Congratulations to yukiblue for winning against TwixtBot while making it look easy! An astonishing feat for sure. 3.i12 doesn't look very characteristic of TwixtBot and 15.s8 seems like a clear mistake. I wonder what happened under the hood?
On
2021-09-03 at 17:13,
yukiblue
said:
Thanks for your comment, mmKALLL. I'm glad I could beat TwixtBot after you in Japan. 3.i12 is not like the usual TwixtBot, I thought so too. But then, it occurred to me that if black was directly linked to something like i9 or i7, he may have wanted to create a route to either v20 or a connection to i12 by occupying the top right corner.
4.k12 was an experimental move for me, but I thought that even if white moved to i8, there would still be a way to get to the left side, and I wanted to be able to connect to the right side from either k8 or k12, so I moved it.
I didn't make any mistakes in my subsequent moves either, which I think led to my victory.
The 15.s8 was the answer with a little thought, but in my opinion it is the move that people are most likely to get wrong. I think that's why TwixtBot resisted.
I feel that coming up with 4.k12 was the factor that allowed me to beat TwixtBot. Since it's a move I don't usually see much of, it's possible that it was a move that wasn't valued in TwixtBot's data.
I've fought over 600 TwixtBot battles before this victory. So happy to have won!
Sorry for the weird English, it's an automatic translation.
On
2021-09-06 at 19:14,
Peyrol
said:
Yes congratulations to yukiblue! Especially for your perseverance in playing so many games against the bot!
Maybe BonyJordan could tell us what the computer evaluations were at each move. |15.r8 looks much better for white, because of the peg on V20. But after 16.p1017.s1018.s1319.t1220.r1521.u1422.u18 it looks like black still wins. Maybe the bot did not realize this until it was already committed to this defense on the right side. Sometimes, when the computer sees that the main line is losing, it will look at alternatives until it runs out of time, and then it will play a bad move because it did not have enough time to see that this bad move also loses.
Even as early as move 7, white apparently assumed that white N10 was a threat. My guess is that 7.k13 was the losing move |7.g8 looks much better.
On
2021-09-06 at 19:40,
Peyrol
said:
Actually, |15.r816.p1017.t12 looks like the crucial line for white. One possible continuation is 18.r1119.s1020.s1421.u1422.u1823.s1524.p1825.q14 and I cannot tell who should win. I am sure the bot looked at this when it played 7.k13. So maybe at that point the bot thought white wins, then later realized it was losing.
3.i12 doesn't look very characteristic of TwixtBot and 15.s8 seems like a clear mistake. I wonder what happened under the hood?
I'm glad I could beat TwixtBot after you in Japan.
3.i12 is not like the usual TwixtBot, I thought so too.
But then, it occurred to me that if black was directly linked to something like i9 or i7, he may have wanted to create a route to either v20 or a connection to i12 by occupying the top right corner.
4.k12 was an experimental move for me, but I thought that even if white moved to i8, there would still be a way to get to the left side, and I wanted to be able to connect to the right side from either k8 or k12, so I moved it.
I didn't make any mistakes in my subsequent moves either, which I think led to my victory.
The 15.s8 was the answer with a little thought, but in my opinion it is the move that people are most likely to get wrong. I think that's why TwixtBot resisted.
I feel that coming up with 4.k12 was the factor that allowed me to beat TwixtBot.
Since it's a move I don't usually see much of, it's possible that it was a move that wasn't valued in TwixtBot's data.
I've fought over 600 TwixtBot battles before this victory.
So happy to have won!
Sorry for the weird English, it's an automatic translation.
3.i12 is too early in the game for me to judge.
Maybe BonyJordan could tell us what the computer evaluations were at each move. |15.r8 looks much better for white, because of the peg on V20. But after 16.p10 17.s10 18.s13 19.t12 20.r15 21.u14 22.u18 it looks like black still wins. Maybe the bot did not realize this until it was already committed to this defense on the right side. Sometimes, when the computer sees that the main line is losing, it will look at alternatives until it runs out of time, and then it will play a bad move because it did not have enough time to see that this bad move also loses.
Even as early as move 7, white apparently assumed that white N10 was a threat. My guess is that 7.k13 was the losing move |7.g8 looks much better.