i just tested Firebird 1.0 beta w32 against Rybka 3.0 on a Quadcore Win32 Computer. At Blitz 1 min, 2 min and 5 min FireBird is clearly superior.

Still not sure about the Robbolito stuff.

One thing about robolito, it doesn't hang my pc, Rybka does. So that's good.

RobboLito is for Single-Processor-Systems only, FireBird is better even with Multiprocessorsystems.

It's actually a clone of Rybka tuned for fast time controls. It's not stronger if you have really long time controls. Either way, it's unfair to the Rybka author.

Ah...but how did they get the source code of Rybka and why doesn't he simply demonstrate it with code snippets explanations etc... since the "secret" is already out. 

Also much of the evaluation is the same -- a detailed analysis may show something...

...yet .. they do differ and the Firebird is many ways seems like an improved Rybka! 

...Reverse engineering is not a crime

You say that the author of firebird is a criminal. What evidence do you have? the both engines behave different, you may read the sourcecode of Firebird. What part is from rybka? Why is the endgamepart of Firebird much better, why are the positions evaluated different from both engines, why is firebird better, why is the depth of firebirdcalculations much higher. I see no hints at all that firebird is a clone. Even the authors of rybka dont say that. If you say something like that, then give some evidence.

Firebird is robbolito, or so I heard, just a different version -- and yes, Rybka authors are upset about robbolito being a clone of Rybka, however if they did it right it would be impossible to prove from the source code.

Vas admits he ignores improving Rybka's endgame, so no surprise there.  I thought they evaluated much the same however.  Search depth doesn't matter because you tell it how far to search, it's strength is odd considering it bursts on the scene as #1 from nowhere, but no, nothing can be proven.

Firebird is the Norman Schmidt version of Ippolit, Robbolito, Igorrit, and Ivanhoe. They are almost 50 elo better than Rybka 3 even at long time controls. The claims of the Rybka crowd are ridiculeous at times, as you cann't gain 50 elo by just decompiling an engine. The Ippolit team did a lot of work to gain the stregnth, but the Rybka guys don't want to admit it. Their wiki is ippolit.wikispaces.com

The link got cutoff for the link to the games at long time controls.

http://immortal223.borda.ru/?1-19-0-00000041-000-0-0#006

46 more elo at 40/40 in an early version

Firebird is not robbolito, but its using code from Robbolito.  Where can i find something about the Info that the Rybkadevelopers would say that Robbolito uses illegal code? What i find as wild rumor: that Rybka uses code from Fruit. But this code is ( more or less ) free and might be used from Robbolito. The depth of Rybka and Firebird are very different at the same machine with the same time. If Firebird would be a Rybkaclone just having a depth +7 of rybka at the same calculationtime then Firebird would be 4000+ Elo.

How could a clone be that much quicker.

I did compare the same Position and analysed with Rybka, Firebird 120m sec and zappa 140 sec each on the same maschine

1b3rk1/5ppp/2p2rq1/1p1n4/3P2P1/1BPbBP2/1P1N2QP/R3R1K1 w - - 0 1

This position has "known" best moves and all 3 engines found the correct result but they behave totaly different in the calculation:

Firebird 2920 kN/s and Depth 19

Rybka 100 kN/s and Depth 12

Thats totaly different. If Firebird would be a clone of Rybka then Firebird would be 29,2 times Quicker then Rybka. Then Rybka would never even draw against Firebird.

I see no hint for Rybka beeing a clone,i hear only repetitions of rumors wich are in the best interest of the developers of rybka.

Analysis by FireBird 1.0 beta w32_2:

