Necronomicon 💀

To collect the deadest of dead things. I was going through some old games when I found this gem, so I’ll start:

From my 11x11 game against thelovingsun. Take a second to count all of my living groups (yes, I decided to go to counting anyway)

Edit: link to game: FOR THE HONOR OF GAYSKULL


A big risky kill wandering through my moyo that was a nerve-wracking few months. This is the kind of thing people tell you not to do with “attack for profit, not to kill”, but then WWLSD*?

* What Would Lee Sedol Do?


Wow that’s indeed very, very dead. A worthy addition to the collection :skull_and_crossbones::headstone:

Is it ever weird looking back at pre-AI games and suddenly having the OGS AI analysis on them? I noticed that the move you considered likely to be the losing move still has an evaluation of Black ahead by 11 points. Would you still consider it a “losing move” in spite of that? (Edit: i just realized this phrasing implies that you lost which I know you didn’t but I hope my question comes across clearly anyway)

I wish I could determine life/death here. I’m only 9 k. To me, in my ignorance, it looks like the game is not over. For example, the status of the upper middle w group depends on whether w or black plays on f8 and whether all the surrounding b groups are alive. Are you claiming that all your w groups are dead? I don’t see that.

1 Like

Black doesn’t have to play f8. If white plays f8 white gets an eye, but white can’t make another eye.
In my opinion, everything is dead


Also worth noting that the lower left black group, which has a dead-ish look to it, is actually locally alive in seki.

So all the black groups are locally alive, which makes the situation easier to analyze (no need to read out a capturing race).


I’ve only looked at a few of my old commented games now with the free AI analysis, but it’s interesting to compare. It does confirm my thoughts that I often won tough games in endgame e.g 19x19 ladder challenge: nrx vs Uberdude. For this move 119 with a big life and death problem that depends on a shortage of liberties to make 1 eye and a ko for the other I would caution about putting too much trust in the free shallow AI analysis. It’s possible my historical comments are more accurate, though do note I said it could still be a game if I killed the right side group in the ko or depends on what happens in lower left. To reach firm conclusions I’d want to explore variations interactively with the bot at thousands of playouts. But it’s morally a lost game if white lives after I chased him through my territory trying to kill :slight_smile:. Something the AI analysis highlights is importance of m3 and White’s missed opportunity to connect there in sente. Even though its over a decade ago, I still remember realising my mistake there and relief I could come back to repair before he exploited it, so it’s nice to see the AI agreeing. You can see in the game how white got the g11 cut and because I played thinly to keep white down to 1 eye he could counterattack and it was me worrying about eyes for my top left group that gave him a chance to live, so if m3 was white g11 would be making two not one unsettled black group and I couldn’t have attacked so aggressively (because if I did one of my groups would probably die before his huge dragon).