| Deepmind drops the bomb. |
h.rode...@gmail.com |
12/5/17 9:14 PM |
Looks like the fun might be over. Next tcec might be a wipeout. |
|
unk...@googlegroups.com |
12/5/17 9:06 PM |
<This message has been deleted.> |
|
unk...@googlegroups.com |
12/5/17 9:30 PM |
<This message has been deleted.> |
| Re: Deepmind drops the bomb. |
mal |
12/5/17 9:36 PM |
Can they port it to multi-core Intel h/w though?
:-)
|
| Re: Deepmind drops the bomb. |
mal |
12/5/17 9:38 PM |
"AlphaZero searches just 80 thousand positions per second in chess ... compared to 70 million for Stockfish"
|
| Re: Deepmind drops the bomb. |
rohitr...@gmail.com |
12/5/17 9:42 PM |
On Wednesday, 6 December 2017 10:36:03 UTC+5:30, Leebot wrote:
> Alpha0 now plays chess.
>
>
> https://arxiv.org/abs/1712.01815
link to paper
https://arxiv.org/pdf/1712.01815.pdf
alphazero scored +28 -0 =72 (+100 elo) against stockfish. Most likely they used stockfish 8. Alphazero played on 4 TPU and sf played on 64 thread with 1 gb hash. |
| Re: Deepmind drops the bomb. |
Lester Antonio Tattersall Rodriguez |
12/5/17 9:46 PM |
I have a great look to it. And although I don't understand it (hard technique so more hard english), it looks super-interesting & exciting for masters and experts in AI. I hope they enjoy it.
On Wednesday, 6 December 2017 06:30:47 UTC+1, rohitr...@gmail.com wrote:
On Wednesday, 6 December 2017 10:36:03 UTC+5:30, Leebot wrote:
> Alpha0 now plays chess.
>
>
> https://arxiv.org/abs/1712.01815
link to paper
https://arxiv.org/pdf/1712.01815.pdf
alphazero scored +25 -3 =72 (+78 elo) against stockfish. Most likely they used stockfish 8
|
| Re: Deepmind drops the bomb. |
joost.van...@gmail.com |
12/5/17 9:51 PM |
On Wednesday, 6 December 2017 06:06:03 UTC+1, Leebot wrote:
> Alpha0 now plays chess. Really interesting. 100 games played against SF 8, not a single win for SF. Nevertheless, the elo increase over SF is surprisingly moderate (and only improves over SF at sufficiently long TC ~1s/move) There is also 10 games for analysis at the end of the paper. |
| Re: Deepmind drops the bomb. |
mal |
12/5/17 9:58 PM |
Might just be an artifact of the ELO calculation, 28 wins in 100 games sounds like 100's of ELO, not just 100
|
| Re: Deepmind drops the bomb. |
Lester Antonio Tattersall Rodriguez |
12/5/17 10:00 PM |
Don't worry, don't hurry. Give it a little more time, and there will be 100ELO gain ;)
On Wednesday, 6 December 2017 06:58:15 UTC+1, mal wrote:Might just be an artifact of the ELO calculation, 28 wins in 100 games sounds like 100's of ELO, not just 100
|
| Re: Deepmind drops the bomb. |
joost.van...@gmail.com |
12/5/17 10:18 PM |
On Wednesday, 6 December 2017 06:51:12 UTC+1, joost.van...@gmail.com wrote:
> On Wednesday, 6 December 2017 06:06:03 UTC+1, Leebot wrote:
> > Alpha0 now plays chess.
> There is also 10 games for analysis at the end of the paper. Also interesting is the opening analysis. Seemingly you better don't play B40 Sicilian with black against SF :-)
|
| Re: Deepmind drops the bomb. |
mal |
12/5/17 10:18 PM |
|
| Re: Deepmind drops the bomb. |
mal |
12/5/17 10:40 PM |
|
| Re: Deepmind drops the bomb. |
mal |
12/5/17 11:09 PM |
|
| Re: Deepmind drops the bomb. |
Leebot |
12/5/17 11:10 PM |
On Tuesday, December 5, 2017 at 10:40:48 PM UTC-8, mal wrote:
Here, as well.
|
| Re: Deepmind drops the bomb. |
mal |
12/5/17 11:21 PM |
|
| Re: Deepmind drops the bomb. |
Михаил Чалый |
12/5/17 11:33 PM |
As for now they are using 4 TPUs of 45 teraflops each, which is somewhat equal to 400 cores of modern xeons in terms of calculating power.
So basically they played 400 cores engine vs 60 cores engine... And got 70 elo difference, which is, I think, what you will get from playing equal engines. |
| Re: Deepmind drops the bomb. |
joost.van...@gmail.com |
12/5/17 11:38 PM |
On Wednesday, 6 December 2017 08:33:21 UTC+1, vizve...@gmail.com wrote:
> As for now they are using 4 TPUs of 45 teraflops each, which is somewhat equal to 400 cores of modern xeons in terms of calculating power.
> So basically they played 400 cores engine vs 60 cores engine... And got 70 elo difference, which is, I think, what you will get from playing equal engines. right, I agree with this analysis.... Nevertheless, well done. |
| Re: Deepmind drops the bomb. |
Михаил Чалый |
12/5/17 11:46 PM |
Sure, it's incredible that NN is avle to become so strong in such a short time.
But actually it played vs sf 8 (correct me if I'm wrong) and had like 400 cores vs 60. So it's maybe actually slightly weaker than current sf. Also looks like it's more weaker the less time you use for move. |
| Re: Deepmind drops the bomb. |
Михаил Чалый |
12/5/17 11:58 PM |
Yeah, found it, it was vs sf8.
Basically to summarize... They had like 400 vs 60 cores and played vs sf8. So +70 elo is probably slightly less than current sf dev will score vs it.
Also looking at their progress graph it doesn't look like it's getting much since 1st half.
Goes to show how strong simple bruteforce is now in chess since shogi programs are getting annihilated by alphazero. |
| Re: Deepmind drops the bomb. |
Михаил Чалый |
12/6/17 12:44 AM |
Also found that sf had 1 GB hash for 64 threads and 1 minute/move.
Okay, this result was somewhat rigged, I don't think that alphazero is stronger than SF dev. |
| Re: Deepmind drops the bomb. |
mal |
12/6/17 12:59 AM |
Yeah. Would be good if TCEC could arrange a match and make sure the conventional engine is set up properly.
On Wednesday, December 6, 2017 at 8:44:08 AM UTC, vizve...@gmail.com wrote:Also found that sf had 1 GB hash for 64 threads and 1 minute/move.
Okay, this result was somewhat rigged, I don't think that alphazero is stronger than SF dev.
|
| Re: Deepmind drops the bomb. |
Lester Antonio Tattersall Rodriguez |
12/6/17 1:05 AM |
I hope next year there is at least 4 super strong engine. And if SFish is the strongest, much better :). |
| Re: Deepmind drops the bomb. |
Gian-Carlo Pascutto |
12/6/17 2:37 AM |
On 06-12-17 06:51, joost.van...@gmail.com wrote:
> Nevertheless, the elo increase over SF is surprisingly moderate (and
> only improves over SF at sufficiently long TC ~1s/move)
They just stopped at a setting that beat it, it's not the limit of the
approach.
The paper is missing tons of details, but from one of the tables you can
see the the NN eval speed is much faster for chess than for Go, in other
words, they used a way smaller network. It topped out in strength just
above SF. What do you think happens if you double the size?
Question is now, can you find a tuned setting that compromises between
network size (search speed) and network strength so that it is stronger
than "SF 9" on mid-high ranged video card...
If not, you'll have to find ways to "speed up" the design, i.e. get a
better quality network with less computational resources.
--
GCP
|
| Re: Deepmind drops the bomb. |
Михаил Чалый |
12/6/17 2:45 AM |
looking at their graph of elo difference I don't see a lot of progress in latter half or 1/3 in any game in terms of getting elo. Even some regression is introduced. Also alphazero becomes worse the worse machine you use.
It's a good experiment but it actually shows how good are modern engines - in other games all engines are surpassed by enormous margin with flattening of elo, in chess it's only like 100~ while having much better computing power and inferior hash for sf. |
| Re: Deepmind drops the bomb. |
Gian-Carlo Pascutto |
12/6/17 2:46 AM |
On 06-12-17 11:45, vizve...@gmail.com wrote:
> looking at their graph of elo difference I don't see a lot of
> progress in latter half or 1/3 in any game in terms of getting elo.
> Even some regression is introduced.
It's what happens when the network reaches "capacity", so to say. You
can restart the experiment and it'll flatten at a higher level.
Obviously they saw no point in doing that.
(See past DM papers)
--
GCP
|
| Re: Deepmind drops the bomb. |
Gian-Carlo Pascutto |
12/6/17 2:48 AM |
On 06-12-17 11:46, Gian-Carlo Pascutto wrote:
> On 06-12-17 11:45, vizve...@gmail.com wrote:
>> looking at their graph of elo difference I don't see a lot of
>> progress in latter half or 1/3 in any game in terms of getting
>> elo. Even some regression is introduced.