1. +/=  (0.61): 1.Bxd5 cxd5 2.Nb3 b4 3.cxb4 Bb5 4.Nc5 h5 5.g5 Rb6 6.f4 Qf5 7.Bd2 Rc8 8.b3 Bxf4 9.Rf1 Bxf1 10.Rxf1
2. +/=  (0.41): 1.Ne4 Rxf3 2.Qxf3 Bxe4 3.Qh3 Nf6 4.Bd1 Bd5 5.Bc1 h6 6.Qh4 b4 7.cxb4 Bc7 8.b3 Bd6 9.Bb2
3. =  (0.15): 1.h4 Nxe3 2.Rxe3 Bf4 3.Re7 Bxd2 4.Rd1 Bh6 5.h5 Qg5 6.Re5 Bf5 7.gxf5 Qxh5 8.Qg4 Qxg4+ 9.fxg4 Bf4 10.Re7 Rh6 11.Ra1 Bh2+ 12.Kg2 Bb8 13.Ra2
4. =  (0.09): 1.h3 Nxe3 2.Rxe3 Bf4 3.Re7 Bxd2 4.Qxd2 Rxf3 5.Qg2 Be4 6.Rf1 Rxf1+ 7.Qxf1 h5 8.Qe1 Bd5 9.Bxd5 cxd5 10.Qe2 hxg4 11.hxg4 Rb8 12.Kf2 b4 13.Qe5 Qc2+ 14.Kg3 Qd3+
5. =  (0.08): 1.Nf1 h5 2.h3 hxg4 3.hxg4 Nxe3 4.Rxe3 Bf4 5.Re7 Bd6 6.Rd7 c5 7.Bd5 cxd4 8.cxd4 Rf4 9.Qg3 Qf6 10.Kh1 Bb8
6. =  (0.00): 1.g5 Nf4 2.Qg4 h5 3.Qh4 Rf5 4.Kh1 Qxg5 5.Qxg5 Rxg5 6.Ra8 Rf5 7.Ne4 Be2 8.Nd2 Bd3
7. =  (0.00): 1.Qf2 h5 2.Bxd5 cxd5 3.g5 Rf5 4.Nb3 Be4 5.Nd2 Bd3
8. =  (0.00): 1.Kh1 Nxe3 2.Rxe3 Bf4 3.Ree1 Bxd2 4.Qxd2 Rxf3 5.Bd1 Be4 6.Rxe4 Rf1+ 7.Kg2 Rxd1 8.Qxd1 Qxe4+ 9.Qf3 Qc2+ 10.Qf2 Qe4+ 11.Qf3
9. =  (0.00): 1.Ba2 Nxe3 2.Rxe3 Bf4 3.Re7 Bxd2 4.Qxd2 Rxf3 5.Qg2 Qf6 6.Rae1 h6 7.R7e3 Rxe3 8.Rxe3 Bg6 9.Qf2 Qg5 10.h3 h5 11.Re5 Qc1+ 12.Re1 Qg5 13.Re5
10. =  (-0.01): 1.Rad1 h5 2.h3 Re8 3.Bxd5 cxd5 4.f4 Ra6 5.f5 Qc6 6.Bg5 hxg4 7.hxg4 Ra2 8.Rxe8+ Qxe8 9.Qf3 Be2 10.Re1
11. =  (-0.23): 1.Bd1 Re6 2.Ne4 f5 3.gxf5 Qxf5 4.Bd2 Bc4 5.Be2 Bxe2 6.Rxe2 Rg6 7.Ng3 Qd3 8.Ree1 b4 9.Kh1 Bf4
12. =/+  (-0.62): 1.Rac1 Nxe3 2.Rxe3 Bf4 3.Re5 h5 4.g5 Rd6 5.Rce1 Bxe5 6.dxe5 Re8 7.Ne4 Rdd8 8.Nc5 Bc4 9.Bxc4 bxc4 10.f4 Qf5 11.Qf2
13. =/+  (-0.66): 1.Ra6 Nxe3 2.Rxe3 Bf4 3.Re1 Bxd2 4.Qxd2 Rxf3 5.Qg2 Be4 6.Bc2 f5 7.Bxe4 fxe4 8.d5 Rf2 9.Qxf2 Rxf2 10.Kxf2 Qf7+ 11.Kg1 Qxd5 12.Ra8+ Kf7 13.Rf1+ Ke6 14.Rh8 h6 15.Re1 Kd7 16.Rf8 Qe6 17.h3 Kc7
14. =/+  (-0.69): 1.Ra5 Nxe3 2.Rxe3 Bf4 3.Re7 Bxd2 4.Qxd2 Rxf3 5.Ra1 Qxg4+ 6.Qg2 Qh5 7.Rae1 Rf6 8.R7e3 Bf5 9.Kh1 Rg6 10.Qf3 Qxf3+ 11.Rxf3 Be6 12.Bxe6 Rxe6 13.Rxe6 fxe6 14.Re3 Kf7 15.Kg2 Ke7 16.Re5
15. -/+  (-0.72): 1.Ra2 Nxe3 2.Rxe3 Bf4 3.Re7 Bxd2 4.Qxd2 Rxf3 5.Ra1 Qxg4+ 6.Qg2 Qh5 7.Rd1 Qf5 8.Rde1 g6 9.R7e5 Qf6 10.Bd1 Rf4 11.R5e3 Bc4 12.Bf3 Rc8 13.Qh3 Be6
16. -/+  (-0.73): 1.Ra3 Nxe3 2.Rxe3 Bf4 3.Re1 Bxd2 4.Qxd2 Rxf3 5.Qg2 Be4 6.Bc2 f5 7.Bxe4 fxe4 8.Raa1 Qe6 9.Rf1 b4 10.cxb4 Qd7 11.Rfd1 Qd5 12.g5 g6 13.h3 Qd6 14.Qh2 Qxb4 15.Ra7 Qb3

(,  30.01.2010)

Analysis by Rybka 3 32-bit-4:

1. +/=  (0.64): 1.Bxd5 cxd5[] 2.Nb3 b4 3.cxb4 h5 4.g5 Rb6 5.Nc5 Qf5 6.Bd2 Bb5 7.Kh1 Rc8
2. +/=  (0.50): 1.Ne4 Rxf3 2.Qxf3[] Bxe4 3.Qh3 Nf6 4.Bd1 h6 5.Qh4 Bd5 6.Bc1 Ne4 7.Bc2
3. =  (0.12): 1.h4 Nxe3 2.Rxe3[] Bf4 3.Re7 Bxd2 4.Rd1 Bf4 5.h5 Qg5 6.Rxd3 c5 7.Bc2 Rh6 8.Re2 cxd4 9.cxd4 Ra6 10.Rd1 Be3+ 11.Kh1 Rd8 12.d5 Raa8
4. =  (0.09): 1.h3 h5 2.Nf1 Re6 3.Bd2 hxg4 4.hxg4 Rfe8 5.Ng3 Rxe1+ 6.Rxe1 Rxe1+ 7.Bxe1 Qg5
5. =  (0.09): 1.Nf1 h5 2.h3 Re6 3.Bd2 hxg4 4.hxg4 Rfe8 5.Ng3 Rxe1+ 6.Rxe1 Rxe1+ 7.Bxe1 Qg5
6. =  (0.03): 1.Ba2 h5 2.g5 Re6 3.Ne4 Bxe4 4.fxe4 Qxe4 5.Qxe4 Rxe4 6.Bd2 Rxe1+ 7.Rxe1 Bf4
7. =  (0.03): 1.Kh1 Nxe3 2.Rxe3[] Bf4 3.Re7 Bxd2 4.Qxd2 Rxf3 5.Rae1 Qxg4 6.Rxf7 R3xf7 7.Qxd3 g6 8.Rf1 Qd7 9.Qe3 Kg7
8. =  (0.01): 1.Qf2 h5 2.Bxd5 cxd5 3.g5 Rf5 4.Nf1 Rf6 5.f4 Re6
9. =  (-0.01): 1.g5 Re6 2.Ne4 Rfe8 3.Bd2 Qh5 4.Ra6 Qh4 5.Qf2
10. =  (-0.09): 1.Rad1 h5 2.h3 Re8 3.Bxd5 cxd5 4.f4 hxg4 5.hxg4 Rc6 6.Ra1 Rce6
11. =  (-0.14): 1.Bd1 Re6 2.Ne4 f5 3.gxf5 Qxf5 4.Bd2 Qh5 5.Ng3 Rxe1+ 6.Bxe1 Qh6 7.Qd2
12. =  (-0.23): 1.Ra3 Re6 2.Bxd5 cxd5 3.Raa1[] h5 4.gxh5 Qxh5 5.Bf2 Rg6 6.Bg3 b4
13. =  (-0.23): 1.Ra5 h5 2.Bxd5 cxd5 3.Raa1[] Re6 4.gxh5 Qxh5 5.Bf2 Rg6 6.Bg3 b4
14. =/+  (-0.46): 1.Ra6 Nxe3 2.Rxe3[] Bf4 3.Re1 Bxd2 4.Qxd2 Rxf3 5.Qg2[] Be4 6.Bc2 f5 7.Bxe4 fxe4
15. =/+  (-0.53): 1.Ra2 Nxe3 2.Rxe3[] Bf4 3.Re7 Bxd2 4.Qxd2 Rxf3 5.Ra1 Qxg4+ 6.Qg2 Qh5 7.Rd1 Bc4 8.Bxc4 bxc4 9.Rde1 Rf6
16. =/+  (-0.53): 1.Rac1 Nxe3 2.Rxe3[] Bf4 3.Re5 Bxe5 4.dxe5 Rf4 5.Re1 Re8 6.Qf2 Qg5 7.h4 Qh6
17. -/+  (-0.88): 1.Kf2 Nxe3 2.Rxe3[] Bf4 3.Rae1 Bc2 4.Re4 Bxd2[] 5.Bxc2[] Bxe1+[] 6.Kxe1 Qg5 7.Kf1 g6 8.Bd3 Qd5 9.Qf2 Ra8
18. -/+  (-1.01): 1.Qh3 Nxe3 2.Rxe3[] Bf4 3.Rae1 Bxe3+ 4.Rxe3 Bc2 5.Qf1 Bxb3 6.Nxb3 Re6 7.Rxe6 Qxe6 8.Nc5 Qe3+ 9.Qf2 Qxf2+ 10.Kxf2 Ra8 11.Ke3 Ra2
19. -/+  (-1.23): 1.f4 Nxe3 2.Rxe3[] Bxf4 3.Rh3 Re8 4.Nf3 Qe4 5.g5 Rf5 6.Kh1 Bxg5 7.Rg3

(,  30.01.2010)

Analysis by Zappa Mexico II:

1. +/-  (0.88): 1.Bxd5 cxd5 2.Nb3 Bd6 3.Nc5 Bxc5 4.dxc5 Re8 5.Bf4 Rxe1+ 6.Rxe1 Re6 7.Ra1
2. +/=  (0.42): 1.Ne4 Rxf3 2.Qxf3 Bxe4 3.Qh3 Nf6 4.g5 h6 5.Qh4 Nh7 6.Qg4 Bf5
3. +/=  (0.36): 1.h4 Nxe3 2.Rxe3 Bf4 3.Re7 Bxd2 4.Rd1 Bxc3 5.bxc3 h5 6.gxh5
4. +/=  (0.35): 1.Nf1 h5 2.g5 Re6 3.Bd2 Nf4 4.Bxf4 Rxe1 5.Rxe1 Bxf4 6.h4 b4
5. +/=  (0.32): 1.g5 Nf4 2.Qg4 Rf5 3.h4 Ne2+ 4.Kh1 Ng3+ 5.Kg2 Bc7 6.Kh3 b4 7.Ne4 Nh5
6. =  (0.24): 1.Rad1 Re6 2.Ne4 Bxe4 3.fxe4 Qxe4 4.Bd2 Qg6 5.h3 Bc7 6.Bxd5 cxd5 7.Qxd5
7. =  (0.20): 1.h3 h5 2.Bxd5 cxd5 3.Bf2 Qg5 4.Nb3 hxg4 5.Nc5 Rxf3 6.hxg4 Qf4 7.Nxd3 Rxd3 8.Qxd5 Qxg4+
8. =  (0.18): 1.Qf2 h5 2.Bxd5 cxd5 3.g5 b4 4.cxb4 Bd6 5.Ra5 Rc8
9. =  (0.17): 1.Kh1 Nxe3 2.Rxe3 Bf4 3.Ree1 Bxd2 4.Qxd2 Rxf3 5.Bd1 Rf6 6.Qg2 Qg5 7.Bb3
10. =/+  (-0.30): 1.Ba2 Nxe3 2.Rxe3 Bf4 3.Re7 Bxd2 4.Qxd2 Rxf3 5.Qg2 Rf4 6.h3 Be4
11. =/+  (-0.41): 1.Ra2 Nxe3 2.Rxe3 Bf4 3.Re1 Bxd2 4.Qxd2 Rxf3 5.Qg2 Be4 6.Bd1 Rf4 7.Qg3 Qg5 8.Ra7
12. =/+  (-0.47): 1.Bd1 Re6 2.Ne4 f5 3.gxf5 Qxf5 4.Bb3 Qxf3 5.Ng5 Qxg2+ 6.Kxg2 Rg6 7.Ra6 Re8
13. =/+  (-0.51): 1.Rac1 Nxe3 2.Rxe3 Bf4 3.Ree1 Bxd2 4.Qxd2 Rxf3 5.Bd1 Qxg4+ 6.Qg2 Qxg2+ 7.Kxg2
14. =/+  (-0.67): 1.Ra6 Nxe3 2.Rxe3 Bf4 3.Qf2 h5 4.h3 Bxe3 5.Qxe3 Bf5 6.Kh2 hxg4 7.fxg4 Bc2
15. -/+  (-0.73): 1.Ra3 Nxe3 2.Rxe3 Bf4 3.Re1 Bxd2 4.Qxd2 Rxf3 5.Qg2 Be4 6.Bd1 Re3 7.Qf2 Rxe1+ 8.Qxe1
16. -/+  (-0.78): 1.Ra5 Nxe3 2.Rxe3 Bf4 3.Re1 Bxd2 4.Qxd2 Rxf3 5.Qg2 Be4 6.Bd1 Re3 7.Qf2 Rxe1+ 8.Qxe1
17. -/+  (-1.06): 1.Qh3 Nxe3 2.Rxe3 Bf4 3.Rae1 Bxe3+ 4.Rxe3 Bc2 5.Ba2 Qg5
18. -/+  (-1.18): 1.Kf2 Nxe3 2.Rxe3 Bf4 3.Rxd3 Qxd3 4.Ne4 Rh6 5.h3 b4 6.Rd1 Qb5 7.g5 Rh4 8.c4
19. -/+  (-1.18): 1.c4 Nxe3 2.Rxe3 Bf4 3.Re5 Bxe5 4.dxe5 Rf4 5.cxb5 cxb5 6.Qf2 Qg5
20. -/+  (-1.24): 1.f4 Nxe3 2.Rxe3 Bxf4 3.Rh3 h5 4.gxh5 Qxg2+ 5.Kxg2 Bf5 6.Nf1 Bxh3+ 7.Kxh3 Bd6
21. -+  (-1.54): 1.Qh1 Nxe3 2.Rxe3 Bf4 3.Rae1 Bc2 4.Qg2 Bxe3+ 5.Rxe3 Bxb3 6.Nxb3 Qb1+ 7.Qf1 Qxb2 8.Nc5 b4 9.Ne4 bxc3 10.Nxf6+ gxf6
22. -+  (-1.78): 1.Bh6 gxh6 2.Bxd5 cxd5 3.Nb3 b4 4.Nc5 bxc3 5.bxc3 Bb5 6.Rab1 Rb6 7.Qb2

(,  30.01.2010)

It is not a rumor that Rybka comes from Fruit. The parsing code and the time management are nearly direct in the copying of code. The chess stuff is also a lot of the same, but the numbres are more tuned in the evaluation. The search changes the pruning and the material tables are also added. The base of it looks still like Fruit. The hash tables are about the same though Rybka has changed those in later versions. The other "major change" from Fruit (Letouzey) to Rybka (Rajlich) was to rewrite it for bitboards to make it 3 times faster on 64 bit CPU, but a programmer said that takes maybe 3 hours once you having a working implementation of both Fruit and bitboards around. Rybka just uses Crafty bitboards so making them is easy. Also the node counts and depth were <<obfuscated>> by "Vas" likely to try to coverup this, plus the sales pitch was that Rybka was a <<smart>> engine rather than a fast one. Mr. Letouzey appears unminding because Rybka only copied ideas and not GPL code.

Also Rybka 1.5.32 in April 2004 was almost 1000 elo worse than Rybka 1.0 in Decmeber 2005 and Fruit 2.1 was in between them. The comparsion of Rybka 1.5.32 can show that Rybka 1.0 was just a <<reboot>> based upon Fruit 2.1 to boost the elo. Rybka was an unknown until December 2005 when the 1.0 suddenly appeared with a huge elo injection.

Wish there was a better place to put this but here is my take on the rybka robolito debate.  First there are two eras in chess programming.  Before fruit and after fruit.  At one time fruit was the best program in the world and was open source.  Every engine author had to raid it or be left behind.

Second, vas does not have to reveil his code.  He claims that robolitos authors were in correspondence with him the whole time they were doing this.   Show us the correspondence and maybe I will beleive.  BUT, (and thats a big but) reverse engineering is not illegial.  For years there have been forrums that have been trying to figure out how rybka works by its playing style.  That is also reverse engineering.  All vas is saying is that robo uses ideas similar to rybka.  He can't or won't prove that these ideas are even in rybka. And even if they are, rybka is a five year old program.  Your telling us that no one could come up with similar ideas on how to make a chess program in five years when hundreds of engines come out each year?????  And of course... similar strong ideas would produce a strong engine.  The programmers who wrote robbolito must be very skilled chess programmers not just hackers because even if they did steal the code they made their program stronger than even vas could improve his own engine.  If they could do that, why couldn't they just make the whole engine from scratch (or at least from fruit, or toga or crafty)?  Throw in some of the conjectured ideas on how rybka works.  Keep the ones that work, throw out the ones that dont and BOOM!  It apears to be a reverse engineered rybka.

What vas is doing is very clever on his part.  Like i said, when fruit came out, everyone stole from it (thats ok its open source).  Now rybka comes out and is significantly stronger.  So, vas sells his engine and makes a lot of money as long as its the strongest engine.  WHOOPS! robbolito comes out, stronger than rybka and open source..... If everyone raids robbolito like they did fruit, rybka will just be run of the mill and the cash cow dries up.  SOOOOO vas claims that the code is stolen from rybka so no one will touch it.  His engine remains the strongest "legitamate" engine (supported by chessbase who won't allow it in their rooms so the pros won't use it) and he gets to keep making money hand over fist. 

Sorry that I sound so harsh but the ideas in rybka are not patented.  Its just vases personal secret that he sells and makes available to anyone who wants to look at it.  Dorothy decided to look behind the curtian.  Of course all software companies state that your not allowed to reverse engineer in their EULA, but, an EULA is not legally binding.  In fact many highly regarded programs use reverse engineered software.  Many emulators use reverse engineered bios files so you can leagaly use them.  Open Office uses a reversed engineered version of Microsofts propriatary document format so you can save and open word documents.  Do you think Microsoft likes THAT?

At any rate, robbo code is going to slowly leak into other programs.  That is unless the chess programming police get really vicious (lol).  If professional software does not use it, the amature programs will and the pros will be left in the dust.  What then would be the advantage in buying the pro programs???? The industry would collapse.  SOOO, the code from Robbolito is going to be used period.  This whole discussion is going to be moot very soon.

Just my two cents.

Joel Schlecht

(p.s. sorry for the gramatical and spelling errors, this page erased my text three times and i just wanted to get it finished before it happened again).

Good discussion - sorry if this is a bit off topic. Could someone explain "hash" and why my software asks if I want to clear the hash? Thanks.

The hash is the size of the stored positions in the memory. You don't need to clear it, except if you are with the testing agencies, and want reproduced results. Most of the testings now are following the censoring to make Rybka appear sitting "#1" still. The size can be about 256+ megabytes if you analysis for awhile. The default hashsize has to be small because the UCI inventor said so when the program loads.

The upward advantage of the pro programs is benefit of users supporting you, and more featrues... and you get to complain. See Rybka never outsold Frtiz or Chessmaster even as being "#1" in the ratings. The principal feature of Rybka has strength, so that's why Ippolit had to be censored, for 5 months no one but "Vas" himself could take their ideas to improve. The Ippolit did way more too, first understanding his code from the reversers mess and then improving it to beat Rybka. They maybe achieved 50 elo additional more quick than "Vas" did from a dismal starting position.

Thank you!

I had Fritz 12 play against Firebird. Fritz 12 got crushed!

Is it a bright idea to purchase Rybka 3, since Firebird is free? Does Rybka have any features that Firebird does not have? I only plan to run engines on one cpu, btw. Will registered users of Rybka 3 receive a discount when upgrading to Rybka 4? Thanks.

It is if you're an honest person.

It looks like I was not clear.

First, from what I've read so far, it doesn't seem like the Firebird team did anything illegal. If they absolutely ripped off Rybka, then I would be happy to remove Firebird from my computer and purchase Rybka 3. No big deal. I've got the money and want to support software developers, as I have always done in the past in the field of music. Music software is far more expensive. Several of my music programs cost between $250.00 to $500.00. One product requires a $100.00 upgrade fee each year to stay current. I don't use cracked software. Period.

My question: Does Rybka 3 have anything to offer that's not found in Firebird? Even if it does, is anyone that doesn't yet own Rybka holding out for the release Rybka 4? Why?

I'm just trying to make an informed purchasing decision. I've already spent money on several licensed chess software products, including ChessBase/Everyman ebooks, Fritz 12, ChessBase Light, Chess Tutor and a few other software products.  All licensed, legit products.  

What website can i download firebird from?

At this point I would wait for Rybka 4 as Rybka 3 has a LOT of bugs that were never fixed (another reason Robbolitto and the like might be stronger...the bugs have been fixed!)

Another FREE and REALLY strong engine that has no controversy surrounding it is Stockfish 1.6.2

This engine is great for practicing endgames against.  I love setting up positions out of the Manual of Chess Endings and trying to play them out against Stockfish!  Again it is free of cost and controversy!