IvanHoe-B57c-SSE2

[Aloha]

http://file027.mylivepage.com/chunk27/3 ... e-B57c.rar

[BTO7]

http://file027.mylivepage.com/chunk27/3 ... e-B57c.rar

This is a great version by Peter Pan and i tuned bishops and knights via STS.

Use settings Knight=298
Bishop pair =35

Regards
BT

[halestorm]

Your modifications increase time to solve following position. Default settings: 2 minutes. Mods: 6 minutes.

r2qrb1k/1p1b2p1/p2ppn1p/8/3NP3/1BN5/PPP3QP/1K3RR1 w - - bm e5

I see no benefit in my fast, incomplete tests for the changes.

[BTO7]

Your modifications increase time to solve following position. Default settings: 2 minutes. Mods: 6 minutes.

r2qrb1k/1p1b2p1/p2ppn1p/8/3NP3/1BN5/PPP3QP/1K3RR1 w - - bm e5

I see no benefit in my fast, incomplete tests for the changes.

The tuning was tested against 200 positions gaining around 4% in over all problem solving in sts. Something doesnt seem right about your post. I also had a nice gain in performance in 4+2 min against houdini confirming the sts results. Just to note in order to gain 4% meaning out of 100 positions the new settings scored 4 more correct results thus was quicker. The only thing that may change is the kind of cpu you have the bishop pair i set at 35 in PP compile and 40 in JRs compile

BT

[halestorm]

My testing mirrors yours. I used 1000 test positions for each.

STS - 10 sec

%age
Ivan57dSSE4.2Arena	        96	96	93	86	91	90	90	80	85	91	898
Ivan57dSSE4.2Arena (mod)	99	89	88	93	92	90	89	83	85	94	902

Speed
Ivan57dSSE4.2Arena	        57	83	81	164	118	134	147	277	207	100	1368
Ivan57dSSE4.2Arena (mod)	58	134	149	105	113	153	145	214	202	95	1368

[BTO7]

Nice test hale. STS is a great thing and it may be that every user has to tune with it to their computer. Sometimes like in the knights 1 or 2 tenths makes a difference. One ivanhoe liked 298 while the other liked 297 ...it can be that sensitive

Regards
BT

[halestorm]

Thank YOU for coming up with settings like that. Must have required a lot of trial and error.