend, so it could be more.)

One of my chess friends, whom I will just call R.R., is going to win a rather ghoulish bet. About three or four weeks ago, he was having a conversation with a friend who was a coronavirus skeptic. This friend thought that the media was blowing the epidemic all out of proportion. R.R. was getting a little bit mad, and he said he did something he often does in similar situation: He asked, “Do you want to bet on it?”

In the end they made three bets of $10 each. The first was whether there would be more than 100 coronavirus deaths on Easter Sunday in the United States. R.R. said yes, his friend said no. R.R. will win this one, because the number will likely be between 800 and 1000. The second bet was whether Disneyland would still be closed on May 1. R.R. said yes, his friend said no. R.R. will almost certainly win this one, too, because California’s shelter-at-home order runs through May 4. Finally, the third bet was whether Disneyland would be closed on December 31. R.R. said yes, his friend said no. I think that R.R. got a little bit carried away here. It’s basically unknowable at this point; the epidemic could come back big time in the fall (as Spanish flu did in 1918) or not.

Anyway, I’m delighted that R.R. will take $20 off his ignorant friend, but I’m still not sure whether the bet accomplished its purpose. The coronavirus skeptics are still saying, and going to keep saying, that the danger of the epidemic has been overestimated. Yes, one skeptic is going to have to cough up ten dollars, but the mass of them are not going to change their tune, and even that one skeptic will forget soon enough that he personally underestimated the threat by a factor of EIGHT (or more). Good grief!

Do any of you have interesting coronavirus stories, either happy or sad?

Okay, now to get to some chess content… I want to show you another example of not-learning, this time by the computer.

My chess activity since the epidemic began has consisted of playing way too many games against Fritz 17 on my computer. It’s somewhat addictive because Fritz plays terrible openings, and I get a huge advantage in most games, but my middlegame tactical play is just awful compared to the computer and I end up losing these beautiful positions. Then I get angry, and the angrier I get the worse I play, and it’s a vicious cycle…

As a nice case in point, a couple weeks ago I played a cool queen sacrifice on move 14 (which could not be accepted), got a position that Fritz evaluates at +7 pawns with best play, and then proceeded to lose. However, there was a happy ending. Fritz’s opening repertoire is so limited (and it has no ability to learn from its mistakes) that I got the chance yesterday to play the exact same queen sacrifice a second time… and this time I managed to bring home the victory.

Here’s how it went down (twice).

White: Dana. Black: Fritz 17. Time control: 40/10. Fritz’s rating set at 2025.

1. e4 d5 2. ed Qxd5 3. Nc3 Qe6+

As I said, Fritz’s repertoire is quite limited. It plays the same variation of the Center Counter (Scandinavian) Defense every game as Black (if I cooperate). So I am now unbelievably familiar with this variation, which I previously had never seen in my life.

4. Be2 Nf6 5. Nf3 h6?!

This seems unforgivably slow to me, in a position where Fritz has already wasted a tempo with his queen.

6. O-O Nc6 7. d4 …

I like to withhold this move until after Black plays … Nc6, so that I can gain another tempo with the threat of d4-d5.

7. … Qg4?!

Position after 7. … Qg4. White to play.

FEN: r1b1kb1r/ppp1ppp1/2n2n1p/8/3P2q1/2N2N2/PPP1BPPP/R1BQ1RK1 w kq – 0 8

Fritz 17 is absolutely berserk. No human would play this move, walking right into a potential discovered attack. This move can’t be good. But I have had lots of frustrating games where I tried to chase its queen around, and did not get much to show for it. Finally, a couple weeks ago, I came up with the idea of striking on the opposite side of the board.

8. Nb5! Qd7

Okay, so the discovered attack won’t happen, but we have made the queen move again, and it’s blocking the development of the queen bishop so it probably “owes” us yet another tempo.

9. c4 …

A good and natural move. However, Fritz points out that the direct 9. d5! is even stronger. Of course Black can’t take because 9. … Nxd5 10. Qxd5! Qxd5 11. Nxc7+ wins a piece. After 9. … Nb4 10. Ne5 Qd8 11. Bc4 the same idea is still in play; even though the pawn on d5 seems inadequately defended, it’s not. If Black tries to drive off the knight with 11. … a6, then 12. d6!! is a killer, threatening 13. Bxf7 mate.

9. … a6 10. Nc3 e6 11. d5! Ne7?!

Everything is bad for Black. However, the next two moves kind of look as if Black is constructing a coffin for his king. Fritz (after the game) thinks 11. … Na7 is best. I would think that 11. … ed 12. cd followed by … Ne7 would avoid the combination played in the game, but on the other hand it opens up the e-file.