>
> It's what happens when the network reaches "capacity", so to say.
> You can restart the experiment
...with a bigger network, that should have said.
> in other games all engines are surpassed by enormous margin
They used bigger networks and trained for longer. See the paper.
--
GCP
|
| Re: Deepmind drops the bomb. |
joost.van...@gmail.com |
12/6/17 6:58 AM |
On Wednesday, 6 December 2017 11:37:06 UTC+1, Gian-Carlo Pascutto wrote: > > Nevertheless, the elo increase over SF is surprisingly moderate (and > > only improves over SF at sufficiently long TC ~1s/move) > > They just stopped at a setting that beat it, it's not the limit of the > approach. yes, I agree, this is just the beginning of the approach, with a nice demo that it also works for chess. Gian-Carlo, given your github repo was mentioned for go.... any plans to come up with an implementation of this chess ? |
| Re: Deepmind drops the bomb. |
protonspring |
12/6/17 7:37 AM |
I've been involved in a new NN for chess with moderate success. The engine will likely be very strong in some positions, and extremely weak in others. The result was some amazingly strong games combined with some amazingly dumb losses. Perhaps a larger network would overcome some issues, but I'm not optimistic about overall success. Pretty interesting, but I don't think I'd worry too much about it. On Tuesday, December 5, 2017 at 10:06:03 PM UTC-7, Leebot wrote: |
| Re: Deepmind drops the bomb. |
Gian-Carlo Pascutto |
12/6/17 7:51 AM |
On 06-12-17 15:58, joost.van...@gmail.com wrote:
> yes, I agree, this is just the beginning of the approach, with a nice
> demo that it also works for chess.
>
> Gian-Carlo, given your github repo was mentioned for go.... any plans
> to come up with an implementation of this chess ?
Well StockFish is GPLv3 like leela-zero, so you're free to take all the
OpenCL, BLAS, TensorFlow, training, UCT/MCTS etc, code for Stockfish.
Replacing one movegen and board structure by another should be pretty
doable. Trickiest part to merge may be constructing the feature planes
for the inputs and outputs. I have to say that DeepMind's choice for
those isn't quite...what I would have done myself.
I have my hands a bit full right now though.
--
GCP
|
| Re: Deepmind drops the bomb. |
Ivan Ivec |
12/6/17 8:17 AM |
All games won with great material advantage:
And great scalability according to paper at Arxiv.
4000 Elo soon... |
| Re: Deepmind drops the bomb. |
fkra...@gmail.com |
12/6/17 9:38 AM |
On Wednesday, December 6, 2017 at 8:33:21 AM UTC+1, vizve...@gmail.com wrote: > As for now they are using 4 TPUs of 45 teraflops each, which is somewhat equal to 400 cores of modern xeons in terms of calculating power. > So basically they played 400 cores engine vs 60 cores engine... And got 70 elo difference, which is, I think, what you will get from playing equal engines. The paper said SF operated at 80 million positions per second and AlphaZero at only 80 thousand per second.
Running sf on my pc I can get around 5.5 million positions per second and that is pretty much enough to level any grandmaster. 80M/s is twice the TCEC power. |
| Re: Deepmind drops the bomb. |
Ivan Ivec |
12/6/17 10:33 AM |
Is this already on GitHub:
|
| Re: Deepmind drops the bomb. |
Михаил Чалый |
12/6/17 11:13 AM |
instead o reading some shit try to read the article.
They used 4 TPUs (which is like 400-2000 cores of normal cores based on how you count them) + they had stupid TC that favors NN (sf has some time allocation features that gives it elo... Yet is was forced to use exactly 1 minute/move), they played this 400-2000 cores vs only 64 cores sf with 1 GB hash (how often do you set 1 GB hash to 64 cores engine? It makes no sence), Alphazero used precalculated "good" openings vs no-book sf (brainfish someone? 50 elo just from adding book)... Basically it played vs insanely handicapped engine and won by 70-100 elo? Well, it's good, but it's not like "omg it's much better than any engine now". Not even close. You can say that it will improve itself further but elo graph looks like a flat line so I doubt it heavily. |
| Re: Deepmind drops the bomb. |
felix....@gmail.com |
12/6/17 11:30 AM |
El miércoles, 6 de diciembre de 2017, 11:13:32 (UTC-8), vizve...@gmail.com escribió:
> instead o reading some shit try to read the article.
> They used 4 TPUs (which is like 400-2000 cores of normal cores based on how you count them) + they had stupid TC that favors NN (sf has some time allocation features that gives it elo... Yet is was forced to use exactly 1 minute/move), they played this 400-2000 cores vs only 64 cores sf with 1 GB hash (how often do you set 1 GB hash to 64 cores engine? It makes no sence), Alphazero used precalculated "good" openings vs no-book sf (brainfish someone? 50 elo just from adding book)... Basically it played vs insanely handicapped engine and won by 70-100 elo? Well, it's good, but it's not like "omg it's much better than any engine now". Not even close. You can say that it will improve itself further but elo graph looks like a flat line so I doubt it heavily.
Yep.
Interesting project, but unfair conditions. Even Fritz with that computer power can crusg SF :D |
| Re: Deepmind drops the bomb. |
vkl...@gmail.com |
12/6/17 5:47 PM |
I honestly think the long Elo flat line you see in the learning/strength graph just shows how incredibly close we are right now to solving chess. 28-0-72 score seems like a total domination, but it's just 100 Elo difference against a year old Stockfish. A couple more years, and Stockfish will be able to grow only in short TC. I think there is a better reason why they selected older Stockfish version despite dev builds being openly available - they simply could not achieve anything dramatic with current Stockfish dev |
| Re: Deepmind drops the bomb. |
Nickolas |
12/6/17 7:36 PM |
On Wednesday, December 6, 2017 at 7:47:38 PM UTC-6, vkl...@gmail.com wrote:
> I honestly think the long Elo flat line you see in the learning/strength graph just shows how incredibly close we are right now to solving chess. 28-0-72 score seems like a total domination, but it's just 100 Elo difference against a year old Stockfish. A couple more years, and Stockfish will be able to grow only in short TC. I think there is a better reason why they selected older Stockfish version despite dev builds being openly available - they simply could not achieve anything dramatic with current Stockfish dev Keep in mind they trained the network for only 9 hours and didn't use endgame tablebases during training or play. A bigger network with more training time and perfect adjudication in TB positions would likely be significantly stronger. This is a sobering moment for engine developers. |
| Re: Deepmind drops the bomb. |
Михаил Чалый |
12/6/17 8:08 PM |
> Keep in mind they trained the network for only 9 hours and didn't use endgame tablebases during training or play. A bigger network with more training time and perfect adjudication in TB positions would likely be significantly stronger. This is a sobering moment for engine developers.
Keep in mind that "only" 9 hours of this 5000 TPUs is like 4 years of fishtest. And if graph has flattened - you can train for 24 more hours or 10 more years and achieve nothing. I don't even see the point of discussion in this case because a) testing match is insanely biased to be in alphazero favor (favourable conditions in both TC, hardware and opening book, directly sucking setting for SF (1 GB hash lol) and an old version), same goes to paper with all that "only for 4 hours" which is how you should never write any scientific document; b) result is like what, 100 elo on max TC with a starting position? Note that on fishtest regression tests are always done with 8 moves book and on 2 moves book they actually provide bigger elo gap. When they actually played from 4 moves starting pos elo gap shrinked to 75. And if you add total bias in test you will get that this alphazero chess engine is most likely weaker than sf dev; c) it's not like any engine developer can reproduce that because it requires absolutely ridiculous hardware and can't be done in any reasonable time on fishtest; d) looking at elo/time graph you can note that in order to be better than sf even with all bias it required 0.5s +/move on an insane hardware. So on ANY real hardware and reasonable TC with equal hardware it will be much weaker than even sf8. After all that I don't see anything "sobering" out of this test. It's actually showing how good stockfish is if it's only getting beaten by 100 elo while shogi program gets beaten by 300. |
| Re: Deepmind drops the bomb. |
Mindbreaker |
12/6/17 10:18 PM |
I looked at the games. The most noticeable difference to me is that Alpha is very good with its rooks and activates them early. Also, I think it is winning some endgames because of the Monte Carlo search. I also suspect it is learning the openings. We already know that if the openings are built in, an engine will play stronger. For a more level playing field they should have played against Brainfish. Also, no comment on endgame tables. Stockfish certainly benefits from endgame tables. On Tuesday, December 5, 2017 at 10:40:48 PM UTC-8, mal wrote: |
| Re: Deepmind drops the bomb. |
Magnus Irestig |
12/7/17 2:11 AM |
AlphaZero is clearly a big breakthrough in computer chess. It is probably still weaker than asmfish on big hardware in tournament time control.
The current SFdev is probably 30-40 ELO stronger than SF8 on long TC.
AsmFish gives perhaps 10 ELO more.
Bigger hash perhaps 10 ELO. Better time management at least another 10 ELO.
