SPCC: Testruns of Velvet 8.1.0 and Patricia dev 241106 finished

[pohl4711]

My UHO-Top15 Ratinglist is the world's first engine-ratinglist, using UHO-openings, and the world's first ratinglist offering additionally Gamepair-statistics.

Testruns of Vevet 8.1.0 normal, Velvet 8.1.0 risky and non-official Patricia dev from 241106 finished. All results (including EAS-scores) only in my full Ratinglist, because the engines are too weak for entering my UHO-Top15 Ratinglist.
https://www.sp-cc.de/files/uho_full_list.txt

https://www.sp-cc.de

Also take a look at the EAS-Ratinglist, the world's first engine-ratinglist not measuring strength of engines but engines's style of play:
https://www.sp-cc.de/eas-ratinglist.htm

(Perhaps you have to clear your browsercache with <strg>+<shift>+<delete> to reload the graphics/diagrams on my website)

[nmcrazyim5]

Hi Stefan, do you mind adding top sac games for engines that don't make it to top UHO-Top-15 after your test runs? I find looking at those games under `view sac games` section quite enjoyable and often engines ranked lower down the rank have more interesting games (and velvet certainly does). Thanks!

[pohl4711]

I did this. Because 3 testruns were added at the same time (Velvet 8.1.0 normal, Velvet 8.1.0 risky and Patricia 241106 dev), I decided to provide games by Patricia. I just forgot to update the news/update-line on the "View games with sacs"-site, so in this line, Integral 5 was still listed. Fixed.

If you like these games (like I do), I recommend downloading my Interesting Wins Search Tool. I use my tool by myself to find these sac-games quickly in my huge gamebases. The IWS-Tool is very fast (because it does no statistics/calculations, like my EAS-tool does) and very easy to use. It only fails, when a game has an irregular ending (timeloss, disconnect, illegal move), but luckily this happens not often in enginegames.
https://www.sp-cc.de/files/interesting_ ... ch_tool.7z

You can download the games of all of my testruns from my GoogleDrive:
https://drive.google.com/drive/folders/ ... yjfrdepJOv

Right now, all games of my latest testruns can be found in the UHO_Ratinglist_Archive_005.zip file. I will switch to a new file, when the file reaches 2GB before zipping. The next file will be UHO_Ratinglist_Archive_006.zip ...
So, in the UHO_Ratinglist_Archive_005.zip you find all played games by Velvet 8.1.0 normal and Velvet 8.1.0 risky (and Patricia 241106 dev).

[nmcrazyim5]

Thank you, I realized that once I saw the full list had 3 updates. I already use your win search tool to find interesting games, even human games in megabase. Much appreciate your thoughts and efforts into creating them.

[pohl4711]

Thank you, I realized that once I saw the full list had 3 updates. I already use your win search tool to find interesting games, even human games in megabase. Much appreciate your thoughts and efforts into creating them.

My pleasure. But please mention, my brutal fast sac-search algorithm can fail in human games, when a human player blunders or loses on time... So, when using my tools on human games, you have to check the found sac-games manually, to avoid such failures. (It also fails in enginegames, when the game ending is not regular (= timeloss, disconnect, illegal move) - luckily this is extremly rare in modern enginegames)

[nmcrazyim5]

My pleasure. But please mention, my brutal fast sac-search algorithm can fail in human games, when a human player blunders or loses on time... So, when using my tools on human games, you have to check the found sac-games manually, to avoid such failures.

That is an excellent point and I have come across a few of those due to the larger mistakes in terms of centipawn loss. However, this isn't that much of a concern because (I) I'm looking at games of 2600+ players at mostly classical time control, and (ii) as those are human games to review for myself or my students, I would be looking at them anyway

[chesskobra]

I would like to ask Stefan if the sac search tool does any additional computing besides running pgn-extract for each filter. Since I don't have windows, I cannot use the sac search tool, but I hacked a little bash script to use Stefan's filters with pgn-extract. I also wrote my own filters to search for only exchange sacrifices. I was thinking of posting my script (with my exchange sac filters) in case some linux users are interested. As nmcrazyim5, I use the filters on 40h database, which has mostly high quality human games of highly rated players. In 40h database, I found nearly 7500 games with an exchange sac by white with white win and nearly 6600 games with an exchange sac by black with black win.

[pohl4711]

I would like to ask Stefan if the sac search tool does any additional computing besides running pgn-extract for each filter. Since I don't have windows, I cannot use the sac search tool, but I hacked a little bash script to use Stefan's filters with pgn-extract. I also wrote my own filters to search for only exchange sacrifices. I was thinking of posting my script (with my exchange sac filters) in case some linux users are interested. As nmcrazyim5, I use the filters on 40h database, which has mostly high quality human games of highly rated players. In 40h database, I found nearly 7500 games with an exchange sac by white with white win and nearly 6600 games with an exchange sac by black with black win.

No, all is done by pgn-extract. This is the reason, why the sac-search is so fast: No evaluation needed and no engine-evals in the comments of the games are needed.
The main idea of my sac-search is very simple: Engines do not blunder and do not lose on time. So, if the winning color of an enginegame had less material for 8 consecutive plies (8 plies works best in my tests, to avoid false sac-detects of simple recaptures with some checks between capture and re-capture) anywhere in the game, this must have been a sacrifice. This works fine in 99% of the time and can be done very fast.

[chesskobra]

I know how the pgn-extract filters work. But I use 6 plies threshold for exchange sacs to get a slightly broader set of games.