12. Ne5 Qd8

The computer’s passivity is somewhat surprising, because it usually errs in the direction of insanely aggressive moves, like 7. … Qg4. But evidently Fritz does not like 12. … Qd6 because of 13. Bf4, threatening Ng6.

13. Qa4+ Bd7

Position after 13. … Bd7. White to move.

FEN: r2qkb1r/1ppbnpp1/p3pn1p/3PN3/Q1P5/2N5/PP2BPPP/R1B2RK1 w kq – 0 14

Now I get to have my moment of fun.

14. de! …

I didn’t say that my queen sacrifice was deep. It’s a two-mover; if 14. … Bxa4?? 15. ef mate! Very appropriate punishment for all of Black’s tempo losses and retreating moves: the Black king is checkmated by a pawn, while surrounded by his own do-nothing pieces.

Of course, computers don’t fall into things like that. In fact, Fritz probably saw this whole thing coming and decided it was better than the other options.

14. … fe 15. Qc2 …

The queen’s sortie accomplished two very useful things: it gave Black an extremely weak isolated pawn on e6, and it opened up the light squares on the kingside for White’s pieces to invade.

15. … c5 16. Rd1 Qc7 17. Nxd7 Nxd7 18. Bh5+ Kd8

Position after 18. … Kd8. White to move.

FEN: r2k1b1r/1pqnn1p1/p3p2p/2p4B/2P5/2N5/PPQ2PPP/R1BR2K1 w – – 0 19

White’s position is so dominant that most chess books would probably say here, “… and White wins.” But believe it or not, I lost the first game after getting this position. I started going wrong right here, when I played 19. Bf7? Of course I wanted to attack Black’s crippled e-pawn, but this is the worst way to do it. The computer played 19. … Nf5 20. Bxe6 Nd4! The Black knight has achieved a very impressive outpost. I played 21. Qe4 Nf6 22. Rxd4+ ed 23. Qxd4+ Qd6.

Although Fritz still evaluates the position at +2.5 pawns for White, that is way down from the initial position after move 18, when the position is +8 pawns for White. This is a big enough change that we can say that White has done something seriously wrong. From there, although I won’t bore you with all the moves, my position kept going downhill. First I traded queens earlier than I should have, a serious mistake because king safety is Black’s biggest problem. Then, in the endgame with two pawns for the exchange, I made a bunch of mistakes in time trouble and lost.

So I’m only human. But being human, I can also learn. And one thing I realized here is that the Black knight has to be kept out of d4. So this time I played

19. Bg4! …

Better. Fritz says my advantage is now 5 pawns. Top marks (8 pawns) are reserved for the move 19. Qe2! The point of this move is that after 19. … Nf5 20. Qxe6 there is no time for 20. … Nd4 because of 21. Qe8 mate.

19. … Nf5?

After this it’s really easy. A good question is what happens if Black tries the other route to d4, 19. … Nc6. Then Fritz’s analysis goes 20. Be3 Qe5 21. Ne2 (note how determined White is to keep the knight out of d4) 21. … h5 22. Bh3 g5. On the surface it seems as if Black has whipped up a great initiative out of nowhere. But just in the nick of time comes a move that reminds us that White is boss: 23. Qd3! Black has no really good defense to the mate threat.

20. Bxf5 ef 21. Nd5! …

After this the finish is just brutal. Although material is even, Fritz evaluates the position at +8 pawns for White!

21. … Qd6 22. Bf4 Qc6 23. Qxf5 g6 24. Qf7 Bd6 25. Bxd6 Qxd6 26. Nf6 …

Here a human would resign. Fritz played

26. … Rf8

and I settled for winning a piece and getting a truly screwup-proof endgame with 27. Qxd7+ Qxd7 28. Nxd7. And White won.

The first time I played this variation with the queen sac, I was thrilled at finding the sac but disappointed with losing the game. The second time, I got to enjoy the sac again and this time win the game. But the sad thing is that, zombie-like, the position will probably come back again and again (because Fritz does not learn) and I will not learn anything more from it.

Print Friendly, PDF & Email

{ 12 comments… read them below or add one }

Greg April 15, 2020 at 8:54 pm

I watch the videos of the ginger gm simon williams. He also likes to gambit and attack. In his seris ” longer games” he always talks about “restrain and push”. Restrain your opponents ideas then push your own. Priority is restrain, so i was looking at Nb2, b4, Be3 as move 19 candidates, because stick a knight on d4 is the most obvious plan for black. Do you think that playing against the computer has made your restrain instincts weak because computer moves dont make sense to humans.

