Moderators: Watchman, Mark Uniacke, mrudolf
Hiarcs 13 does not exist yet - and fixed depth v Rybka is a waste of time as Rybka does not display its true depth - you probably have to add 3 so when rybka says depth 1 it is searching closer to 4. This has been well doccumented on the Rybka forum even Larry K accepts it.yanquis1972 wrote:vas probably hasnt miniaturized rybka for the same reason it's the least customizable, most platform specific (windows PC) chess engine out there right now (save fritz i guess). he's clearly very driven & he has one priority, which is engine strength. thats his business & he's very, very good at it. mark has more diverse priorities which is why he's taken the time to code for Mac (something vas has stated he'll probably never do), pocket pc, iphone, etc. my guess is that when rybka 4 is tidied up, we'll see an iphone version. you could be right that it's too difficult to pull off, ive no idea -- i just think its more likely that vas hasnt even really tried yet.
as for the idea that hiarcs is better on weak hardware, i see no reason to think that; if anythign the opposite is true. its easy to simulate this on any PC if you know approximately what depths each program woudld reach on average on pocket hardware. eg i am running a match right now at fixed depth -- HIARCS 12.1 depth 7, rybka 3 depth 6 (chose those randomly as i have no idea about approx. depths).
rybka so far is +44 =22 -9 +176 elo.
btw, is it true that hiarcs iphone still runs 12.1 instead of 13?? this is really disappointing if so. i have a feeling 13 is going to be a substantial upgrade.
You can add depth 1 as much as you like but why don't you read where it is has been proved that this wont help. Members of the Rybka team accept that Rybka depth displayed is wrong.yanquis1972 wrote:well thats why i staggered the depths a little bit -- right now im doing a d10 v d9 match, 1 CPU & they seem to be using roughly = time (v hard to say w/ any certainty) & the results are similar thus far. in rare positions at this depth ive noticed hiarcs take 3-4 seconds to make a move on my machine & similar w/ rybka, though the vast majority of moves come within 1 second. i think adding 3 would give hiarcs an unrealistic advantage as 1 seems about right at low depth anyway.
im under the impression PF4 has the hiarcs 13 beta (i think ive seen it referred to as 'hiarcs x'), couldnt iphone users get the royal treatment as well?
I presume you are using 32bit Rybka for your test?yanquis1972 wrote:you seem to be misinterpreting the purposes of my test, which is to roughly simulate low hardware conditions, not to give hiarcs = fighting chances. im doing a 0+1 match now, but i expect rybka to dominate even further. i suppose i should use 32 bit rybka however.
im not sure what your point is re glaurung. obviously it was ported successfully, & while it may be much weaker than hiarcs at the moment, stockfish 1.4 is much stronger than glaurung. its also a highly tactical fast-search engine that probably doesnt fare quite as well at very short time controls/slow hardware as hiarcs or rybka which have much more advanced searches.
I don't think it would be particularly interesting with the current iPod version of Glaurung. It will be a little more interesting with the upcoming Stockfish-based version, which is a lot stronger (in a match between the two on my first generation iPod Touch, Stockfish won by 77-23, with 66 wins, 28 draws and only 8 losses). I still expect Hiarcs to be clearly stronger, though. Stockfish 1.4 seems to be a little stronger than Hiarcs 12.1 on fast 64-bit desktop computers, but on the ARM CPU it is crippled by its bitboard infrastructure, and runs extremely slowly.
The test is worthless for any purpose, with the possible exception of entertaining yourself (which is, of course, a very valid purpose). The only way to simulate slow hardware conditions is to benchmark the programs on the slow hardware you are interested in, finding out precisely how much the programs slow down (it will differ widely depending on the program), and then run a fast time handicap match on a faster machine.yanquis1972 wrote:you seem to be misinterpreting the purposes of my test, which is to roughly simulate low hardware conditions, not to give hiarcs = fighting chances. im doing a 0+1 match now, but i expect rybka to dominate even further. i suppose i should use 32 bit rybka however.
It wasn't really "ported" at all, it was just a recompile.im not sure what your point is re glaurung. obviously it was ported successfully,
It is, and the current development version of Stockfish is stronger still. It's significantly stronger than Hiarcs 12.1 on my Mac. However, you are missing the point: Stockfish slows down far more than Hiarcs when running on the ARM CPU found in the iPhone. This would be the case for Rybka as well. The reason for this is the low-level implementation details of the programs: Stockfish and Rybka both make extensive use of 64-bit arithmetic and big lookup tables. This works well on powerful 64-bit computers with huge amounts of cache memory (like the Core 2 Duo), but very badly on the ARM. Matching Stockfish against Hiarcs on the iPhone is roughly equivalent to matching them on a desktop computer with a 3 to 1 time handicap in Hiarcs' favor.& while it may be much weaker than hiarcs at the moment, stockfish 1.4 is much stronger than glaurung.
The tactical speed of Stockfish compared to other modern chess engines is fairly close to the average, I think. Hiarcs, on the other hand, is the tactically fastest chess engine I know.its also a highly tactical fast-search engine
Stockfish does just as well against Hiarcs (and most other engines) at fast time controls as at slow time controls. At some types of slow hardware, on the other hand, it will have a serious disadvantage against Hiarcs, for the reasons described above. For the same reasons, Rybka will have a similar disadvantage against Hiarcs on these types of hardware.that probably doesnt fare quite as well at very short time controls/slow hardware as hiarcs or rybka which have much more advanced searches.