With help from Jonny-developer SF would clearly be able to benefit from extreme high core computer setups to match the computing power och the four TPUs of AlphaZero.
In the long run, I think it is more or less impossible to compete with the NN approach. This is their first shot at the problem and more progress is sure to come if they put effort into it but the match will be closer in chess, compared to Shogi or GO. |
| Re: Deepmind drops the bomb. |
mal |
12/7/17 11:50 AM |
I think the time control weakened sf noticeably - it's gonna get stuffed when it fails low.
I ran a test of sf with time control 40/40s vs sf with 1/1s, result: 19 - 9 - 72
No prizes for guessing which one lost.
|
| Re: Deepmind drops the bomb. |
mal |
12/7/17 2:51 PM |
Following up on the previous comment, I tried a test of latest SF against SF8. Just to give SFdev a hand, I gave it the 40/40s time control, 4MB hash and SF8 had 1/1s and 1MB hash. (Should hash be higher? I found some calc suggesting ~20MB)
Oh, and SFdev had Contempt=30 :-)
Result:
Score of sf8_1in1 vs sf_master: 11 - 45 - 44 [0.330] 100
Ok, this was only 1s / move, but still, beating SF8 45-11, i.e. winning 45% of the games seems like a pretty strong result. |
| Re: Deepmind drops the bomb. |
mal |
12/8/17 6:22 AM |
And another with slightly longer time controls (40/400s vs 1/10s) with latest SF in drawfish mode (i.e. Contempt=0):
Score of sf8_1in1 vs sf_master: 4 - 31 - 65 [0.365] 100
Elo difference: -96.19 +/- 38.81
SFlatest still managed to score 31 wins, but this time SF8 only managed 4 - fairly close to Alphazero's performance. (I think SF8 only managed 2% wins against Alphazero.)
No question that Alphazero is a big accomplishment, but it sounds like it's roughly the same level as SFlatest, not vastly better, and that's with the aid of a considerable hardware advantage. It would be interesting to see a well organised match between SF on some really big hardware and Alphazero ... |
| Re: Deepmind drops the bomb. |
Lester Antonio Tattersall Rodriguez |
12/8/17 7:56 AM |
Hello mal. Should be higher.
The size of Hash must be one size that the engine can be near to fullfilled its Hash at "every move" (the theoreticall optimus Hash). How many Hash it is your engine able to fullfill at every move? It depends about 2 things: time for every move/analysis (so... if time control it is not about a fixed time, then it must be use the "average time per move") and about the Speed (Kn/s). As much time and/or Speed, much possibilities to fullfill a bigger Hash.
The formula then it is (just aproximate formula)... for every move/analysis it is TIME x SPEED / 100 = Hash in Mb
Time must be in seconds. Speed must be in Kn (per second).
So how many (average) speed the engine achieves with the core(s) for an average movement? (it depends of number of threads, of course, it is using...) Per exemple...
Time control: 1' + increment 1'/move Speed: the engine normally with every "normal" move runs about 7800Kn/s.
Hash must be...
60'' for the whole game, plus an average game of 60 (moves) x 1' = 120 '' has the engine to make an average game (about 60 moves).
120'' (seconds) / 60 (moves) == 2''/move (two seconds per move, the average time of every move).
So just make:
2 x 7800 / 100 = 15600 / 100 = 156 (Megas for Hash). Probably the engine needes (or it is better) to round to a double of 2 (2, 4, 8, 16, 32, 64, 128, 256, 512...) in the Hash configuration...
So if round is needed: normally it is rounded down:
***** Hash must be set to 128Mb *** with 2'' per move (average) and with a speed of 7800Kn/s
(Could/should be round up: 256Mb Hash, if you can see how % of Hash the engine fullfill at every move, and you see it arrives too soon near 90-100% fullfilled, and still has time enough to increase 1 whole ply its depth analysis before it plays its move).
|
|
unk...@googlegroups.com |
12/8/17 7:57 AM |
<This message has been deleted.> |
|
unk...@googlegroups.com |
12/8/17 7:58 AM |
<This message has been deleted.> |
| Re: Deepmind drops the bomb. |
Ivan Ivec |
12/8/17 8:32 AM |
So, I need at least 2 TB of Hash for my analysis... |
| Re: Deepmind drops the bomb. |
Lester Antonio Tattersall Rodriguez |
12/8/17 9:14 AM |
If you want the optimum... yes. And so... whatever it is the Speed, for any "infinite analysis" mode, then... the hash should be as well "infinite" :) Just a theoretical formula for the (optimal) configuration in any standard/current hardwares and games. If you are planning a long "infinite analysis" you must put all the Hash your Hardware, and OSystem, and engine, allow to use for it. |
| Re: Deepmind drops the bomb. |
Uri Blass |
12/8/17 9:39 AM |
I expect a clearly smaller number wins for sf master when you use longer time control and the time control google used for the games is clearly longer(1 minute per move for 64 cores). If you use 40/4000 vs 1/100 seconds that is still faster than their time control I believe that you will get less than 25 wins for sf_master out of 100 games. |
| Re: Deepmind drops the bomb. |
mal |
12/8/17 11:43 AM |
Yes, I'm sure you're right there, especially about 64 threads vs my single-threaded tests. I did 1 move in 10s because it allowed 100 games in relatively short time, I don't really have the hardware to test longer than that. I might try some LTC games next week to compare with this 10s test, but they'll still be single-threaded to keep the total time down (I will use concurrency=3). I'm sure Deep Mind have come up with something big here, it's just annoying how it's like a lot of modern "science" - they seem to have rigged several things in their favour and then over-hyped the result instead of just doing a neutral test and a factual report.
On Friday, December 8, 2017 at 5:39:50 PM UTC, Uri Blass wrote:I expect a clearly smaller number wins for sf master when you use longer time control and the time control google used for the games is clearly longer(1 minute per move for 64 cores). If you use 40/4000 vs 1/100 seconds that is still faster than their time control I believe that you will get less than 25 wins for sf_master out of 100 games.
|
| Re: Deepmind drops the bomb. |
mal |
12/8/17 11:50 AM |
So with say 8x faster speed, we need 1Gb, then 60s per move is 30x so that's 32Gb hash. I think this calc probably overstates the requirement because it allows for variable time taken per move, but if we divide by 4 to allow for that, that's still 8Gb for correct sizing. Maybe it's only a minor disadvantage anyway, I haven't seen any reliable figures for the impact of under-sizing of the cache, but it obviously doesn't help and it makes Deep Mind look bad to me. On Friday, December 8, 2017 at 3:56:51 PM UTC, Lester Antonio Tattersall Rodriguez wrote: ...
***** Hash must be set to 128Mb *** with 2'' per move (average) and with a speed of 7800Kn/s
|
| Re: Deepmind drops the bomb. |
protonspring |
12/8/17 5:01 PM |
I took some time to look at a few games. In the first 10 moves, Stockfish consistently put itself into horrible positions and could never recover.
Seems like lack of an opening book crippled Stockfish making the results almost completely useless. |
| Re: Deepmind drops the bomb. |
arfa....@gmail.com |
12/24/17 6:05 PM |
Le mercredi 6 décembre 2017 06:06:03 UTC+1, Leebot a écrit :
> Alpha0 now plays chess.
>
>
> https://arxiv.org/abs/1712.01815
is it possible to have all the games played between AZ & SF |
| Re: Deepmind drops the bomb. |
zatar...@gmail.com |
12/25/17 5:35 AM |
You can complain all you like about the conditions of the setup. But I reran some of those positions on 36 cores Xeon with 32GB hash and it simply took Asmfish bmi2 about an hour to find 21 Bg5!! similar results with Houdini and regular Stockfish. Actually Houdini finds Bg5 in about 17 mins which seems to set the record. Opening book, tablebases and such aside, Alpha0 shows some great capabilities. |
| Re: Deepmind drops the bomb. |
Nickolas |
12/25/17 1:59 PM |
On Monday, December 25, 2017 at 7:35:38 AM UTC-6, zatar...@gmail.com wrote:
> You can complain all you like about the conditions of the setup. But I reran some of those positions on 36 cores Xeon with 32GB hash and it simply took Asmfish bmi2 about an hour to find 21 Bg5!! similar results with Houdini and regular Stockfish. Actually Houdini finds Bg5 in about 17 mins which seems to set the record.
>
> Opening book, tablebases and such aside, Alpha0 shows some great capabilities. There's no doubt that AlphaZero is capturing and effectively using chess knowledge that modern engines don't understand as readily, but the difference in playing strength isn't as stark as the paper makes it seem. (Indeed, it's very possible -- even probable -- that AlphaZero as tested wasn't actually the strongest chess playing entity in the world, despite running on millions of dollars of specialized hardware.) For one, Stockfish Master achieved a similarly dominating score against Stockfish 8 in a test run by someone on these forums, in which Stockfish Master had reasonable settings and Stockfish 8 was handicapped in the same way it was against AlphaZero. Two, while 21. Bg5 is a cool move, it's definitely not as impossible for modern engines to find as you're claiming it to be. I put the position into lichess's browser Stockfish and it showed up in the top 2 PVs at depth 40, which is a depth routinely reached in computer chess tournaments at that stage of the game. For me, the main takeaway from the oh-so-poorly run AlphaZero experiment is that manual selection and tuning of algorithims/features in position evaluation functions is probably on the way out. |
| Re: Deepmind drops the bomb. |
zatar...@gmail.com |
12/26/17 5:32 AM |
Yes there are many tricks that could be used to show 21 Bg5. But under standard engine settings and a single PV, it will take 1 hour for asmfish/stockfish to select 21. Bg5!! and thats with 2x Xeon, 32G hash. AsmFish
00:00:00.0 0.85 1 1043 Bf4
00:00:00.0 0.73 2 1173 Bf4 c5
00:00:00.0 -0.14 3 3566 Bf4 d5 Nd2
00:00:00.0 5.07 4 6705 Nf6 gxf6 Be4 d5 Bxh7 Kxh7
00:00:00.0 3.78 5 7785 Nf6 gxf6 Be4 Nd6 Bxh7
00:00:00.0 0.56 6 17717 Bf4 d5 Nd6 f5 Qd1 Nd7
00:00:00.0 -0.04 7 45003 Qf4 Kg8 Bh3 d5 Bf5 Qh8 Nd6
00:00:00.0 -0.38 8 48180 Qf4 d5 Nd6 Nd7 Nxb7 Bxb7 Re7 Nc5
00:00:00.0 -0.42 9 61771 Qf4 d5 Nd6 Nd7 Nxb7 Bxb7 Re7 Rad8 Rxf7 Rxf7 Qxf7
00:00:00.0 -0.16 10 86633 Qf4 d5 Nd6 Nd7 Nxf7+ Kg8 Re7 Ndc5 Be3 Nd3 Qf3 Nxb2
00:00:00.1 -1.06 11 245303 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 Bf4 Qf5 Rad1 Qxh5 Rd4 Rc8 Qa3 Qf5
00:00:00.1 -0.19 12 314678 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 Bf4 Qd3 a4 Rc8 Rac1 Qf5 Bf3 Qd3 Be2
00:00:00.1 -0.11 13 1064756 b3 d5 Ba3 Rd8 Nc3 Qg8 Qd4 f6 Be7 Rd7 Bxf6 Qf7 Be5 Qxh5
00:00:00.2 0.15 14 4692639 Nc3 d5 b4 Nd6 Qf4 Rd8 b5 Qc2 Nxd5 cxd5 bxa6 Nxa6 Bxd5 Rac8 Ba3 Nc5 Rac1 Qf5 Qxf5 Nxf5 Bxf7 Nd3
00:00:00.2 0.17 15 5078074 b4 d5 Nc3 Bd3 b5 Nc5 Ba3 Qf5 Qxf5 Bxf5 Na4 Rc8 Bxc5 bxc5 Nxc5 Nd7
00:00:00.2 0.32 16 8220222 b4 d5 Nc3 Nd6 a4 Bc4 b5 Qf5 Qxf5 Nxf5 Bf4 Nd4 Rac1 Rc8 Re7 Ne6 Rxf7 Nxf4 Rxf4
00:00:00.3 0.09 17 13950268 b4 d5 Nc3 Bc4 Bf4 Qg8 Rad1 Re8 b5 Rxe1+ Rxe1 Nc5 Bxb8 Qxb8 bxc6 Qc8 Qxc8+ Rxc8 Nxd5 Bxa2 Ne7 Rc7 Nf5 Be6 Nd6 Re7 Re3 Kh7
00:00:00.4 0.06 18 20235648 b4 d5 Nc3 Bc4 Bf4 Qg8 Re7 Nd8 b5 Re8 Rxe8 Qxe8 bxc6 Qxc6 Qd1 Qd7 Nxd5 Nbc6 Rc1 Bxa2 Ne3 Qxd1+ Nxd1 Rc8 Nc3
00:00:00.8 0.30 19 40387336 b4 d5 Nc3 Nd6 a4 Bb7 b5 Rd8 Ba3 Nc4 Be7 Re8 Bf6 Re6 Rxe6 fxe6 Qxe6 gxf6 bxc6 Nxc6 Qxf6+ Qg7
00:00:01.0 0.36 20 54316605 b4 d5 Nc3 Bc4 Bf4 Qg8 Re7 Nd8 Rae1 Na6 a3 Ne6 R1xe6 fxe6 Be5 Rf7 Rxe6 Qh7 Rxc6 Re8 f4 Qc2 Rxh6+ Kg8 Qg6 Qxg6 Rxg6 Rxe5 fxe5
00:00:02.2 0.65 21 115113930 b4 d5 Nc3 Nd6 a4 Bc8 Qf4 Rd8 b5 Qf5 Qh4 f6 Bf4 Be6 Ne2 c5 g4 Qc2 Bxh6 Nd7 Nd4 cxd4 Rxe6 Qb2 Rae1 Nxb5 axb5 gxh6 Bxd5 Qxb5 Bxa8 Rxa8
00:00:02.8 0.60 22 152729702 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Re7 Nd3 Bd2 Re8 Rxf7 d4 Rc7 dxc3 Bxc3
00:00:02.9 0.65 23 156973075 b4 d5 Nc3 Nd6 a4 Bd3 b5 Nc4 Bf4 Na5 Qh4 Nb3 Ra2 f6 Re7 Bc4 Qg4 Rd8 Rb2 Nc5 Bc7 Rg8 Bxb8 Raxb8 bxc6
00:00:03.5 0.57 24 189015810 b4 d5
00:00:03.8 0.65 24 206711258 b4
00:00:04.3 0.53 24 230285600 b4 d5 Nc3 Nd6 a4 Bd3 b5 Nc4 Bf4 Na5 Qh4 Nb3 Bd6 Rc8 Ra2 Nc5 Bxc5 bxc5 Rd2 c4 Qf4 Qf5 Qxf5 Bxf5 bxc6 Be6 c7 Rxc7 Bxd5
00:00:04.4 0.60 25 237453293 b4
00:00:05.0 0.68 25 272290321 b4
00:00:05.4 0.65 25 291484822 b4 d5 Nc3 Nd6 a4 Qd3 Bf4 Bc8 Qd1 Qxd1 Raxd1 Nb7 Re7 Nd8 Bd6 Bg4 Rc1 Na6 Bf1 Nxb4 Bxb4 c5 Ba3 d4 Bg2 Rc8 Nb5 Nc6 Rb7 Bxh5 Nxa7 Nxa7
00:00:06.9 0.61 26 372778157 b4 d5 Nc3 Bc4 Bf4 Qg8 Re7 Nd8 Rae1 Na6 a3 Ne6 R1xe6 fxe6 Be5 Rf7 Rxe6 Qc8 Bh3 Kg8 Qg6 Nc7 Bf5 Nxe6 Bxg7 Nxg7 Qh7+ Kf8 Bxc8 Ke7 Bb7 Nf5
00:00:09.7 0.65 27 529787662 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qc2 Rec1 Qf5 Ne4 Ba6 Nd6 Qxh5 Qd7 c5 Re1 Bc8 Nxc8 Raxc8 Re5 Qg6 a3 Rcd8 Qf5 Nd3 Rxd5 Rxd5 Qxd5 Nxf4
00:00:10.9 0.57 28 590538051 b4 d5
00:00:11.5 0.50 28 627698096 b4 d5
00:00:11.5 0.57 28 628374058 b4
00:00:11.9 0.58 28 649165123 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Nd1 d4 a3 Nd3 Qxc6 Nxe5 Bxe5 f6 Bxd4 Qd3 Bb2 Qe2 Rb1 Bd3 Nc3 Qxh5 Rd1 Rad8 Qb7 Rf7