Reply

admin April 16, 2020 at 7:56 am

This is an interesting stylistic question. My natural style is to always fight for the initiative, and I think this causes me to undervalue restraining move sometimes. Playing the computer actually helps me appreciate restraint, because the computer is simply better than me at tactics. So taking away its threats in advance becomes a matter of survival.

The position in diagram 3 is a really good example of the two kinds of thinking. Player 1 (me) says, “What are the weaknesses in Black’s camp that I can attack?” He sees the isolated pawn on e6, and behind it the pinned knight on d7, and he thinks, “Wow! Putting my bishop on that diagonal looks like a great idea!”

Player 2 (you) says, “What does my opponent want to do in this position and how can I take it away from him?” He sees that bringing a knight to d4 would solidify Black’s position and defend the aforementioned targets, so he immediately sets about preventing that with the moves you mentioned.

In reality, there is no one right way to think; both of these approaches are valid. Sometimes the first approach will work better and sometimes the second, and a versatile player should be able to think in both ways. But I think that I need to work on my ability to think in the second way, especially in speed or rapid chess. (In slow, tournament chess I think I do a better job of restraint … at least until I get into time trouble.)

Reply

Greg April 15, 2020 at 8:56 pm

Why dont you use a chess board plugin for wordpress. So readers can play the moves.

Reply

admin April 16, 2020 at 8:02 am

I’ve used one a little bit in the past, but decided that I prefer the old-fashioned way. It forces the reader to actually get out a chessboard or a program and make the moves — to slow down and think instead of just clicking. Also, if I print out a PDF, everything is there.

Reply

Mary Kuhner April 16, 2020 at 7:29 am

Is there any chance that a small change in Fritz’ rating would jar it out of its fixed opening choice here?

My coronavirus story is that I started a new job on April 1, so I am working in a lab I’ve never set foot in, with co-workers I’ve never met in person, using a computer I’ve never seen. It’s a little difficult to believe in all this!

I’m also teaching after-school “chess club” to K-5 students via Zoom. If anyone has tips or tricks to share, I would love to hear them. We hoped to use chesskid.com to have the kids play each other, but that has been a logistic nightmare, so instead we do puzzles together and I show short games. Their attention span has actually been quite remarkable, better than it was in person.

Reply

admin April 16, 2020 at 8:06 am

Yes, Mary, I suspect that changing the rating might increase the variety of openings a little bit. I haven’t done that yet partly because I’m stubborn. I feel as if I ought to win at least 75 percent of the games in this 1. e4 d5 2. ed Qxd5 3. Nc3 Qe6+ 4. Be2 variation, but I haven’t gotten even close to that point yet.

Do you have a paid subscription to Zoom so that you can have a call with several people at once?

Reply

Mary Kuhner April 20, 2020 at 5:57 pm

Yes, I have a paid subscription, since I’m actually getting paid to teach these classes. (I think even a free subscription will give you multiple people, but there is a tight limit on meeting length, and other issues.)

I’m struggling with the fact that I am fairly face-blind, and had only been teaching these kids for a short time when the shutdown began. So I’m confronted with a bunch of tiny pictures of faces that I can’t tell apart, with screen names that are not their usual names (which are not their names of record–they’re a nickname-prone bunch), and…well, it’s confusing, and I probably let the strongest one dominate the conversation because I can’t use my usual trick of calling on someone else by name. Also zoom will only let me see a few of them at a time.

All of us need to learn not to point at our screens, as it’s quite useless. But their use of notation is improving by leaps and bounds as we do puzzles together and they shout out answers, so that’s cool!

