HasiLover_Test wrote:

-089- wrote:

hey, can someone with a fast system please try out this engine real quick https://turbowarp.org/1038827110?stuck and tell me what NPS you're getting? for me it's around 100k-110k.
just a couple moves is enough, don't bother with a full game. something somewhere clearly isn't quite working, I'm thinking I probably broke some of the old code…

Its at 50K n/s for me, but it plays awful chess, not to insult you.

-089- wrote:

HasiLover_Test wrote:

-089- wrote:

hey, can someone with a fast system please try out this engine real quick https://turbowarp.org/1038827110?stuck and tell me what NPS you're getting? for me it's around 100k-110k.
just a couple moves is enough, don't bother with a full game. something somewhere clearly isn't quite working, I'm thinking I probably broke some of the old code…

Its at 50K n/s for me, but it plays awful chess, not to insult you.

yeah, I know… there's a bug somewhere that I haven't been able to fix. I know what doesn't work, I just don't know why lol

interesting that you and AZURUS both got 50k, makes me think there might be even more problems than I thought.
I don't think your system is actually half as fast as mine, there has to be another reason for that massive difference.
thanks a lot for the feedback

HasiLover_Test wrote:

-089- wrote:

hey, can someone with a fast system please try out this engine real quick https://turbowarp.org/1038827110?stuck and tell me what NPS you're getting? for me it's around 100k-110k.
just a couple moves is enough, don't bother with a full game. something somewhere clearly isn't quite working, I'm thinking I probably broke some of the old code…

Its at 50K n/s for me, but it plays awful chess, not to insult you.

ArnoHu wrote:

HasiLover_Test wrote:

-089- wrote:

hey, can someone with a fast system please try out this engine real quick https://turbowarp.org/1038827110?stuck and tell me what NPS you're getting? for me it's around 100k-110k.
just a couple moves is enough, don't bother with a full game. something somewhere clearly isn't quite working, I'm thinking I probably broke some of the old code…

Its at 50K n/s for me, but it plays awful chess, not to insult you.

300k midgame, 400k start board.

Your link activates warp timer setting (stuck); interestingly this does not seem to influence your engine. GoK would lose two plies with that one.

-089- wrote:

ArnoHu wrote:

HasiLover_Test wrote:

-089- wrote:

hey, can someone with a fast system please try out this engine real quick https://turbowarp.org/1038827110?stuck and tell me what NPS you're getting? for me it's around 100k-110k.
just a couple moves is enough, don't bother with a full game. something somewhere clearly isn't quite working, I'm thinking I probably broke some of the old code…

Its at 50K n/s for me, but it plays awful chess, not to insult you.

300k midgame, 400k start board.

Your link activates warp timer setting (stuck); interestingly this does not seem to influence your engine. GoK would lose two plies with that one.

great! that's actually better than what I was hoping for, thanks for testing it!

about the warp timer setting, I work in turbowarp desktop, which has warp timer on by default and I've never really tried turning it off before. I didn't know that it could be a problem.
honestly the only reason I had that setting enabled in the link is to show the search progress thingy…

-089- wrote:

-089- wrote:

HasiLover_Test wrote:

-089- wrote:

hey, can someone with a fast system please try out this engine real quick https://turbowarp.org/1038827110?stuck and tell me what NPS you're getting? for me it's around 100k-110k.
just a couple moves is enough, don't bother with a full game. something somewhere clearly isn't quite working, I'm thinking I probably broke some of the old code…

Its at 50K n/s for me, but it plays awful chess, not to insult you.

yeah, I know… there's a bug somewhere that I haven't been able to fix. I know what doesn't work, I just don't know why lol

interesting that you and AZURUS both got 50k, makes me think there might be even more problems than I thought.
I don't think your system is actually half as fast as mine, there has to be another reason for that massive difference.
thanks a lot for the feedback

my own stupidity never fails to amaze me, that was such an obvious mistake…

it should be fixed now. it still plays pretty bad positionally, but that's (more or less) by design. I'm still not sure why it would be so much slower on your system, would you mind giving it another try? I want to see if fixing that bug changed anything

AZURUS41 wrote:

-089- wrote:

-089- wrote:

HasiLover_Test wrote:

