New engine releases & news 2021

[niel5946]

After a lot of work and testing, Loki 3.0 is now released. Thanks to Marcel (who pointed out how underrated Loki was, its feature set taken into account), I have been working tirelessly to rework the search and evaluation function. This has paid off greatly!

I have tested the new version to have a strength of around 2510 (>600 elo points stronger than v1.2.0). This was done in three separate gauntlets against MadChess 2.2 and Raven 1.10.

The release: https://github.com/BimmerBass/Loki/releases/tag/v3.0.0

[Gabor Szots]

After a lot of work and testing, Loki 3.0 is now released. Thanks to Marcel (who pointed out how underrated Loki was, its feature set taken into account), I have been working tirelessly to rework the search and evaluation function. This has paid off greatly!

I have tested the new version to have a strength of around 2510 (>600 elo points stronger than v1.2.0). This was done in three separate gauntlets against MadChess 2.2 and Raven 1.10.

The release: https://github.com/BimmerBass/Loki/releases/tag/v3.0.0

Splendid news, Niels.

Unfortunately, even my Loki 2.0.0 results could not appear on our list yet so 3.0 will have to wait a bit I'm afraid.

[emadsen]

After a lot of work and testing, Loki 3.0 is now released. Thanks to Marcel (who pointed out how underrated Loki was, its feature set taken into account), I have been working tirelessly to rework the search and evaluation function. This has paid off greatly!

I have tested the new version to have a strength of around 2510 (>600 elo points stronger than v1.2.0). This was done in three separate gauntlets against MadChess 2.2 and Raven 1.10.

The release: https://github.com/BimmerBass/Loki/releases/tag/v3.0.0

Sounds like great progress, Niels. However, I saw in your recent commit you tested Loki versus MadChess 2.2 with a time control of 5s + 0.1s. I don't test MadChess at lightning time control. The fastest games I use are 2m+1s, so MadChess' rating at 5s + 0.1s is unreliable. My guess is MadChess performs much worse at 5s + 0.1s than 2m + 1s. I don't know why. I've never investigated it. Perhaps due to the disadvantage of a managed memory runtime (C#) versus a native runtime (C++). Perhaps due to poor implementation on my part that manifests itself only in very fast games.

Anyhow, if you're basing your ELO estimate of Loki off a gauntlet versus MadChess at 5s + 0.1s, you're likely inflating the rating of Loki. You can answer the question definitively by running games at 2m+1s, which is what CCRL uses for blitz ratings.

[niel5946]

Sounds like great progress, Niels. However, I saw in your recent commit you tested Loki versus MadChess 2.2 with a time control of 5s + 0.1s. I don't test MadChess at lightning time control. The fastest games I use are 2m+1s, so MadChess' rating at 5s + 0.1s is unreliable. My guess is MadChess performs much worse at 5s + 0.1s than 2m + 1s. I don't know why. I've never investigated it. Perhaps due to the disadvantage of a managed memory runtime (C#) versus a native runtime (C++). Perhaps due to poor implementation on my part that manifests itself only in very fast games.

Anyhow, if you're basing your ELO estimate of Loki off a gauntlet versus MadChess at 5s + 0.1s, you're likely inflating the rating of Loki. You can answer the question definitively by running games at 2m+1s, which is what CCRL uses for blitz ratings.

Thank you for your response! I don't know if this is worth anything, but when I ran the gauntlets, I anchored the ratings to Raven's 2min+1sec CCRL rating, and the rating of MadChess 2.2 (in all three tests) was nearly identical to the one on CCRL Blitz, so I assumed the results would at least be partly valid. I also tested Loki 3.0 against some weaker engines (because my self-play against v2.0 and v1.2.0 showed lower elo changes), and they showed an elo in the same range.
Therefore I think the rating results of Loki 3.0 are, as said, at least within a reasonable range of its real rating. But unfortunately, due to school and only owning one computer, it is rather hard for me to conduct longer tc tests.

[emadsen]

Therefore I think the rating results of Loki 3.0 are, as said, at least within a reasonable range of its real rating. But unfortunately, due to school and only owning one computer, it is rather hard for me to conduct longer tc tests.

OK. I wanted to warn you were wandering into uncharted territory using a very fast time control with my engine. I understand testing is difficult with limited computing resources. We'll see what CCRL finds.

[jshriver]

After a lot of work and testing, Loki 3.0 is now released.

The engine looks wonderful, just played a game and enjoyed its style.

I made linux binaries for 64bit armv8 and 32bit armv7 (pi4 and pi0)
https://chess.nyc3.cdn.digitaloceanspac ... i3-arm.zip

[WinPooh]

GreKo PSQ 1.1 - experimental version with piece/square only evaluation:
http://greko.su/greko-psq-1.1.zip

[Damir]

Thanks Vladimir!

[Gabor Szots]

Unfortunately, even my Loki 2.0.0 results could not appear on our list yet so 3.0 will have to wait a bit I'm afraid.

Well, I've made an exception, +500 Elo was to tempting, so I started a gauntlet with 3.0.0.

[niel5946]

Unfortunately, even my Loki 2.0.0 results could not appear on our list yet so 3.0 will have to wait a bit I'm afraid.

Well, I've made an exception, +500 Elo was to tempting, so I started a gauntlet with 3.0.0.

I am happy to hear that! I hope the results on CCRL will be as good as the ones from my own tests!