Reply

R.R. April 26, 2020 at 8:18 am

An update on the coronavirus bet. The after-Easter reply from my friend.
===============
So when we made the Easter bet, we were all looking at essentially the same info, except you still put some faith in the projections of the covid alarmists that have been shown to be way off. Why were these projections, so high? Why were there even some comparisons to the Spanish flu, when even the most casual knowledge of the Spanish flu shows that such a pandemic today is virtually impossible? When we made the Easter bet, total US covid deaths were projected to be well over 100,000, 250,000 or even higher, so projecting the curve to Easter, meant that daily deaths would have to in the many thousands. I suppose if I just wanted to win the bet I might have suggested that daily deaths would have been say 2000 and whoever was closest would win. You might have taken the bet at that time because projected US deaths were so much higher on that date. But this was a casual bet, which I of course will still pay, but the point was to keep us in contact and, not make money, and if by chance I was right, I could claim to be clairvoyent. Flattening the curve was meant to give the US healthcare system a chance to prepare for the wildly exaggerated expected number of cases. Since this has not materialized and it is now clear that the US healthcare system will be able to handle covid, why are we still in lockdown? Why has it been so difficult to access covid deaths by age group? It appears that covid deaths under 50 are only one percent and covid has killed as many people under 30 as over 100. Why is this not more widely known? Because it illustrates that this is (similar to the flu) that primarily kills people who are already sick. Although I expect Sweden to come under increasing pressure to lockdown, if it doesn’t and ultimately if Sweden’s mortality rate is not much different than the final US mortality rate then what was the point of the lockdown? While flattening the curve lengthens the covid outbreak, (and as a result not overload the healthcare system and hopefully save some lives) it was never intended or expected to limit the spread of the virus, which is something that is just not possible, unless the population stays in quarantine. Will the covid response greatly expand the govt? Yes! Did China have an interest in exaggerating the mortality of the virus to clear the streets of Hong Kong and possibly reduce Trump’s chance of reelection. Yes! Did the CCP do this? Who knows, but if something is in a nation’s interest there’s a good chance they might try it. Also, Covid deaths are way below projections and people are wondering what all the hysteria was about. Expect to see a big jump in Covid deaths because it is in the bureaucracy’s interest to count and reclassify as many deaths as possible as Covid deaths. Will China try to create or exaggerate a second wave to squash the Hong Kong protests again? It would seem to be in the CCP interest. Will Russia and China do everything possible to diminish Trump’s chances of reelection? I think so. Russia’s main source of income is fossil fuel. A democrat may ban fracking in the US, which will increase the price of fossil fuels (but not reduce CO2) and as result, Putin’s income. Will the lockdown ultimately be shown to have killed more people than it saves? I think this is almost a foregone conclusion. Your pal throughout the centuries, J
====================
Make of that what you will.

This did teach me several things. Most importantly, what I want from my friend is to put a stake in the ground about something that represents his world view, that is testable, that differs from my world view. Lastly and most importantly, we need to tell each other in advance what a failure of that test means for our world view. Instead of deciding to agree on the bet based on how much money we are willing to risk it should be based on how much of our world view are we willing to risk. If no result of the test results in a change of world view, a renewed questioning, then there is no point in the bet.

Reply

admin April 26, 2020 at 11:15 am

Hi R.R., When I read this, I see, “Yada yada yada yada yada yada yada yada yada yada yada yada yada I will pay the bet yada yada yada yada yada yada…”

It’s disappointing but not surprising. At least he said he will pay the bet. I half-expected him to deny he made the bet, or deny that he lost.

By the way, he was off by a factor of 15 (not 8 — I was too charitable to him in my post). It’s like driving 300 mph in a 20 mph zone. Not exactly a small mistake.

Reply

R.R. April 26, 2020 at 1:49 pm

See, I think he almost claimed that he won.

“I suppose if I just wanted to win the bet I might have suggested that daily deaths would have been say 2000 and whoever was closest would win. You might have taken the bet at that time because projected US deaths were so much higher on that date.”

He seems to be saying that at the time he knew 2000 was a more reasonable guess and that I probably did not know that. I probably would think 2000 is too small and instead place my money on 20000 covid-19 deaths on a single day. In other words, if he had wanted to, he already knew enough to make a bet where he would be close and I would not know enough to make a better guess. He just did not want to. He has always understood this better than I did. Retroactively, I declare myself the winner!

This is straining an old friendship….

Reply

admin April 26, 2020 at 3:02 pm

I’m sorry to hear the last sentence, although again I’m not surprised. One thing I thought about asking, both when you made the original bet and then this morning when you gave me the update was, “How come you’re still friends with this guy?” But I didn’t say it, in part because I think it’s very important to recognize the humanity in each other and be able to embrace each other in spite of our differences.

Nevertheless, his behavior to me is very reminiscent of the 7-year-olds in my chess club who, after losing a game by checkmate, knock down all of their opponents’ pieces and proclaim, “I won!” This happens fairly often, and every time the child thinks he is being so clever and original. It is perhaps a necessary emotional stage for 7-year-olds to pass through, but it is sad to see it in an adult.

Reply

R.R. April 26, 2020 at 2:25 pm

By the way, from the Los Angeles Times on April 23 2020

Swiss banking giant UBS told clients Monday that Walt Disney Co. is likely to wait until Jan. 1 to open its theme parks and predicted the Burbank media company will see only about 50% of 2019 attendance.

No, I was never *sure* of this bet. I just did not see yet how they could open until, somehow, they could be sure they would not be a Covid-19 supersite. Why go to the trouble and expense of re-opening until they had a sure fire plan to avoid a headline about Disneyland and new covid-19 cases plastered together. And it could all change in one direction or another.

Reply

Leave a Comment

Previous post:

Next post: