What is the earliest move AI has said you lost in an even game?

So it went wrong, meaning that it was a bad move, but it wasn’t so bad to lose the game, so not accurate?

(I’m useless with AI and I’ll never get better. :frowning: )

I do not put my games through AI so far, but I’d guess this game since I surrendered at move 29 :stuck_out_tongue:

https://www.dragongoserver.net/game.php?gid=1160686

it’s been three years since then ? wow … time flies.

3 Likes

Were you even losing?!

2 Likes

Were you even losing?!

Yup. The opponent was 4 ranks better and had huuuuuuge moyos and two corners, so I’d have to reduce all game long against a much better opponent. Pretty horrible direction of play on my behalf.

3 Likes

I’ll take over from black if you are interested. See how badly I can screw up a game that warranties resigning at move 29. :slight_smile:

Move 43: the opponent (White) is at 99% winrate.

How long does it take for your games to be decisive? .

3 Likes

My earliest recently was move 9. Just 18% to win. :sweat_smile:

4 Likes

Someone play a game and pass for a couple of moves with standard komi. You’ll win this comp :slight_smile:

I once attempted to watch someone pass after they played one move. It was the most frustrating thing.

There is no rules about how much the AI says you have to be losing by. I could play a normal 9x9 game and do this, not even on purpose. I don’t need to pass for a bunch of turns. That also just defeats the purpose of playing if you aren’t on to troll people. Why sign up for the game if you don’t care and are going to pass the whole game.

1 Like

I have a question somewhat relating to this topic.

Why does the AI give me 100% chances of losing in some games, even if I win that same game(I am sure it has happened the opposite way, but I don’t look at all of my games that I lose. I understand that my opponent made a bad move or something, but clearly they don’t have a 100% chance of winning if I win these games.

Some examples:

https://online-go.com/game/view/24124575
https://online-go.com/game/24020102
https://online-go.com/game/22755357

1 Like

AI assumes its playing with itself, and assumes that both players play the optimal end-game. It does not take the human-errors as a factor. Like on that forked game with gia, black had won the game from where you start and AI saw 0% change of white winning as long as the black plays the best responses (which didnt happen)

TBH i think you’re right saying that it should say “99.9%” and not “100%” at best, because of couse there is always a tiny chance of losing, regardless how much you are ahead on board. Like you know, maybe your network connection goes down when youre about to pass and you lose on time. Stuff like that, the outcome is never fully 100% certain before both players have passed.

5 Likes

I assume it displays 100% when the winrate gets equal to or above 99.95%.

And of course, technically the outcome is actually either 100% or 0% at any point, assuming optimal play. Bar from the absolute end of the game (at which point it becomes combinatorial game theory and strategies can be proven to be optimal by us mortals), it’s just us that can’t really be certain.

4 Likes

Then why does it still have 99.9% chances?

Probably for when the win rate is between 99.85% and 99.94%.

4 Likes

Just to put my shame on full display, may I remind everyone we switched colors at the beginning of the forked game.

1 Like

Hey! I am sure I can find a game that I look terrible in. I’ll post it after I feel like finding the game I am thinking of(and I am not looking right now)

You know what, screw that. I don’t feel like digging for games.

What about largest percentage change?

I have -195.6% in 1 move.

I think this beats that


Also it isn’t a change of 195%, it is a change of about 96% in your game.

Oh you’re right! The winning percentage sum is always 100 :smiley: