Stockfish 4

Now available.   Should be interesting.

Cool! I just tried it on Analyze This app and it is quite faster compared to Critter (my earlier fav!)

In this amazing game played by Kamsky yest, Stockfish is nearly 10-20s faster in finding mate in 9, compared to Critter!

Hi pfren,

There are more x64 builds on the Development page, if you want to try them.

http://abrok.eu/stockfish/

Based on the early results I'm seeing, it looks like Stockfish 4 might be edging out Houdini 1.5a.   (Not that it matters a whole lot.)

Update - I'm almost at 40 games, which isn't a lot, I know, but my SF4 is just smashing H1.5a. I saw another report from some testers showing maybe +12 elo above Houdini, but I'm getting something like +100+. I know this can't hold, it's gotta tighten up. I've rechecked my settings, and everything looks OK. (I'm only using 1 thread per engine; I'm using the Windows 64-bit sse42 compile.)

@EscherehcsE: Maybe Stockfish 4 is relatively strong on machines that don't have "unlimited" resources?

So Houdini 1.5a might need more resources to get close to the level of Stockfish 4, and with limited resources Stockfish 4 may outplay Houdini.

As I suspected it might, my Stockfish 4 vs. Houdini 1.5a tournament has tightened up. At 100 games, Stockfish leads Houdini by +56 elo (58% score). That's close to what the IPON rating test is around right now (except he's testing Stockfish 4 vs. Houdini 3; atm he's showing Stockfish at 14 elo over Houdini 3). If this is close to right, that puts Stockfish 4 close to, if not slightly above, Houdini 3. Houdart had better pull a Houdini 4 out of his hat real fast. (And I'd expect him to.)

Do you happen to have a link to the games or anything like that? And also, will the Stockfish app be updated to Stockfish 4 soon?

it does look like stockfish is having some trouble with komodo cct

I don't see a link to the actual games, but here's one to the ongoing IPON results:  http://www.inwoba.de/

And my games are on my computer.

btw, it looks like the CCRL testing is showing Stockfish not doing quite as well as in the IPON testing. We'll just have to wait and see where everything shakes out in time. But Stockfish is really getting close.

Sorry, I don't know anything about Stockfish apps - I just download and install the engines for PC use.

Oh, your just a tester... You made it seem like you were the creator of it, with all your information and testing...

HaHa, I'm not even an official tester...just idle curiosity. Sorry if I caused you to think I was somebody important.

hi pfren, what GUI are you using? i'm not having any problem with Stockfish 4 x64 SSE42 with Fritz 11 GUI under sandybridge celeron

I used the instructions here to compile it on my Debian Linux system.  It works fine.  I am not familar enough with chess engines to comment on it's performance.

well i made stockfish-4 play  some  bullet games against houdini 1.5a with no book for either engines and ponder on.

i found that houdini is better than stockfish-4 in bullet games. however as soon as i made both of them play longer games.  stockfish-4  clearly outplayed houdini

well later on, i had stockfish 3 with me. in the downloads, they offer 2 types of stockfish engines. one is stockfish-3 32 bit  and another one is stockfish-3 legacy-ja.  

so i made stockfish-3 legacy-ja  play some bullet games with houdini1.5a  . i found that stockfish-3 legacy-ja was better than houdini. houdini was able to draw some games but could never defeat stockfish-3 legacy-ja

i thought that stockfish-3-legacy-ja might be stronger than stockfish-4. so to check this fact, i made stockfish-4 and stockfish-3-legacy-ja play some bullet games. the result stockfish-4 was better than stockfish-3-legacy-ja.

quite surprising and weird results.

houdini beats stockfish-4 .  stockfish-3-legacy-ja beats houdini.  stockfish-4 beats stockfish-3-legacy-ja.   

after some time, i will hold a bullet tournament with all the 3 engines participating it. only then i can get a concrete conclusion.

well i finished the tournament of bullet games . there were 4 rounds and 12 games were played .so here is the table of tournament

engines             houdini1.5a     stockfish-4     stockfish-3-legacy-ja  points

houdini1.5a        ~~~~~~~~        1-1-0-1                   1-0-0-0               4/8

stockfish-3         0-1-1-1           0-0-0-0                  ~~~~~~~               3/8 

stockfish-4         0-0-1-0          ~~~~~~~                  1-1-1-1                5/8

 so i can say that stockfish-4 is better than stockfish-3 . no doubt about it.

however, it is still not better than houdini in bullet games

also , stockfish-3 is quite good at playing at bullet games than houdini1.5

but as soon as i make them play games of 30 min.  stockfish-4 clearly beats stockfish-3 and houdini1.5a

so, i can say that in bullet , houdini is better than stockfish-4  but not better than stockfish-3-legacy-ja

but in standard games, stockfish-4 rocks!!

(PS: i know that in order to judge an engine, i should made them play more than 100 games. however i don't have time to make them play 100 games )

I ran a bunch at 4 + 2 between Stockfish 4 and Stockfish 3. Stockfish 3 +10 =25 -1 over Stockfish 4. Ran these in the Fritz 11 GUI. Key difference is that Stockfish is using all eight cores, while S 4 is using only one.

Don't know how to fix this.

Edit:

The answer was on the Stockfish help page. Stockfish 3 automatically detected the number of CPUs. With S 4 this must be set manually. I'm now running more games.

How much stronger can these "computers" get before we have skynet, and they take control of the planet's computing infrastructure in order to solve chess once and for all? And then it's just one small step further to start using people as batteries to get even more power, in order to compute a few ply deeper.

Think of the children, people.

My three commercial engines are single processor, so it's no surprise to find them at the bottom. But, I'm surprised at the order of the top three, all of which were set to use all eight processors. Time control was 4 2.

You need to play lots more games for the results to mean anything.