If we look at the rather restrictive (in terms of excluding dozens of Ipponians) IPON list, we find Junior 13.3 on rank 17. To become top-10, the new version needs to have gained 78 elo points. Good luck.
Seems it gained huge compared to any known version. The opponents were completely surprised by this, getting in positions where they thought they had a winning advantage, and then losing. Consensus amongst the participants was that they must have invented some new search improvement.
hgm wrote:Seems it gained huge compared to any known version. The opponents were completely surprised by this, getting in positions where they thought they had a winning advantage, and then losing. Consensus amongst the participants was that they must have invented some new search improvement.
No doubt that Junior was the best program in this tournament, but what else can we conclude from playing 10 games?
Yes with 10 games maybe you can't conclude too much, but the "sudden death" nature of the small amount of games can make it exciting.
Of note also, that the hardware of Jonny and Junior probably did not make much difference. We know that scaling beyond 16 cores is not that good, and chances are they are slower cores too beyond that point. But the computer geek side of me loves to see engines running on such hardware.
hgm wrote:Seems it gained huge compared to any known version. The opponents were completely surprised by this, getting in positions where they thought they had a winning advantage, and then losing. Consensus amongst the participants was that they must have invented some new search improvement.
No doubt that Junior was the best program in this tournament, but what else can we conclude from playing 10 games?
So, with errors 2SD of 100 points (or error of difference 140 points), Junior seems superior to Hiarcs, Pandix and Shredder, even with this small sample.
fern wrote:BUT I understand that playing in equal hardware, Hiarcs was the winner. Wasn't so?
Fern
What was the equal hardware ? On very low-end 32-bit hardware as was the case in the past, Hiarcs's performance was not degraded as much as some other engines. If the equal hardware had been an 8-core 64-bit box, or even a quad 64-bit, Hiarcs would not have stood a chance. Plenty of equal hardware ratings lists out there with hundreds of games (IPON, CEGT. CCRL) that show Hiarcs as barely a top-10 engine. But I don't know what the hardware was this year. Maybe someone can clarify, and that isn't the case this time.
SzG wrote:
All programs play on the following hardware: laptop with I7-3740, 2.7 GHz and 16 GB RAM.
With a 32-bit or 64-bit operating system ? We know that Hiarcs does not benefit from 64-bit, but some of it's competitors gain 50% + speedup from 64-bit.