SedatChess

[Sedat Canbaz]

UPDATE 5

The Winner (via UHO_Lichess_4852_v1 Epd openings) is
Kookaburra 3.10 - My Congratulations to DorsZ, SF team!
In short, Kookaburra is just UNSTOPPABLE.. as real BEAST for sure !!

Where unfortunately Omega is last place..and now 100% clear that
Omega MPV engine suffers a lot via UHO_Lichess_48521 openings..
Btw, via Unique openings: Omega's performance is one of best !!

On other hand, just I's like to point out that too:
HypnoS++ 1.02 at any recent opening test cond. suffers much too!
And let's hope M.Z will improve his engine (in future releases...)

Code: Select all

Rank Name                Elo    +    - games score oppo. draws 
   1 Kookaburra 3.10    3869   13   13  1040   54%  3840   49% 
   2 Corchess 210325    3864   13   13  1040   53%  3841   48% 
   3 SF-X 130425        3859   13   13  1040   53%  3841   48% 
   4 Stockfish 020425   3859   13   13  1040   53%  3841   50% 
   5 SF PRO2 190325     3859   13   13  1040   53%  3841   49% 
   6 KillfishPB 110425  3848   14   14  1040   51%  3843   48% 
   7 Artemis 12TR       3845   13   13  1040   50%  3843   50% 
   8 HypnoS++ 1.02      3796   13   13  1040   42%  3849   50% 
   9 Omega MPV          3790   13   13  1040   41%  3850   49% 

GAMES:
https://mega.nz/file/iloDCLwJ#2RPVI4Uyw ... G_F56n4Duc

And please stay tuned..soon as possible,
New and more details plus stats coming...

Greetings

[Sedat Canbaz]

Meanwhile,
Let Me say also please some notes more,
And just for our Computer Chess's progress!

1st of all,
We have to ask ourselves, in recognizing truth about NNUE:
- It is simply playing/gaining with previous trained data!
If I am missing something here...please correct Me...

On other hand,
The well-updated NNUE files + strong openings lead to
Much more draws, this is as real fact! e.g in the recent
Official tour, we noticed about 92% overall draw ratio!
Otherwise, if Unique suite was based on weak lines, then
Full of blunder moves would be born... and here just only
Who are Born-Idiots may have troubles to see the reality...!)

E.g as clear proof that Unique openings are not so Drawish:
https://sites.google.com/site/computers ... 1990s-cs-i
And if you need more statistics data...just let me know please...)

By the way, and once more:
Blaming the strong openings is just another MADNESS!
And please..I did not start the testings from yesterday!

Moreover,
Why we see different draw statistics (via my recent tours) ?
Simply because due to the 'strength' of the played openings!
In short, it seems Unique suite has less blunder moves in..!

For examples,
If X opening suite is full with blunders, critical, weak moves..
Then of course it's expecting to appear much more blunder moves!
In same time that means more wins...but more wins do not mean
Better..sure I refer here if determining the strength of engines!
In other words, it all depends..such as on our used tour cond.!

But this is also true that in case of using strong openings,
Less blunder moves are on scene for sure! proved, guaranteed!
And it is not required many thousands of games (per player),
E.g about 1.000 (per player) is enough for any kind of testing,
And it does not matter 2000+ or 3000+ or 3800+ Elo eng tours!

And as final words, for better eng Elo metrics:
I would prefer Bullet + Strong opening lines than
Testing the engines via Slower TC + Weak openings

In other words,
No way to run serious competitions via 'blundered' openings!
And it does not matter what you will use e.g SLOW TC, MP etc.

If still not so clear,
Weak Opening Line = The blunder move is already included in opening!
Where via strong opening suites..no much blunder moves are there...
And this is NOT as joke !! Hope helps )

And I hope too that
All my latest produced data to be useful as well !)

Greetings

[Anton101]

All the data is useful, and I'm very grateful for the final test. I can only congratulate Dorsz on this derivative engine, and I have no choice but to take an X-ray to Kokaburra; the work is extraordinary.

And thank you very much, Mr. Sedat, for taking the time and resources to do it.

Best regards and have a great weekend!

Thanks a lot.

[Sedat Canbaz]

Mr. Anton,

As usual, you are welcome )

Meanwhile,
You may notice or not e.g as good news is that
The win. per.s stats by UHO+Lichess 4852 are
56%-44%, where via Unique openings: 52%-48%

Just I'd like to point out that too,
52% - 48% stats are counted as most usual stats,
Which belonging to the strongest opening theory!

Btw, exactly same win perc. stats are produced in
Most SCCT Book Tours, and one of them with same
52%-48% are produced in latest organized XLIII CS:
https://sites.google.com/site/computers ... nnue-cs-43

One thing more, sometimes I wonder too,
If I run a new DUEL but with Time Control Odds:
Strong openings 30s+0.5s Vs Weak openings 2m+1s

Let's say Top books 30s+0.5s Vs Depth4 2025 2m+1s
Any idea who will win ?) I assume Depth4 since it
Will use much slower time control: Blitz 2m+1s

Moreover, according to our Troll on chess forums,
Deeds (botunnet) loves to say that fast Bullet Time
Controls are meaningless, plus full with blunders..but
To know about this mystery..(instead of Bla-Bla-Bla..)
A real 'proof' data is required, right ?)

Wishing you a wonderful weekend too )

Greetings

[supernova]

Hello, Sedat

I hope you are going and still thanks for your contest you have been running over the years.

In this reply, I will clarify the role and benefits of UHO (Unbalanced Human Openings) and address several points raised by yu for additional context.

1. Understanding UHO (Unbalanced Human Openings)

The UHO concept is simple: select openings that are fully legal, chess‑theory based, yet inherently unbalanced, giving White a tangible (sometimes extreme) advantage. These lines:

• Are entirely “human” in origin—no custom modifications like preventing castling or inventing unnatural configurations.

• Can potentially create more decisive outcomes (fewer draws) in engine testing, because White begins with a clear lead or dynamic potential.

• Serve as a way to highlight performance differences among engines quickly.

• One important misunderstanding is to equate “UHO” with “blundered lines.” In reality, “unbalanced” does not mean “inaccurate.” It simply means that from the very first moves, one side attempts a riskier setup. Such lines are still legitimate opening theory.

Your main arguments

• “Weak openings lead to more decisive results, but those are ‘blunders’ and do not reflect engines’ strength.”

While it is true that extreme blunders or outright mistakes serve little purpose in engine testing (because the outcome becomes trivial), UHO openings are not the same as “inserting deliberate blunders.” They are riskier or unusual but remain valid within standard theory. In engine competitions, these aggressive setups can demonstrate how well an engine handles imbalances—often more revealing than playing symmetrical or overly “perfect” opening lines.

• “Strong openings reduce blunders and thus produce more draws, giving us better data.“

Fewer decisive results do not always guarantee “better” or more reliable data. Indeed, obtaining a 92% draw ratio can flatten out meaningful rating fluctuations, forcing testers to require tens of thousands of games before seeing differences. On the contrary, obtaining a balanced number of decisive games (especially from unbalanced openings) can make rating gaps more statistically significant with fewer games, which is precisely a key reason behind adopting UHO sets.

• “Faster time controls plus strongest openings is always better.”

While bullet or blitz time controls can be entertaining and produce high volumes of games quickly, there is an ongoing debate in computer chess circles about whether such rapid conditions reveal the true strength of an engine. Many would suggest that slower time controls allow deeper calculation, revealing how engines handle complex middlegame or endgame scenarios. The choice between bullet and slower classical time controls depends heavily on the testing goal:

• If the goal is rapid throughput, bullet helps.

• If the aim is maximum depth of search, slower is preferable.

One last point.

• A Subtle Point About “Blunders in the Opening”

Sedat’s statements suggest that any line that is not “theoretically strong” amounts to a blunder. However:

• Chess theory is vast, and many offbeat lines (like UHO) are still far from outright blunders.

• Some lines are simply riskier and lead to sharper positions; that does not necessarily make them “bad testing lines.”

• Ultimately, a test suite that includes both “strong” and “unbalanced” lines can prevent engines from only memorizing “main lines” and drawish continuations.

Clarifying Anton101

• Multiple Opening Suites – The data references Unique, UHO, and UHO+Lichess 4852. These sets vary in average draw ratio and in the typical advantage for White.

• Bayeselo Program Measurements – Anton is basically saying that the derivative engine (e.g., “Dorsz…Kokaburra…”) performed impressively compared to others. The “final test” to which he refers is presumably the set of matches run on these various opening suites.

• Anton acknowledges the value of having multiple testing sets (including unbalanced ones) to cross-verify engine strength.

• He also points out that no single set of data is a perfect reflection of engine skill—rather, they collectively help us draw better conclusions.

Key Reasons Why UHO is Useful

• Greater Imbalance = Faster Distinction: Highly unbalanced openings help engines diverge in their evaluations more quickly, revealing their deeper planning capabilities.

• Less Risk of Artificially Inflated Draws: Purely “strong” or symmetrical openings can artificially inflate draw rates, requiring enormous sample sizes to differentiate closely matched engines.

• Real-World Human Testing: Human players rarely stick solely to the “absolute main lines,” so featuring practical or unusual lines can simulate more varied, real-world chess scenarios.

Summary

While Sedat’s emphasis on “strong openings” and bullet tests caters to a particular style of rating measurement, there are strong counterarguments in favor of unbalanced lines such as UHO:

• They do not represent forced blunders but calculated imbalances that expose engines’ abilities to handle sharp positions.

• They help generate more decisive outcomes without sinking into draws, which can make rating differences clearer in fewer games.

• They reflect a broader range of “real” positions humans might explore, rather than restricting engine tests to narrow, theoretically “perfect” lines.

• Recommended best practices typically involve mixing different test suites—some featuring main lines, others featuring UHO or critical unbalanced positions—to fully measure an engine’s capability across the entire spectrum of chess. This approach avoids overfitting, ensures that we see both how an engine handles best play as well as how it navigates unusual or risky structures.

