Patzer AI 020223

[Sedat Canbaz]

U r welcome.

Though i would appreciate if you could give me feedback on how it is (performance) one on one i know patzer is crazy. Once you hv tested it on ur free time.

I test my engines on their default settings + 2min+2s / +1s (on the final test) + tests suites i like DOTS6 and 5 and Nooman 24 n mine. these have interesting positions where i hv seen engines struggling to win and my engines (officially released ones) perform better, other many test suites give alot of draws because their not imbalanced, these i use for book making. am aware that other testers like to add their own settings, large pages etc its fyn.

Well, nowadays I have no much free time for eng testings
But, as I stated earlier,in September I hope to test it...

Regarding opening test suites, (if 3700+ Bullet, fast Blitz..)
According to my experience, be careful with weak ones!
Yes..they will bring more fun such as 50%-75% draw ratio!

But then note also that e.g just in case of weak ones ,
You have to run at least 2500-3000 games (per player)
Otherwise, the ERROR margin can be too high!
And please don't forget to use various weak suites!
You may know..more opening positions better !!

Btw, in case of strong opening suites,
The draw ratio will be higher for sure, but
Then about 500-1000 games is enough...!

In other words:
It all depends on tester, on conditions etc: such as:
More fun..more games required, less fun..less games..!)

Btw, for more details over these issues:
https://sites.google.com/site/computers ... testings-1
https://sites.google.com/site/computers ... testings-2

Greetings )

I agree with you. And that UHO 2022 i have it as well, i just forgot where i got it, but am pretty sure its from your website, bcz i follow some tournaments there. And i hv also downloaded some test suites there though perfect 2022 gives me errors idk why. 2019 seems good.

There are newer book, suite releases on my site...

Perfect 2023 book:
https://sites.google.com/site/computers ... 2023-books

Balsa v270423 suite:
https://sites.google.com/site/computers ... ite-270423

Unique v170623 suite:
https://sites.google.com/site/computers ... test-suite

Anyhow, if 3700+ Elo tours...sure for more fun,
Then I suggest simply to be used Unique openings
Especially Unique v50 should produce more wins

But in case of serious tours,
Balsa, Perfect books seems to be not so bad choice
But WARNING: if 3700+, may appear many draws !)

Sure there are many other openings out there..
So..you can try to test them as well..

Greetings )

[OrgZ]

Thanks, but i have already downloaded these from your site i think last week if am nt mistaken. Ill try unique haven't gotten a chance to try that one

[Sedat Canbaz]

perfect 2022 gives me errors.

Can you inform Me please,
What kind of error ?

I wonder... because,
I did not release Perfect 2022 books!
I mean not available at least online )

Best,
Sedat

[OrgZ]

on Arena Gui when I tried to load it, it did not load, so I deleted it. am left with perfect 2019 and 2021 these ones load

[OrgZ]

Score of Patzer AI 220823_bmi2 vs TACTICAL 220823_b: 12 - 6 - 22 [0.575]
... Patzer AI 220823_bmi2 playing White: 8 - 0 - 12 [0.700] 20
... Patzer AI 220823_bmi2 playing Black: 4 - 6 - 10 [0.450] 20
... White vs Black: 14 - 4 - 22 [0.625] 40
Elo difference: 52.5 +/- 73.0, LOS: 92.1 %, DrawRatio: 55.0 %
40 of 40 games finished.

[Sedat Canbaz]

on Arena Gui when I tried to load it, it did not load, so I deleted it. am left with perfect 2019 and 2021 these ones load

I see, but you can NOT be right, because
Perfect 2022 opening book is not released!

And very likely,
You confused the book versions...)

Greetings

[Sedat Canbaz]

Score of Patzer AI 220823_bmi2 vs TACTICAL 220823_b: 12 - 6 - 22 [0.575]
... Patzer AI 220823_bmi2 playing White: 8 - 0 - 12 [0.700] 20
... Patzer AI 220823_bmi2 playing Black: 4 - 6 - 10 [0.450] 20
... White vs Black: 14 - 4 - 22 [0.625] 40
Elo difference: 52.5 +/- 73.0, LOS: 92.1 %, DrawRatio: 55.0 %
40 of 40 games finished.

Thanks a lot for the current test...

Very good.. such as + 52 Elo for Patzer AI, but just 40 games...
So can you please run it at least 2000 games (per player)
Why 2000 games? 1st + 52 Elo can not be true !!
2nd current draw ratio: 55% so it seems you use weak suite
And against less Error Elo margin...you need thousands of games
And after test... can you share your played games as well ?

Btw,
Can you inform us please about your used conditions ?

Greetings

[Sedat Canbaz]

Meanwhile, to be more clear...
I wonder about your test conditions:
- Hardware / Processor ?
- Hyper-Threading ON/OFF ?
- Large-Pages ON/OFF ?
- MP or 1 Core or Thread ?
- Opening Book/Suite ?
- Ponder ON/OFF ?
- Time Control ?
- Hash-Table Size ?
- Endgames ?
- GUI info not needed (very likely Cutechess you use )

Thanks in advance )

[OrgZ]

on Arena Gui when I tried to load it, it did not load, so I deleted it. am left with perfect 2019 and 2021 these ones load

I see, but you can NOT be right, because
Perfect 2022 opening book is not released!

And very likely,
You confused the book versions...)

Greetings

you are correct there, as am unsure which perfect book version was it. but the ones that i have now are good.

boths engines were running on their default settings, exp OFF (& starting from scratch). 2min+ 1s (DOTS5)
AMD Athlon Gold 3150U with Radeon Graphics, 2400 Mhz, 2 Core(s), 4 Logical Processor(s)

[Sedat Canbaz]

on Arena Gui when I tried to load it, it did not load, so I deleted it. am left with perfect 2019 and 2021 these ones load

I see, but you can NOT be right, because
Perfect 2022 opening book is not released!

And very likely,
You confused the book versions...)

Greetings

you are correct there, as am unsure which perfect book version was it. but the ones that i have now are good.

boths engines were running on their default settings, exp OFF (& starting from scratch). 2min+ 1s (DOTS5)
AMD Athlon Gold 3150U with Radeon Graphics, 2400 Mhz, 2 Core(s), 4 Logical Processor(s)

Thanks for your replay...

But however,
Mostly of my conditions questions are missing (in your answer)..why?)
Instead of giving so much processor details, you'd try to share other..)
But not so BIG problem.. )

By the way...very soon,
I will run the same engines, but under SCCT cond.
Sure I will use stronger openings than yours..so
It is not required +2000 games...very likely about
400 games (per player) will be enough...if nothing
Else..we will have more data for comparing...)
You know..more better, right ?)

And I wonder much what will be the results..
Let's see 50 Elo difference will appear or not..

Good luck to both engines authors )