00:00:14.8 0.66 29 807543120 b4
00:00:18.1 0.55 29 985779954 b4 d5 Nc3 Nd6 a4 Qd3 Bf4 Bc8 Qd1 Qxd1 Raxd1 Nb7 Re7 Nd8 Bd6 Bg4 Rc1 Na6 Bf1 Nxb4 Bxb4 Ne6 Be2 Bxe2 Nxe2 c5 Bd2 Rfe8 Rxe8+ Rxe8 Kg2 d4
00:00:21.2 0.48 30 1155051585 b4 d5
00:00:21.5 0.40 30 1176898083 b4 d5
00:00:21.8 0.33 30 1189776519 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Nd1 d4 a3 Nd3 Qxc6 Nxe5 Bxe5 f6 Bxd4 Qd3 Bb2 Rac8 Qf3 Qd2 Bc3 Qc2 Bd4 Qd2 Qe3 Qxe3
00:00:25.6 0.40 31 1396355746 b4
00:00:26.4 0.48 31 1439847827 b4
00:00:30.0 0.59 31 1633717577 b4
00:00:32.7 0.60 31 1780674345 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Nd1 d4 Be4 Bd3 Bf3 Bc4 Bg2 Rad8 Qxa7 Nd3 Be4 Qe2 Bxd3 Qxd3 Nb2 Qc3 Rc1 Qxb2 Rxc4 d3
00:00:41.1 0.68 32 2147483647 b4
00:00:41.8 0.53 32 2147483647 b4 d5
00:00:42.0 0.53 32 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Re7 Kg8 Bf1 Bxf1 Kxf1 d4 a3 Nd3 Ne4 Re8 Rxe8+ Rxe8 Bc7 c5 g4 Re6
00:00:43.6 0.60 33 2147483647 b4
00:00:46.8 0.60 33 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Re7 Kg8 Bf1 Bxf1 Kxf1 d4 a3 Nd3 Ne4 Re8 Rd7 c5 Bc7 f5 Nd6 Ne5
00:00:53.2 0.68 34 2147483647 b4
00:00:58.3 0.61 34 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qc2 Rac1 Qf5 Ne4 Qxh5 Nd6 Nd3 Qxc6 Qg6 Qc7 Bxa2 Ra1 Nxe1 Rxe1 b5 Re7 Kg8 Qa5 Bc4 Nxc4 bxc4 Bxd5 Qf5
00:01:03.9 0.54 35 2147483647 b4 d5
00:01:12.9 0.54 35 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qc2 Rec1 Qf5 Ne4 Ba6 Qe7 c5 Nd6 Qxh5 Rc3 Nc6 Bf3 Nxe7 Bxh5 Ng6 Bd2 Rad8 Nxf7+ Rxf7 Bxg6 Re7 Ra3 Rd6
00:01:33.8 0.46 36 2147483647 b4 d5
00:01:36.5 0.54 36 2147483647 b4
00:01:38.7 0.65 36 2147483647 b4
00:02:04.2 0.71 36 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Re7 Kg8 Bh3 Re8 Rxe8+ Rxe8 a3 Nc2 Rc1 Ne1 Bd7 Nxf3+ Kf2 Rd8 Bxc6 Nd4
00:02:35.8 0.78 37 2147483647 b4
00:02:51.8 0.86 37 2147483647 b4
00:03:04.6 0.63 37 2147483647 b4 d5
00:03:25.8 0.47 37 2147483647 b4 d5
00:03:58.3 0.26 37 2147483647 b4 d5
00:03:59.3 0.45 37 2147483647 b4
00:04:06.2 0.57 37 2147483647 b4 d5 Nc3 Bc4 Bf4 Qg8 Re7 Nd8 Rae1 Na6 a3 Ne6 R1xe6 fxe6 Be5 Rf7 Rxe6 Qc8 Qg6 Kg8 Bh3 Nc7 Bf5 Nxe6 Bxg7 Nxg7 Qh7+ Kf8 Bxc8 Ke7 Bb7 Nf5
00:04:38.8 0.59 38 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Nd1 d4 Be4 Qe2 Rc1 Nd5 Nb2 Qxb2 Rxc4 Nxf4 gxf4 Qd2 Qd7 f5 Bxf5 c5 Qe6 Qd1+ Kg2 Qxh5
00:04:51.7 0.67 39 2147483647 b4
00:05:11.7 0.74 39 2147483647 b4
00:06:40.5 0.81 39 2147483647 b4 d5 Nc3 Nd6 a4 Bd3 b5 Nc4 Bf4 d4 bxc6 dxc3 c7 Na6 Bxa8 Nxc7 Bxc7 Rxa8 Re7 Qf5 Qxf5 Bxf5 Rxf7 Bh7 Bf4 c2 Re1 Rd8 Kg2 Nb2 Ree7 Nd3
00:08:14.0 0.89 40 2147483647 b4
00:09:58.7 0.90 40 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc8 Qxc8 Rfxc8 Re7 Nd3 Bh3 Rd8 Rc7 c5 Bg2 Nxf4 gxf4 Rab8 Bf1 Bxf1 Kxf1 d4 Ne4 d3 Rxf7 Rd4
00:10:06.1 0.82 41 2147483647 b4 d5
00:10:06.9 0.75 41 2147483647 b4 d5
00:10:10.7 0.64 41 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Re7 Re8 Rc7 Rec8 Rxc8+ Rxc8 Rd1 Rd8 Bf1 Bxf1 Kxf1 d4 a3 Nc2 Ne4 Rd5
00:10:30.5 0.72 42 2147483647 b4
00:10:34.5 0.79 42 2147483647 b4
00:11:14.0 0.77 42 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc8 Qxc8 Rfxc8 Re7 Nd3 Bh3 Rd8 a4 d4 Ne4 Bd5 Bg2 f5 Nd2 Bxg2 Kxg2 Re8 Rd7 Re2 Rxd4 Nxf2
00:11:27.5 0.70 43 2147483647 b4 d5
00:11:46.5 0.62 43 2147483647 b4 d5
00:13:55.5 0.70 43 2147483647 b4
00:14:57.9 0.86 43 2147483647 b4
00:17:54.9 1.03 43 2147483647 b4 d5 Nd2 Nd6 Qf4 Nc4 a4 Nd7 b5 cxb5 Nxc4 bxc4 Ba3 Nc5 Bxc5 bxc5 Qd6 Bb7 Qc7 Bc8 Bxd5 Be6 Be4 Rfc8 Qxc8+ Rxc8 Bxh7 Kxh7 Rab1 Rc6 a5 c3
00:18:19.7 0.95 44 2147483647 b4 d5
00:19:09.2 1.01 44 2147483647 b4 d5 Nd2 Nd8 Nf3 Bd3 Nh4 Ne6 Rxe6 fxe6 Ng6+ Bxg6 hxg6 Qg8 Qh4 Rf6 Bxh6 Rxg6 Bf4+ Qh7 Qe7 Na6 Qb7 Re8 Qxa6 e5 Qe2 Rge6 Bd2 Qf5 b5 e4
00:21:21.5 1.08 45 2147483647 b4
00:21:30.1 0.93 45 2147483647 b4 d5
00:24:35.7 0.98 45 2147483647 b4 d5 Nd2 Nd6 Qf4 Nc4 a4 Nd7 b5 cxb5 Nxc4 bxc4 Ba3 Nc5 Bxc5 bxc5 Qd6 Bb7 Qc7 Bc8 Bxd5 Be6 Be4 Rfc8 Qxc8+ Rxc8 Bxh7 Kxh7 Re5 Rc6 a5 g6
00:24:45.2 0.91 46 2147483647 b4 d5
00:26:32.0 0.83 46 2147483647 b4 d5
00:26:35.7 0.91 46 2147483647 b4
00:30:02.1 1.07 46 2147483647 b4
00:30:57.8 0.86 46 2147483647 b4 d5 Nd2 Nd6 Qf4 Nc4 a4 Nd7 b5 cxb5 Nxc4 bxc4 Ba3 Nc5 Bxc5 bxc5 Qd6 Bb7 Qc7 Bc8 Bxd5 Be6 Be4 Rfc8 Qxc8+ Rxc8 Bxh7 Kxh7 Re5 Rc6 a5 g6
00:31:30.1 0.93 47 2147483647 b4
00:32:36.4 1.01 47 2147483647 b4
00:33:18.9 0.88 47 2147483647 b4 d5 Nd2 Nd6 Qf4 Nc4 a4 Nd7 b5 cxb5 Nxc4 bxc4 Ba3 Nc5 Bxc5 bxc5 Qd6 Bb7 Qc7 Bc8 Bxd5 Be6 Be4 Rfc8 Qxc8+ Rxc8 Bxh7 Kxh7 Re5 Rc6 a5 g6
00:35:32.8 0.83 48 2147483647 b4 d5 Nd2 Nd6 Qf4 Nc4 a4 Nd7 b5 cxb5 Nxc4 bxc4 Ba3 Nc5 Bxc5 bxc5 Qd6 Bb7 Qc7 Bc8 Bxd5 Be6 Be4 Rfc8 Qxc8+ Rxc8 Bxh7 Kxh7 Re5 Rc6 a5 g6
00:38:49.3 0.91 49 2147483647 b4
00:47:53.2 0.98 49 2147483647 b4
00:56:06.9 1.10 49 2147483647 Bg5
01:01:09.3 1.25 49 2147483647 Bg5 ******************************
Stockfish 00:00:00.0 0.62 1 425 Bf4
00:00:00.0 0.62 2 604 Bf4 c5
00:00:00.0 -0.47 3 6102 Nf6 gxf6 Be4
00:00:00.0 -0.75 4 6512 Qd1 d5 Nc3 Nd7
00:00:00.0 -1.14 5 8206 Qh3 d5 Nc3 Nc5 g4
00:00:00.0 -1.44 6 12469 Kh1 d5 Nc3 f5 Qf3 d4
00:00:00.1 -0.11 7 30049 Bf4 d5 Nd2 Nc5 Bd6 Nbd7 Bxf8 Rxf8
00:00:00.1 -0.11 8 34831 Bf4 d5 Nd2 Nc5 Bd6 Nbd7 Bxf8 Rxf8
00:00:00.4 -0.44 9 125910 Qf4 d5 Nd6 Nxd6 Qxd6 Rg8 Qf4 Nd7 Qxf7 Nc5
00:00:01.9 -0.23 10 570885 Nd2 Nc5 Nb3 Nd3 Re7 d6 Bf4 d5 Nd4 Nxb2 Bxb8 Raxb8 Nxc6
00:00:02.6 -0.13 11 787921 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 Bf4 Qc2 Re7 Bb5 b3 Na6 a4
00:00:04.2 -0.19 12 1259571 Qf4
00:00:04.2 -0.34 12 1270986 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 Re7 Qf5 Bd2 Qxh5 Rae1 Qg6 Bf4 Qxd6 Bxd6
00:00:04.7 -0.27 13 1419068 Qf4
00:00:04.8 -0.28 13 1428498 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 b3 Qf5 Ba3 Rc8 Re5 Qg4 Rae1 Nd7 Re7 Nc5 Rxf7
00:00:05.6 -0.19 14 1639317 Qf4
00:00:05.6 -0.10 14 1647404 Qf4
00:00:06.2 -0.22 14 1786398 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 Re5 Bc8 Bf4 Bg4 Rae1 Nd7 Re7 Nc5 Qxc6 Be6 b3 Qf5
00:00:09.2 -0.14 15 2521291 Qf4
00:00:09.6 -0.06 15 2605181 Qf4
00:00:10.1 -0.17 15 2717071 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 b3 Qf5 Ba3 Rc8 Re5 Qf6 Qxf6 gxf6 Re7 Rd8 Bb2 Nd7 Bh3
00:00:27.3 -0.09 16 6498388 b4
00:00:30.1 -0.02 16 7215982 b4
00:00:36.5 0.10 16 8893493 b4
00:00:38.5 0.20 16 9656419 b4 d5 Nc3 Nd6 a4 Bc8 Qf3 Qf5 b5 Be6 Bf4 Rd8 Rad1 Qg4 Rd4 Qxf3 Bxf3
00:00:47.0 0.27 17 13623262 b4
00:00:53.4 0.35 17 17783776 b4
00:00:54.4 0.15 17 18542327 b4 d5 Nc3 Nd6 Bf4 Ne8 Be5 f5 Qf3 Nd7 b5 Bb7 Rad1 Nxe5 Rxe5 cxb5 Re7 Bc6 Nxd5 Nf6
00:00:55.5 0.23 18 19574926 b4
00:00:56.3 0.30 18 20342831 b4
00:01:04.7 0.08 18 32183794 b4 d5
00:01:04.7 0.25 18 32378139 b4
00:01:07.2 0.39 18 39837379 b4 d5 Nc3 Nd6 a4 Bb7 b5 Rd8 Ba3 c5 Bxd5 Nd7 Bxb7 Nxb7 Nd5 Nf8 Ne7 Rd3 Nc6 Nd6 Ne5 Rb3 Rad1
00:01:08.1 0.44 19 44220973 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nd8 Bd6 Nb8 Qg4 Rg8 b5 Ne6 bxc6 Nxc6 Nxd5 Bd3 Rac1 Ncd4 Be5 Nc2 Red1 Rad8 Rxd3 Qxd3
00:01:09.9 0.51 20 66736962 b4
00:01:10.2 0.53 20 75771688 b4 d5 Nc3 Nd6 a4 Bd3 b5 Nc4 Bf4 Na5 Bd6 Rd8 Bc7 Rg8 Re7 Bf5 Qf4 Nd7 Rxf7 Be6 bxc6 Nxc6 Bxd5 Bxf7 Bxc6
00:01:10.6 0.46 21 87391833 b4 d5
00:01:10.7 0.53 21 89113538 b4
00:01:10.7 0.54 21 90752495 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qc2 Rac1 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Bf1 Bxf1 Rxf1 d4 Ne4 Rd8 Re7 Kg8 Rb7 c5 Nd6 Nd3 Nf5 Re8
00:01:11.0 0.47 22 99327410 b4 d5
00:01:11.6 0.54 22 119473702 b4
00:01:13.6 0.39 22 181330738 b4 d5
00:01:13.6 0.52 22 182520275 b4
00:01:14.3 0.70 22 204264802 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Bf1 Bxf1 Kxf1 d4 Ne4 c5 Nd6 Rc7 Rae1 Nxa2 Re7 Rxe7 Rxe7
00:01:14.8 0.62 23 218386495 b4 d5
00:01:14.8 0.55 23 218613506 b4 d5
00:01:14.9 0.58 23 219319636 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Rc1 Qb2 Qe7 Nd3 Re2 Rae8 Qxf8+ Rxf8 Rxb2 Nxb2 Nd1 Nd3 Rc2 Nxf4 gxf4 b5 Ne3 g6
00:01:14.9 0.50 24 220676919 b4 d5
00:01:14.9 0.54 24 222097917 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Rc1 Qb2 Qe7 Nd3 Re2 Rae8 Qxf8+ Rxf8 Rxb2 Nxb2 Nd1 Nd3 Rc3 Nxf4 gxf4 a5 Nb2 b5 Ra3 Re8
00:01:16.0 0.47 25 260067563 b4 d5
00:01:16.1 0.39 25 265159982 b4 d5
00:01:16.2 0.47 25 267790447 b4
00:01:16.9 0.57 25 292058430 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Re7 Kg8 Rd1 Nd3 Bf1 Nxf4 gxf4 b5 Kf2 Kf8 Rb7 Rcb8 Rc7 Rc8 Rxc8+ Rxc8
00:01:17.5 0.55 26 314045554 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qc2 Rec1 Qf5 Ne4 Ba6 Nd6 Qxh5 Qd7 c5 Re1 Bc8 Nxc8 Raxc8 Re5 Qg6 Rd1 Qc2 Rd2 Rcd8 Qg4 Qb1+ Qd1 Qxd1+
00:01:19.3 0.63 27 378558826 b4
00:01:20.6 0.58 27 425838391 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qc2 Rec1 Qf5 Ne4 Ba6 Nd6 Qxf4 gxf4 Bxb7 Nxb7 Rab8 Nd6 Rbd8 a3 Nc2 Rxc2 Rxd6 Kf1 c5 Ke2 g6 Rg1 gxh5
00:01:21.1 0.50 28 443250920 b4 d5
00:01:22.5 0.43 28 494283203 b4 d5
00:01:23.7 0.39 28 542249096 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Rc1 Qb2 Qe7 Nd3 Re2 Rae8 Qxf8+ Rxf8 Rxb2 Nxb2 Nd1 Nd3 Rc2 Nxf4 gxf4 b5 Bf3 g5 f5 Kg7
00:01:27.1 0.31 29 668376164 b4 d5
00:01:27.6 0.31 29 686967067 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Nd1 d4 Re7 Bd5 Bxd5 Nxd5 Rxf7 Rxf7 Qxa8+ Kh7 Rc1 Qf5 Qxc6 Re7 Qg6+ Qxg6 hxg6+ Kxg6 Kf1 Nxf4
00:01:32.3 0.38 30 866521864 b4
00:01:32.9 0.39 30 887988087 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Nd1 d4 a3 Nd3 Qxc6 Nxe5 Bxe5 f6 Bxd4 Qd3 Bb2 Qe2 Bc3 Rac8 Qf3 Qc2 Qf4 Rfd8 Bd4 Be6
