Are you changing openings as a result of Alpha Zero?

From my reading and video viewing about this AZ, I’m trying out 1. C4 to replace e4.  Haven’t started playing 1. E5 as black yet, as I am a suicidal Sicilian guy.  I know AZ plays d4 rather than c4, but I don’t care, yet.  Are you players throwing out your openings yet?

 Personally I used to play the French Defense a lot as black, but I gradually gave it up for reasons such as lack of space. My favorite opening is likely the Queen's Gambit as white (although I only began to play it because I did not understand it at all, and wanted to learn it - which there is no faster way to learn an opening than to play it. Of course, I was pleasantly surprised to see my personal story have a lot of parallelism with Alpha Zero. Alpha Zero also slowly gave up the French Defense, and seems to like the Queen's Gambit (and English Opening too). However, I see this as perhaps only a cool coincidence. I would not recommend changing openings (especially a repertoire) just because an AI seems to use it often. People should play solid, respectable, openings that feel more natural to them. If you love 1.e4 then continue to use it (I used to play 1.e4 before I fell in love with 1.d4). If you like the Reti, or some other solid opening - I encourage that. Additionally, Alpha Zero "learns" from its games, so maybe in the future it will play the Queen's Gambit or English Opening less, and shift to an opening that you loved. As my opening choices mostly matched Alpha Zero's I am not changing, but if Alpha Zero later decides on another opening repertoire, I may not change still. I am curious to hear other thoughts regarding your question though. Obviously I would not let it impact me, but perhaps others will. Computers have such a high ability compared to me in chess, so I am not going to learn an opening that I feel uncomfortable with, just because a computer uses it often. The computer has tactical awareness that I do not, therefore I would rather play in my comfort zone, than to wander into what a computer knows but I see as unknown. Great forum topic, although I am really more interested in what Alpha Zero can do outside of chess, in math, or science etc.

Thanks.  I’ve only played maybe 120 games on this site, so haven’t made a big investment yet.  I also disliked French defense as black as well.  I started with e4 because of the common belief that it generally leads to quick tactics, which need to be learned.  And because of Fischer’s saying that it’s most principled.  I hope more people weigh in on this, however I imagine most posters have already invested time in whatever openings, so they will stay with them.  I don’t know what AZ preferred against 1. D4, but I just go with d5 because it’s safe.  The KID has rep for danger, and I think the Dutch is just weaker for black.

Unless you play like AZ, which you don't, then it makes no sense to model your openings off of AZ. 

I'm too new to have established any set of standard openings; however, I have been trying a variation of the Retie (I believe) that I read about in a chess book. It is a little different and at my level of play tends to be unrecognized by my (only 4 so far) opponents. I think what Hotspur said make a lot of sense, why change what you know if you're not up against AZ?  

I am using AZ to change all facets of my play.

Well the idea is AZ is the strongest chess player on the planet, so logically, if it plays an opening, it’s theoretically a better than others, and if it doesn’t play, for example, the French, it’s for a good reason.  It’s weaker than other defenses.

it's silly in general to use a computer as an openings reference.  it is a great tool, but only if you know how to use it although a competent statistician can do reasonably well regardless as I have seen in corr.  the pt is to get your pieces out and have active plans.  a computer can be very stupid in this regard and can be rolled over if it fails to follow these principles.  of course give a computer an active plan and another computer without the appropriate heuristics or horizon or a human will not be a match. 

I play the Caro as black, so I was quite interested in seeing the statistics on that opening. AlphaZero was happy to play it quite a lot, but for some reason it tailed off at the end.

I'm curious about whether the reason for this was because AlphaZero decided the Caro wasn't all that good, or if it's because AlphaZero shifted away from playing e4 as white. After all, you don't get many Caros if white doesn't play e4. 

It would be interesting to see what would happen if you forced AlphaZero to play some of its less favored openings, like the King's Indian. 

My point is, if Google is to believed, they gave AZ the rules.  It played itself a bunch of times.  It then beat the strongest chess playing machine on the planet (setting aside objections over software versions and hardware), and showed distinct preferences regarding openings and defenses.  This seems significant, even to poor humans who may endeavor to push wood.

I also play Caro-Kann a lot as black. Sometimes, if white plays d4, I play c6 inviting white to play Caro-Kann, but often I answer d4 with Nf6 and play Benoni. In the past I often played Kings Indian Defence then I decided some years ago to explore Benoni. I don't think Benoni was part of AlphaZero's play but Kings Indian Defence was. I will probably still explore Benoni though.

Of course, Im now starting my games with 1d4. Unfortunately, none of my opponents play the queens indian, which A0 demonstrated is crap for black by refuting it (games 3, 5 and 9)

Hi, they released only 10 games, so that is a very small and biased sample. If you look at the paper,... I would prefer if the graphs showed some convergence... A0 is by definition trying to achieve perfect play which is an extremely non-linear and discontinuous objective, so its preference will change over time. The percentages they provide do not add up to 100%. In fact merely ~30% just looking at the plots so most of the self-plays are still hidden.
They should train it on even bigger hardware, and for weeks on end. It will eventually stabilize and always play the same game over and over. We will be excited about the results. If they repeat this often enough (and is repeated independently elsewhere) and we always get the same game, then we will have 'statistically' solved chess. I think it will be a draw...
And no, I still study openings so I try to be good against casual players, not yet trying to play perfectly...

from c5 to c6 make plays like this

 no

That was a beautiful attack Adumbrate!

Let us assume that the above statement is correct (which we have no prove it is): unless you are playing someone > 2700 (and you are that high rated yourself), you could pretty much win or lose with every opening.

There is no reason to change my opening repertoire due to the existence of a machine that I'll never play against. But if everyone else stops playing french I expect my statistics with that opening to improve thanks to the lack of knowledge on my opponents side. So be my guest and follow in Alpha Zero's footsteps.

It's not...never in your lifetime will you or any opponent you play be within 1000 miles of competing at that level, so use whatever opening you like.  You might as well try to sprint a 5 second 100 yard dash because of how fast a Tesla goes 0-60.  It's utterly meaningless to you.

As for AlphaZero vs. other engines...wait for something official before you crown any gods.

The ONLY players this alpha zero thing will affect are the world elite.  The rest of us?  Lets all quit pretending we are good enough that we need to change things because of it.