PRIVATE ProteusSFX-AI tests

Code, algorithms, languages, construction...
AlexChess
Posts: 504
Joined: Sat Jan 08, 2022 9:25 am
Real Name: Alex Morales

Re: PRIVATE ProteusSFX-AI tests

Post by AlexChess » Mon Dec 25, 2023 9:15 am

Proteus-SFX-AI 221109 vs Incognito 5
Partial result: 5-0 / 187 games

Even without unbalanced suite it's weak. Looking for a leaked daily released Big Light 15. Maybe Ediot has been inspired by Santa for his last clowne :mrgreen:

Image

AlexChess
Posts: 504
Joined: Sat Jan 08, 2022 9:25 am
Real Name: Alex Morales

Re: PRIVATE ProteusSFX-AI tests

Post by AlexChess » Mon Dec 25, 2023 9:52 am

PLAYCHESS RANKING - 231225 UPDATE


Image

Right now ProteusSFX-AI is first with 2517 bullet, 3339 blitz and 2505 long. You can check all the results on my profile, NO SIBLINGS .
Big Light 15 even running on IG88-A (88 threads) , A2-A3 and Detlef Uter powerful hardware are very far from it.
EDIOT has been removed from Banned from Leaderboard, due to sibling even with Leaderboard Team leaders,
that change their rules according to what they need for their friends :lol: Ibaibur you are disappointing.. I was thinking that we were friends again..

Image

Martin_1969
Posts: 27
Joined: Mon Jul 03, 2023 6:53 am

Re: PRIVATE ProteusSFX-AI tests

Post by Martin_1969 » Mon Dec 25, 2023 11:42 am

AlexChess wrote:
Mon Dec 25, 2023 9:15 am
Proteus-SFX-AI 221109 vs Incognito 5
Partial result: 5-0 / 187 games

Even without unbalanced suite it's weak. Looking for a leaked daily released Big Light 15. Maybe Ediot has been inspired by Santa for his last clowne :mrgreen:

Image
+/- 17 ;)

AlexChess
Posts: 504
Joined: Sat Jan 08, 2022 9:25 am
Real Name: Alex Morales

Re: PRIVATE ProteusSFX-AI tests

Post by AlexChess » Mon Dec 25, 2023 10:09 pm

HAPPY FARMING CHALLENGE 3200 - 231225 UPDATE

REACH US HERE !!!

Image

AlexChess
Posts: 504
Joined: Sat Jan 08, 2022 9:25 am
Real Name: Alex Morales

Re: PRIVATE ProteusSFX-AI tests

Post by AlexChess » Mon Dec 25, 2023 10:15 pm

Martin_1969 wrote:
Mon Dec 25, 2023 11:42 am
+/- 17 ;)
It's the ELO error margin playing 187 games, 300 games = +-14 ELO. Playing 4200 games my error margin is only 8 ELO.

Also a +10 ELO points difference is a lot between Stockfish and derivatives.

AlexChess
Posts: 504
Joined: Sat Jan 08, 2022 9:25 am
Real Name: Alex Morales

Re: PRIVATE ProteusSFX-AI tests

Post by AlexChess » Tue Dec 26, 2023 7:55 am

Proteus-SFX-AI-231109-Incognito 5
Blitz 5+0
Final result: 10-0

Ediot, +/- 1 is normal statistically. 10-0 / 300 is a debacle!
You should consider to focus only on an engine, testing it
deeply for months, instead of releasing a new clowne each day :mrgreen:

(I will send all games to SARONA and dorsz, to verify the performance.)

Image

Martin_1969
Posts: 27
Joined: Mon Jul 03, 2023 6:53 am

Re: PRIVATE ProteusSFX-AI tests

Post by Martin_1969 » Tue Dec 26, 2023 9:37 am

AlexChess wrote:
Mon Dec 25, 2023 10:15 pm
Martin_1969 wrote:
Mon Dec 25, 2023 11:42 am
+/- 17 ;)
It's the ELO error margin playing 187 games, 300 games = +-14 ELO. Playing 4200 games my error margin is only 8 ELO.