00:01:41.1 0.32 31 1200605798 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Nd1 d4 a3 Nd3 Qxc6 Nxe5 Bxe5 f6 Bxd4 Qd3 Bb2 Qe2 Bc3 Rac8 Qf3 Qc2 Qf4 Rfd8 Bd4 Be6
00:01:45.4 0.39 32 1367284947 b4
00:01:46.9 0.47 32 1425100062 b4
00:01:48.6 0.51 32 1492741623 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Nd1 d4 a3 Nd3 Qxc6 Nxe5 Bxe5 f6 Bxd4 Qd3 Bb2 Rae8 Ne3 Bf7 Bf1 Qb3 Bd4 Bxh5 Qc7 Rd8
00:01:55.5 0.55 33 1762253048 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Bh3 Rd8 Re7 Kg8 Rd1 Nd3 Bc7 Re8 Rxe8+ Rxe8 Bf5 Re3 Bxd3 Rxd3 Rxd3 Bxd3
00:02:05.6 0.62 34 2147483647 b4
00:02:06.7 0.70 34 2147483647 b4
00:02:12.1 0.59 34 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Re7 Re8 Rc7 Rec8 Rxc8+ Rxc8 Rd1 Rd8 Bh3 Kg8 Bf5 c5 a3 Nc6 Bc7 Re8
00:02:18.6 0.66 35 2147483647 b4
00:02:42.7 0.74 35 2147483647 b4
00:03:14.5 0.85 35 2147483647 b4
00:03:33.5 0.87 35 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Nd1 Nd3 Qxc6 Nxe5 Bxe5 f6 Ne3 Qe2 Bd6 Rfe8 Bh3 Qxh5 Bd7 Be2 Bxe8 Rxe8 Qxd5 Qxd5 Nxd5 Bf3
00:03:42.0 0.80 36 2147483647 b4 d5
00:03:56.6 0.72 36 2147483647 b4 d5
00:03:59.3 0.61 36 2147483647 b4 d5
00:04:16.3 0.45 36 2147483647 b4 d5
00:04:33.7 0.57 36 2147483647 b4
00:04:54.1 0.77 36 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qd3 Re3 Qf5 Bd6 Ba6 g4 Qf6 Qe7 Nc2 Qxf6 gxf6 Bxf8 Rxf8 Rd1 Nxe3 fxe3 Bc8 Bf3 Rg8 Kf2 Bxg4
00:05:29.0 0.84 37 2147483647 b4
00:05:43.7 0.69 37 2147483647 b4 d5
00:05:45.0 0.58 37 2147483647 b4 d5
00:06:01.8 0.42 37 2147483647 b4 d5
00:06:06.5 0.55 37 2147483647 b4
00:06:25.1 0.78 37 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc2 Nd1 d4 Be4 Bd3 Rc1 Qxa2 Bxc6 Rad8 Bg2 Bc2 Qe7 Nd3 Bxh6 Nxe5 Qxe5 f6 Bxg7+ Kxg7 Qc7+ Kh8
00:07:25.4 0.86 38 2147483647 b4
00:07:49.9 0.87 38 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qf6 Qe7 Qxe7 Rxe7 Kg8 a4 Rfd8 Bc7 Re8 Rae1 Rxe7 Rxe7 d4 Nb5 d3 Nc3 d2 Rd7 Nd5 Bxd5 Bxd5
00:08:00.0 0.80 39 2147483647 b4 d5
00:08:06.5 0.72 39 2147483647 b4 d5
00:08:10.8 0.72 39 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc8 Qxc8 Rfxc8 Re7 Kg8 Bd6 Nd3 Nd1 Re8 Rxe8+ Rxe8 Ne3 b5 Rd1 g6 Bf1 Nb2 Rc1 Bxf1 Kxf1 d4
00:08:30.7 0.65 40 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc8 Qxc8 Rfxc8 Re7 Kg8 Bd6 Nd3 Nd1 Rd8 Ba3 Re8 Ne3 b5 Rd1 Rxe7 Bxe7 Re8 Bd6 g6 Bf1 Nb2
00:09:49.1 0.73 41 2147483647 b4
00:10:15.2 0.80 41 2147483647 b4
00:11:10.0 0.92 41 2147483647 b4
00:11:31.2 0.75 41 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc8 Qxc8 Rfxc8 Re7 Nd3 Bh3 Rd8 Rc7 d4 Rxc6 b5 Nd1 Re8 Rd6 Rad8 Rxd8 Rxd8 a4 Nxf4 gxf4 bxa4
00:12:23.0 0.69 42 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc8 Qxc8 Rfxc8 Re7 Nd3 Bh3 Rd8 Rc7 d4 Rxc6 b5 Nd1 Re8 Rd6 Rad8 Rxd8 Rxd8 a4 Nxf4 gxf4 bxa4
00:13:23.3 0.74 43 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc8 Qxc8 Rfxc8 Re7 Nd3 Bh3 Rd8 Rc7 d4 Rxc6 b5 Nd1 Re8 Rd6 Rad8 Rxd8 Rxd8 a4 Nxf4 gxf4 bxa4
00:16:55.0 0.81 44 2147483647 b4
00:17:21.7 0.66 44 2147483647 b4 d5
00:18:54.7 0.55 44 2147483647 b4 d5
00:19:20.9 0.66 44 2147483647 b4
00:20:18.7 0.75 44 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 f3 Qc8 Qxc8 Rfxc8 Bh3 Rd8 a3 Nd3 Re7 c5 Rc7 d4 Ne4 Nxf4 gxf4 Be6 Bxe6 fxe6 Re1 Rdc8
00:21:15.3 0.82 45 2147483647 b4
00:23:35.2 0.90 45 2147483647 b4
00:30:23.2 0.67 45 2147483647 b4 d5
00:30:43.5 0.84 45 2147483647 b4
00:32:28.6 0.76 45 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc8 Qxc8 Rfxc8 Re7 Nd3 Bh3 Rd8 Rc7 d4 Ne4 Bd5 Bf5 Nxf4 gxf4 c5 f3 Be6 Bxe6 fxe6 Kf2 Rdc8
00:34:16.5 0.76 46 2147483647 b4 d5 Nc3 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qc8 Qxc8 Rfxc8 Re7 Nd3 Bh3 Rd8 Rc7 d4 Rxc6 b5 Nd1 Re8 Rd6 Rad8 Rxd8 Rxd8 a4 Nxf4 gxf4 bxa4
00:34:53.8 0.83 47 2147483647 b4
00:41:35.0 0.91 47 2147483647 Bg5 ************************
Houdini 00:00:06.3 0.25 12 9089844 b3 d5 Ba3 Rd8 Nd2 Nc5 Bxc5 bxc5 Qf4 Nd7 Qxf7 Nf6 Nf3
00:00:06.3 0.26 12 10511178 b3 d5 Ba3 Rd8 Nd2 Nc5 Bxc5 bxc5 Qf4 Bd3 Qxf7 Na6 Nf3 Rf8 Qb7
00:00:06.4 0.34 13 16364002 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 b3 Qf5 Ba3 Rc8 Re5 Qg4 Re7 Bd3 Rae1 Qxh5 Re8+
00:00:06.5 0.34 14 18338846 Qf4 d5 Nd6 Nxd6 Qxd6 Rg8 Re7 Qf5 Bxh6 gxh6 Qxh6+ Qh7 Qf4 Rf8 Rae1 Qxh5 R1e6 Qd1+ Kh2
00:00:06.6 0.27 15 27995043 Qf4 d5 Nd6 Nxd6 Qxd6 Rg8 Bf4 Bb5 Re7 f6 b3 Qf5 Bf3 Qd3 Bg4 Qc3 Rc1 Qb2
00:00:06.9 0.21 16 42758550 b3 d5 Ba3 Rd8 Qf4 Nd7 Nd6 Nxd6 Bxd6 Nc5 Bxc5 bxc5 Re7 f6 Qc7 Bb5 a4 Bd3 Qxc6
00:00:08.9 0.49 17 160749374 Nc3 d5 b4 Nd6 b5 Nxb5 Nxd5 cxd5 Bxd5 Nc7 Bxa8 Nxa8 Ba3 Rc8 Qf3 Qg8 Rac1 Nc7 Qf4 Nd5 Rxc8 Qxc8
00:00:09.7 0.48 18 209803382 Nc3 d5 b4 Bc4 Bf4 Na6 Qd7 Nxb4 Qxb7 Qf5 Re5 Qg4 Qc7 d4 Ne4 Nd5 Qxc6 Rac8 Qd6 Rfd8 f3 Qd7
00:00:10.4 0.40 18 249157105 Nc3 Nc5
00:00:11.5 0.31 18 314266025 Nc3 d5