-089- wrote:

hey, can someone with a fast system please try out this engine real quick https://turbowarp.org/1038827110?stuck and tell me what NPS you're getting? for me it's around 100k-110k.
just a couple moves is enough, don't bother with a full game. something somewhere clearly isn't quite working, I'm thinking I probably broke some of the old code…

Its at 50K n/s for me, but it plays awful chess, not to insult you.

yeah, I know… there's a bug somewhere that I haven't been able to fix. I know what doesn't work, I just don't know why lol

interesting that you and AZURUS both got 50k, makes me think there might be even more problems than I thought.
I don't think your system is actually half as fast as mine, there has to be another reason for that massive difference.
thanks a lot for the feedback

my own stupidity never fails to amaze me, that was such an obvious mistake…

it should be fixed now. it still plays pretty bad positionally, but that's (more or less) by design. I'm still not sure why it would be so much slower on your system, would you mind giving it another try? I want to see if fixing that bug changed anything

- 75k with the “?stuck” and without the turbo
- 78k with the “?stuck” and with the turbo
- 104k without the “?stuck” and without turbo
- 120k without the “?stuck” and with the turbo

I'm faaaaaaaar away from the 400k that Arno has, but we all know that he has a whole datacenter to run our bots so it doesn't count…

ArnoHu wrote:

AZURUS41 wrote:

-089- wrote:

-089- wrote:

HasiLover_Test wrote:

-089- wrote:

hey, can someone with a fast system please try out this engine real quick https://turbowarp.org/1038827110?stuck and tell me what NPS you're getting? for me it's around 100k-110k.
just a couple moves is enough, don't bother with a full game. something somewhere clearly isn't quite working, I'm thinking I probably broke some of the old code…

Its at 50K n/s for me, but it plays awful chess, not to insult you.

yeah, I know… there's a bug somewhere that I haven't been able to fix. I know what doesn't work, I just don't know why lol

interesting that you and AZURUS both got 50k, makes me think there might be even more problems than I thought.
I don't think your system is actually half as fast as mine, there has to be another reason for that massive difference.
thanks a lot for the feedback

my own stupidity never fails to amaze me, that was such an obvious mistake…

it should be fixed now. it still plays pretty bad positionally, but that's (more or less) by design. I'm still not sure why it would be so much slower on your system, would you mind giving it another try? I want to see if fixing that bug changed anything

- 75k with the “?stuck” and without the turbo
- 78k with the “?stuck” and with the turbo
- 104k without the “?stuck” and without turbo
- 120k without the “?stuck” and with the turbo

I'm faaaaaaaar away from the 400k that Arno has, but we all know that he has a whole datacenter to run our bots so it doesn't count…

Haha, how do you guys even work in the Scratch editor with thousands of blocks? I only have a small notebook (ASUS Zenbook, OK it has an Intel i9-13900H, and 16g memory)

ArnoHu wrote:

AZURUS41 wrote:

-089- wrote:

-089- wrote:

HasiLover_Test wrote:

-089- wrote:

hey, can someone with a fast system please try out this engine real quick https://turbowarp.org/1038827110?stuck and tell me what NPS you're getting? for me it's around 100k-110k.
just a couple moves is enough, don't bother with a full game. something somewhere clearly isn't quite working, I'm thinking I probably broke some of the old code…

Its at 50K n/s for me, but it plays awful chess, not to insult you.

yeah, I know… there's a bug somewhere that I haven't been able to fix. I know what doesn't work, I just don't know why lol

interesting that you and AZURUS both got 50k, makes me think there might be even more problems than I thought.
I don't think your system is actually half as fast as mine, there has to be another reason for that massive difference.
thanks a lot for the feedback

my own stupidity never fails to amaze me, that was such an obvious mistake…

it should be fixed now. it still plays pretty bad positionally, but that's (more or less) by design. I'm still not sure why it would be so much slower on your system, would you mind giving it another try? I want to see if fixing that bug changed anything

- 75k with the “?stuck” and without the turbo
- 78k with the “?stuck” and with the turbo
- 104k without the “?stuck” and without turbo
- 120k without the “?stuck” and with the turbo

I'm faaaaaaaar away from the 400k that Arno has, but we all know that he has a whole datacenter to run our bots so it doesn't count…

Haha, how do you guys even work in the Scratch editor with thousands of blocks? I only have a small notebook (ASUS Zenbook, OK it has an Intel i9-13900H, and 16g memory)

ArnoHu wrote:

