Page 4 of 13
Re: TCEC S15, END of an ERA event is much more Brutal than I thought!
Posted: Fri May 17, 2019 4:06 am
by Leo
If SF loses can I use the excuse that it only was a 100 game match and that if it was thousands of games SF might have won?
Re: TCEC S15, END of an ERA event is much more Brutal than I thought!
Posted: Fri May 17, 2019 4:41 am
by jp
Leo wrote: ↑Fri May 17, 2019 4:06 am
If SF loses can I use the excuse that it only was a 100 game match and that if it was thousands of games SF might have won?
If the margin is 4, yes. If the margin is 25, not really.
But opening choices and hardware are the bigger issues that will never be agreed on.
Re: TCEC S15, END of an ERA event is much more Brutal than I thought!
Posted: Fri May 17, 2019 4:53 am
by Leo
I really want to see a match with SF allowed to use Brianfish.
Re: TCEC S15, END of an ERA event is much more Brutal than I thought!
Posted: Fri May 17, 2019 6:46 am
by MikeB
jp wrote: ↑Fri May 17, 2019 4:41 am
Leo wrote: ↑Fri May 17, 2019 4:06 am
If SF loses can I use the excuse that it only was a 100 game match and that if it was thousands of games SF might have won?
If the margin is 4, yes. If the margin is 25, not really.
But opening choices and hardware are the bigger issues that will never be agreed on.
If the margin is only 9 games, for most people that would be pretty convincing.
But if you run enough long matches ( say 50 100 game matches a day ) , you will get a false positive on average once a day.
Say the current trend continues and the final score looks like this.
Code: Select all
Rank Name Rating Δ + - # Σ Σ% W L D W% =% OppR
---------------------------------------------------------------------------------------------------------
1 LCZero v0.21.1-nT40.T8.610 3115 0.0 28 28 100 54.5 54.5 15 6 79 15.0 79.0 3085
2 Stockfish 19050918 3085 30.6 28 28 100 45.5 45.5 6 15 79 6.0 79.0 3115
---------------------------------------------------------------------------------------------------------
Δ = delta from the next higher rated opponent
# = number of games played
Σ = total score, 1 point for win, 1/2 point for draw
ResultSet-EloRating>los
LC St
LCZero v0.21.1-nT40.T8.610 98
Stockfish 19050918 1
ResultSet-EloRating>
The difficult part is that how does one say the hardware is equal - and one cannot. Essentially ,we're comparing two different breeds of cats and they both have a lot of plusses for what they can do. The good news is that regardless of how one feels. it brings people back into computer chess and to me, that's a good thing.
Re: TCEC S15, END of an ERA event is much more Brutal than I thought!
Posted: Fri May 17, 2019 7:45 am
by glennsamuel32
MikeB wrote: ↑Fri May 17, 2019 6:46 am
The difficult part is that how does one say the hardware is equal - and one cannot. Essentially ,we're comparing two different breeds of cats and they both have a lot of plusses for what they can do.
Thanks, that's an excellent explanation...
But people have found ways to do it, using formulas based on node ratios, cost comparisons, energy consumption and what not.
Next may come the comparison of mass
Re: TCEC S15, END of an ERA event is much more Brutal than I thought!
Posted: Fri May 17, 2019 8:05 am
by glennsamuel32
Here's another comparison...
Leela used 32.2 billion transistors while SF used 14.4 billion
Re: TCEC S15, END of an ERA event is much more Brutal than I thought!
Posted: Fri May 17, 2019 1:16 pm
by giovanni
glennsamuel32 wrote: ↑Fri May 17, 2019 7:45 am
MikeB wrote: ↑Fri May 17, 2019 6:46 am
The difficult part is that how does one say the hardware is equal - and one cannot. Essentially ,we're comparing two different breeds of cats and they both have a lot of plusses for what they can do.
Thanks, that's an excellent explanation...
But people have found ways to do it, using formulas based on node ratios, cost comparisons, energy consumption and what not.
Next may come the comparison of mass
I understand that it is quite naive, but couldn't we simply compare how much they cost?
Re: TCEC S15, END of an ERA event is much more Brutal than I thought!
Posted: Fri May 17, 2019 1:43 pm
by Werewolf
giovanni wrote: ↑Fri May 17, 2019 1:16 pm
glennsamuel32 wrote: ↑Fri May 17, 2019 7:45 am
MikeB wrote: ↑Fri May 17, 2019 6:46 am
The difficult part is that how does one say the hardware is equal - and one cannot. Essentially ,we're comparing two different breeds of cats and they both have a lot of plusses for what they can do.
Thanks, that's an excellent explanation...
But people have found ways to do it, using formulas based on node ratios, cost comparisons, energy consumption and what not.
Next may come the comparison of mass
I understand that it is quite naive, but couldn't we simply compare how much they cost?
There are lots of comparisons:
- cost
- power consumption (i.e. limit both to 1 KW or something)
- or just what you can build
Personally I favour the latter. If it can be built, build it.
Re: TCEC S15, END of an ERA event is much more Brutal than I thought!
Posted: Fri May 17, 2019 5:07 pm
by Leo
I will use AB engines for a long time. I cant afford and don't understand how to get LZ working on a computer. SF with a great book is very close if not better than LZ anyway.
Re: TCEC S15, END of an ERA event is much more Brutal than I thought!
Posted: Fri May 17, 2019 8:20 pm
by leavenfish
Leo wrote: ↑Fri May 17, 2019 5:07 pm
I will use AB engines for a long time. I cant afford and don't understand how to get LZ working on a computer. SF with a great book is very close if not better than LZ anyway.
True, unless one gets their jollies off of thinking 'they' are doing something in actual play (why I abandoned 'correspondence' chess) when it is really their engines, it does not matter - fast is fast and good is good enough. It is said that those people are 'cheating themselves'...I won't weigh in on that - but I will say that chess is a fascinating game in itself, which I enjoy using my own brain in combat vs an opponent; but I do like trying to prepare opening lines to give better arm myself for that combat.