• While I prefer personally balance and less biased opening test sets, UHO has demonstrated the value for chess programmers for improving their test engine

Regards.

[supernova]

Sorry, I corrected a couple of errors from the original message. This Forum does not let you remove your messages.

Hello, Sedat

I hope you are going and still thanks for your contests you have been running over the years.

In this reply, I will clarify the role and benefits of UHO (Unbalanced Human Openings) and address several points raised by you for additional context.

Understanding UHO (Unbalanced Human Openings)

The UHO concept is simple: select openings that are fully legal, chess‑theory based, yet inherently unbalanced, giving White a tangible (sometimes extreme) advantage. These lines:

• Are entirely “human” in origin—no custom modifications like preventing castling or inventing unnatural configurations.

• Can potentially create more decisive outcomes (fewer draws) in engine testing, because White begins with a clear lead or dynamic potential.

• Serve as a way to highlight performance differences among engines quickly.

• One important misunderstanding is to equate “UHO” with “blundered lines.” In reality, “unbalanced” does not mean “inaccurate.” It simply means that from the very first moves, one side attempts a riskier setup. Such lines are still legitimate opening theory.

Your main arguments

• “Weak openings lead to more decisive results, but those are ‘blunders’ and do not reflect engines’ strength.”

While it is true that extreme blunders or outright mistakes serve little purpose in engine testing (because the outcome becomes trivial), UHO openings are not the same as “inserting deliberate blunders.” They are riskier or unusual but remain valid within standard theory. In engine competitions, these aggressive setups can demonstrate how well an engine handles imbalances—often more revealing than playing symmetrical or overly “perfect” opening lines.

• “Strong openings reduce blunders and thus produce more draws, giving us better data.“

Fewer decisive results do not always guarantee “better” or more reliable data. Indeed, obtaining a 92% draw ratio can flatten out meaningful rating fluctuations, forcing testers to require tens of thousands of games before seeing differences. On the contrary, obtaining a balanced number of decisive games (especially from unbalanced openings) can make rating gaps more statistically significant with fewer games, which is precisely a key reason behind adopting UHO sets.

• “Faster time controls plus strongest openings is always better.”

While bullet or blitz time controls can be entertaining and produce high volumes of games quickly, there is an ongoing debate in computer chess circles about whether such rapid conditions reveal the true strength of an engine. Many would suggest that slower time controls allow deeper calculation, revealing how engines handle complex middlegame or endgame scenarios. The choice between bullet and slower classical time controls depends heavily on the testing goal:

• If the goal is rapid throughput, bullet helps.

• If the aim is maximum depth of search, slower is preferable.

One last point.

• A Subtle Point About “Blunders in the Opening”

Sedat’s statements suggest that any line that is not “theoretically strong” amounts to a blunder. However:

• Chess theory is vast, and many offbeat lines (like UHO) are still far from outright blunders.

• Some lines are simply riskier and lead to sharper positions; that does not necessarily make them “bad testing lines.”

• Ultimately, a test suite that includes both “strong” and “unbalanced” lines can prevent engines from only memorizing “main lines” and drawish continuations.

Clarifying Anton101

• Multiple Opening Suites – The data references Unique, UHO, and UHO+Lichess 4852. These sets vary in average draw ratio and in the typical advantage for White.

• Bayeselo Program Measurements – Anton is basically saying that the derivative engine (e.g., “Dorsz…Kokaburra…”) performed impressively compared to others. The “final test” to which he refers is presumably the set of matches run on these various opening suites.

• Anton acknowledges the value of having multiple testing sets (including unbalanced ones) to cross-verify engine strength.

• He also points out that no single set of data is a perfect reflection of engine skill—rather, they collectively help us draw better conclusions.

Key Reasons Why UHO is Useful

• Greater Imbalance = Faster Distinction: Highly unbalanced openings help engines diverge in their evaluations more quickly, revealing their deeper planning capabilities.

• Less Risk of Artificially Inflated Draws: Purely “strong” or symmetrical openings can artificially inflate draw rates, requiring enormous sample sizes to differentiate closely matched engines.

• Real-World Human Testing: Human players rarely stick solely to the “absolute main lines,” so featuring practical or unusual lines can simulate more varied, real-world chess scenarios.

Summary

While Sedat’s emphasis on “strong openings” and bullet tests caters to a particular style of rating measurement, there are strong counterarguments in favor of unbalanced lines such as UHO:

• They do not represent forced blunders but calculated imbalances that expose engines’ abilities to handle sharp positions.

• They help generate more decisive outcomes without sinking into draws, which can make rating differences clearer in fewer games.

• They reflect a broader range of “real” positions humans might explore, rather than restricting engine tests to narrow, theoretically “perfect” lines.

• Recommended best practices typically involve mixing different test suites—some featuring main lines, others featuring UHO or critical unbalanced positions—to fully measure an engine’s capability across the entire spectrum of chess. This approach avoids overfitting, ensures that we see both how an engine handles best play as well as how it navigates unusual or risky structures.

• While I prefer personally balance and less biased opening test sets, UHO has demonstrated the value for chess programmers for improving their test engine

Regards.