GoK NNUE (Difficult, white) against Element (6+8), several blunders, one decisive, 91% vs. 80% accuracy: https://lichess.org/lePZmNOv#115

Replayed same opening with reduced search extensions, completely different game, GoK wins in 19 moves: https://lichess.org/ewhsnzmB#37

GoK NNUE (10sec, white) vs. White Dove (P2), GoK NNUE wins in 25 moves: https://lichess.org/ZokvjTpV#49

birdracerthree wrote:

ArnoHu wrote:

GoK NNUE (Difficult, white) against Element (6+8), several blunders, one decisive, 91% vs. 80% accuracy: https://lichess.org/lePZmNOv#115

Replayed same opening with reduced search extensions, completely different game, GoK wins in 19 moves: https://lichess.org/ewhsnzmB#37

GoK NNUE (10sec, white) vs. White Dove (P2), GoK NNUE wins in 25 moves: https://lichess.org/ZokvjTpV#49

I tried to replicate 19…Qd8?? and White Dove finished ply 7 and selected Qh5 at the last moment (turbo mode, warp timer for cleaner progress bar).
Is White Dove running with Turbo Mode?

Element sees Bxh6 on depth 6 If only I could run king ring without attack tables…

Quick question, I am currently planning to make a change to Element, but I want to change the second decimal digit to represent the update. What do I do?

ArnoHu wrote:

GoK NNUE (Difficult, white) against Element (6+8), several blunders, one decisive, 91% vs. 80% accuracy: https://lichess.org/lePZmNOv#115

Replayed same opening with reduced search extensions, completely different game, GoK wins in 19 moves (90% vs 62% accuracy): https://lichess.org/ewhsnzmB#37

GoK NNUE (10sec, white) vs. White Dove (P2), GoK NNUE wins in 25 moves (94% vs 78% accuracy): https://lichess.org/ZokvjTpV#49

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

GoK NNUE (Difficult, white) against Element (6+8), several blunders, one decisive, 91% vs. 80% accuracy: https://lichess.org/lePZmNOv#115

Replayed same opening with reduced search extensions, completely different game, GoK wins in 19 moves: https://lichess.org/ewhsnzmB#37

GoK NNUE (10sec, white) vs. White Dove (P2), GoK NNUE wins in 25 moves: https://lichess.org/ZokvjTpV#49

I tried to replicate 19…Qd8?? and White Dove finished ply 7 and selected Qh5 at the last moment (turbo mode, warp timer for cleaner progress bar).
Is White Dove running with Turbo Mode?

Element sees Bxh6 on depth 6 If only I could run king ring without attack tables…

Quick question, I am currently planning to make a change to Element, but I want to change the second decimal digit to represent the update. What do I do?

Yes, I ran it on turbo mode. And when I just import the FEN, it also plays Qh5, easily within time. But try to import the board one move before that and replay two moves, and WD will go for Qd8. The transposition table might have an effect here.

I don't quite understand the last question though.

birdracerthree wrote:

ArnoHu wrote:

birdracerthree wrote:

ArnoHu wrote:

GoK NNUE (Difficult, white) against Element (6+8), several blunders, one decisive, 91% vs. 80% accuracy: https://lichess.org/lePZmNOv#115

Replayed same opening with reduced search extensions, completely different game, GoK wins in 19 moves: https://lichess.org/ewhsnzmB#37

GoK NNUE (10sec, white) vs. White Dove (P2), GoK NNUE wins in 25 moves: https://lichess.org/ZokvjTpV#49

I tried to replicate 19…Qd8?? and White Dove finished ply 7 and selected Qh5 at the last moment (turbo mode, warp timer for cleaner progress bar).
Is White Dove running with Turbo Mode?

Element sees Bxh6 on depth 6 If only I could run king ring without attack tables…

Quick question, I am currently planning to make a change to Element, but I want to change the second decimal digit to represent the update. What do I do?

Yes, I ran it on turbo mode. And when I just import the FEN, it also plays Qh5, easily within time. But try to import the board one move before that and replay two moves, and WD will go for Qd8. The transposition table might have an effect here.

I don't quite understand the last question though.

I want to change the ‘9’ in v1.4921, but I can’t really do that (base 10).

ArnoHu wrote:

I gave GoK NNUE now that particular AI Look & Feel (intro screen, board), hope you like it: https://scratch.mit.edu/projects/1033379869/