Also a +10 ELO points difference is a lot between Stockfish and derivatives.
1. you only test derivatives against derivatives.
The father is still the strongest.
2. if there is a +10 but the errorbar is +/- 14 it means the next 300 games could also result in a -4. Or am I wrong in my interpretation of the errorbar?
But keep up your testing. I'm not going to waste your thread and my time interpreting your results correctly.
Bye

AlexChess
Posts: 504
Joined: Sat Jan 08, 2022 9:25 am
Real Name: Alex Morales

Re: PRIVATE ProteusSFX-AI tests

Post by AlexChess » Tue Dec 26, 2023 9:39 am

ProteusSFX-AI 231109 Banksiagui Gauntlet
231226 Update - 4500 games

Image

Next match is against Catropoly 3.3 (Sorry, I cannot find 4 2xNNUE)

Looking for leaked Predator AI

AlexChess
Posts: 504
Joined: Sat Jan 08, 2022 9:25 am
Real Name: Alex Morales

Re: PRIVATE ProteusSFX-AI tests

Post by AlexChess » Tue Dec 26, 2023 9:42 am

Martin_1969 wrote:
Tue Dec 26, 2023 9:37 am
AlexChess wrote:
Mon Dec 25, 2023 10:15 pm
Martin_1969 wrote:
Mon Dec 25, 2023 11:42 am
+/- 17 ;)
It's the ELO error margin playing 187 games, 300 games = +-14 ELO. Playing 4200 games my error margin is only 8 ELO.

Also a +10 ELO points difference is a lot between Stockfish and derivatives.
1. you only test derivatives against derivatives.
The father is still the strongest.
2. if there is a +10 but the errorbar is +/- 14 it means the next 300 games could also result in a -4. Or am I wrong in my interpretation of the errorbar?
But keep up your testing. I'm not going to waste your thread and my time interpreting your results correctly.
Bye
I can only say that after 4500 games ProteusSFX-AI is very likely 3500 ELO, with an error margin of +/- 7 ELO and it performs very well compared to all other derivatives and last Stockfish itself. But my main purpose is to build two very strong books to compete on Playchess and Lichess.

tt-StockFish
Posts: 24
Joined: Tue Nov 07, 2023 7:24 pm
Contact:

Re: PRIVATE ProteusSFX-AI tests

Post by tt-StockFish » Tue Dec 26, 2023 2:23 pm

AlexChess wrote:
Mon Dec 25, 2023 9:52 am
PLAYCHESS RANKING - 231225 UPDATE


Image

Right now ProteusSFX-AI is first with 2517 bullet, 3339 blitz and 2505 long. You can check all the results on my profile, NO SIBLINGS .
Big Light 15 even running on IG88-A (88 threads) , A2-A3 and Detlef Uter powerful hardware are very far from it.
EDIOT has been removed from Banned from Leaderboard, due to sibling even with Leaderboard Team leaders,
that change their rules according to what they need for their friends :lol: Ibaibur you are disappointing.. I was thinking that we were friends again..

Image
Hi,
Just to clarify, we change our rules to make the leaderboards the more fair possible. Some others won't agree but we are doing a lot of same things than on lichess : We (lichess and us) ban all farming/cheating users. Also, only one account is allowed in the leaderboards on lichess (like nihalsarin, who is on the leaderboards and not his second account Junglebook1). We allow multiple accounts but we don't tolere farming. So, we banned all users farming on all their accounts. It's not like on lichess but it's almost the same system. For sure, only people which farmed were against our rules.

We removed Eduard from the banned from leaderboard of bots because he was too annoying with us, and because he wasn't eligible for the leaderboards (Rating deviation > 75). If his BOTs are eligible again for our leaderboards, he'll banned for sure.

Friendly,
tt-stockfish

Post Reply