00:00:12.4 0.40 18 366287317 Nc3
00:00:12.6 0.28 18 374358778 Nc3 d5
00:00:13.4 0.25 18 419790461 Nc3 d5 b4 Bc4 Bf4 Qg8 Re7 Nd8 b5 Ne6 Bxb8 Rfxb8 bxc6 d4 Nd5 Re8 Rc1 Bxd5 Rxe8 Qxe8 Bxd5 d3
00:00:13.5 0.25 18 429517320 Nc3 d5 b4 Bc4 Bf4 Qg8 Re7 Nd8 b5 Ne6 Bxb8 Rfxb8 bxc6 d4 Nd5 Re8 Rc1 Bxd5 Rxe8 Qxe8 Bxd5 d3
00:00:13.7 0.33 19 439234904 Nc3
00:00:13.8 0.25 19 444601929 Nc3 d5
00:00:14.1 0.33 19 462991251 Nc3
00:00:15.5 0.50 19 547520878 Nc3 d5 b4 Nd6 a4 Bc4 Qf4 Ne8 Nd1 Na6 Ba3 c5 bxc5 Nxc5 Bxc5 bxc5 Ne3 Rd8 Nxc4 dxc4 Rac1 Nf6 Qxc4 Qf5 Qxc5 Qxc5
00:00:16.0 0.50 19 572756841 Nc3 d5 b4 Nd6 a4 Bc4 Qf4 Ne8 Nd1 Na6 Ba3 c5 bxc5 Nxc5 Bxc5 bxc5 Ne3 Rd8 Nxc4 dxc4 Rac1 Nf6 Qxc4 Qf5 Qxc5 Qxc5
00:00:16.9 0.52 19 624469329 b4 d5 Nc3 Bc4 Bb2 Nd6 Qf4 Rd8 Rad1 Na6 b5 Nxb5 Nxb5 cxb5 Qxf7 Nc5 Re7 Rg8 Bxd5 Rad8 Be5 Rxd5 Rxd5 Bxd5
00:00:17.1 0.52 19 633891330 b4 d5 Nc3 Bc4 Bb2 Nd6 Qf4 Rd8 Rad1 Na6 b5 Nxb5 Nxb5 cxb5 Qxf7 Nc5 Re7 Rg8 Bxd5 Rad8 Be5 Rxd5 Rxd5 Bxd5
00:00:17.1 0.43 20 638066278 b4 d5
00:00:17.5 0.52 20 662094617 b4
00:00:18.3 0.65 20 703199620 b4
00:00:19.9 0.64 20 794850162 b4 d5 Nc3 Bc4 Bb2 Nd6 Qf4 Rd8 Rad1 Na6 b5 Nxb5 Nxb5 cxb5 Qxf7 Nc5 Bxd5 Rxd5 Rxd5 Rg8 Qf3 Nd3 Rxd3
00:00:20.3 0.64 20 819925466 b4 d5 Nc3 Bc4 Bb2 Nd6 Qf4 Rd8 Rad1 Na6 b5 Nxb5 Nxb5 cxb5 Qxf7 Nc5 Bxd5 Rxd5 Rxd5 Rg8 Qf3 Nd3 Rxd3
00:00:20.6 0.55 21 837648913 b4 d5
00:00:20.7 0.46 21 846552772 b4 d5
00:00:21.0 0.53 21 860357806 b4 d5 Nc3 Bc4 Bb2 Nd6 Qf4 Rd8 Rad1 Na6 b5 Nxb5 Nxb5 cxb5 Re7 Rf8 Rxf7 Kg8 Rxf8+ Rxf8 Qe5 Nc5 Bd4 Ne4 Qe7 Rf7
00:00:21.0 0.53 21 860872478 b4 d5 Nc3 Bc4 Bb2 Nd6 Qf4 Rd8 Rad1 Na6 b5 Nxb5 Nxb5 cxb5 Re7 Rf8 Rxf7 Kg8 Rxf8+ Rxf8 Qe5 Nc5 Bd4 Ne4 Qe7 Rf7
00:00:22.3 0.61 22 936060891 b4
00:00:22.8 0.53 22 963063956 b4 d5
00:00:24.3 0.40 22 1048361161 b4 d5
00:00:25.5 0.33 22 1117126650 b4 d5 Nc3 Bc4 Bf4 Qg8 Rad1 Re8 Bxh6 gxh6 Qd4+ Kh7 Ne4 Nd7 Bh3 Qg7 Bxd7 Qxd4 Rxd4 Rxe4 Bf5+ Kg7 Bxe4 a5 bxa5 Rxa5 Bc2 Rxa2 Rg4+ Kf6 Rf4+ Kg7
00:00:25.7 0.33 22 1127947690 b4 d5 Nc3 Bc4 Bf4 Qg8 Rad1 Re8 Bxh6 gxh6 Qd4+ Kh7 Ne4 Nd7 Bh3 Qg7 Bxd7 Qxd4 Rxd4 Rxe4 Bf5+ Kg7 Bxe4 a5 bxa5 Rxa5 Bc2 Rxa2 Rg4+ Kf6 Rf4+ Kg7
00:00:27.5 0.24 23 1228786852 b4 d5
00:00:31.3 0.15 23 1451081155 b4 d5
00:00:37.2 0.24 23 1790435948 b4
00:00:37.4 0.28 23 1801572934 b4 d5 Nc3 Bc4 Bf4 Qg8 Re7 Nd8 b5 Ne6 Bxb8 Rfxb8 bxc6 d4 Nd5 Re8 Rxe6 Rxe6 Nc7 Be2 Qh4 d3 Nxe6 fxe6
00:00:39.4 0.28 23 1916784261 b4 d5 Nc3 Bc4 Bf4 Qg8 Re7 Nd8 b5 Ne6 Bxb8 Rfxb8 bxc6 d4 Nd5 Re8 Rxe6 Rxe6 Nc7 Be2 Qh4 d3 Nxe6 fxe6
00:00:41.1 0.45 23 2014407388 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 b3 Qf5 Ba3 Rc8 Bb2 Qg5 Re7 Bd3 Rd1 Be4 Bxe4 dxe4 Qd4 Na6 Re5 Qf6 Qxe4 Rd8 Rb1 Rac8 Re7 Nc5 Bxf6 Nxe4 Bb2
00:00:41.2 0.45 23 2015890749 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 b3 Qf5 Ba3 Rc8 Bb2 Qg5 Re7 Bd3 Rd1 Be4 Bxe4 dxe4 Qd4 Na6 Re5 Qf6 Qxe4 Rd8 Rb1 Rac8 Re7 Nc5 Bxf6 Nxe4 Bb2
00:00:41.4 0.36 24 2030868684 Qf4 d5
00:00:41.7 0.30 24 2049775966 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 b3 Qf5 Ba3 Rc8 Bb2 Qg5 Re7 Bd3 Rd1 Be4 Bxe4 dxe4 Qd4 Na6 Re5 Qf6 Qxe4 Rd8 Rxd8+ Qxd8 Qxc6 Nc5 Bd4 Rc8 Rd5 Qf8
00:00:44.2 0.30 24 2147483647 Qf4 d5 Nd6 Nxd6 Qxd6 Kg8 b3 Qf5 Ba3 Rc8 Bb2 Qg5 Re7 Bd3 Rd1 Be4 Bxe4 dxe4 Qd4 Na6 Re5 Qf6 Qxe4 Rd8 Rxd8+ Qxd8 Qxc6 Nc5 Bd4 Rc8 Rd5 Qf8
00:00:44.3 0.21 24 2147483647 Qf4 d5
00:00:44.6 0.21 24 2147483647 Qf4 d5
00:00:45.7 0.33 24 2147483647 b4 d5 Nc3 Bc4 Bb2 Nd6 Qf4 Rd8 Rad1 Na6 b5 Bxb5 Nxd5 cxd5 Rxd5 Bc6 Rxd6 Rxd6 Qxd6 Bxg2 Kxg2 Nc5 Re7 Rf8 Qc6 Ne6 Rxa7 Qb1 Ba3 Qxa2 Qb7 Rd8
00:00:45.8 0.33 24 2147483647 b4 d5 Nc3 Bc4 Bb2 Nd6 Qf4 Rd8 Rad1 Na6 b5 Bxb5 Nxd5 cxd5 Rxd5 Bc6 Rxd6 Rxd6 Qxd6 Bxg2 Kxg2 Nc5 Re7 Rf8 Qc6 Ne6 Rxa7 Qb1 Ba3 Qxa2 Qb7 Rd8
00:00:47.0 0.41 25 2147483647 b4
00:00:48.7 0.36 25 2147483647 b4 d5 Nc3 Bc4 Bb2 Nd6 Qf4 Rd8 Rad1 Na6 b5 Bxb5 Nxd5 cxd5 Rxd5 Bc6 Rxd6 Rxd6 Qxd6 Bxg2 Kxg2 Nc5 Re7 Rf8 Qc6 Ne6 Rxa7 Qd3 Qf3 Qxf3+ Kxf3 Kg8
00:00:48.8 0.36 25 2147483647 b4 d5 Nc3 Bc4 Bb2 Nd6 Qf4 Rd8 Rad1 Na6 b5 Bxb5 Nxd5 cxd5 Rxd5 Bc6 Rxd6 Rxd6 Qxd6 Bxg2 Kxg2 Nc5 Re7 Rf8 Qc6 Ne6 Rxa7 Qd3 Qf3 Qxf3+ Kxf3 Kg8
00:01:01.7 0.45 26 2147483647 b4
00:01:08.2 0.36 26 2147483647 b4 d5
00:01:24.7 0.45 26 2147483647 b4
00:01:30.0 0.64 26 2147483647 b4
00:01:52.5 0.49 26 2147483647 b4 d5
00:01:58.9 0.38 26 2147483647 b4 d5 Nc3 Bc4 Bf4 Qg8 Rad1 Re8 Bxh6 gxh6 Qd4+ Kh7 Ne4 Nd7 Bh3 Qg7 Bxd7 Qxd4 Rxd4 Rxe4 Bf5+ Kh8 Bxe4 Re8 Rc1 Nd6 Bg2 Re5 a3 Kg7 Rg4+ Kh7
00:01:58.9 0.38 26 2147483647 b4 d5 Nc3 Bc4 Bf4 Qg8 Rad1 Re8 Bxh6 gxh6 Qd4+ Kh7 Ne4 Nd7 Bh3 Qg7 Bxd7 Qxd4 Rxd4 Rxe4 Bf5+ Kh8 Bxe4 Re8 Rc1 Nd6 Bg2 Re5 a3 Kg7 Rg4+ Kh7
00:02:02.8 0.46 27 2147483647 b4 d5 Nc3 Bc4 Bf4 Qg8 Rad1 Re8 Bxh6 gxh6 Qd4+ Kh7 Ne4 Nd7 Bh3 Qg7 Bxd7 Qxd4 Rxd4 Rxe4 Bf5+ Kg7 Bxe4 Re8 Rc1 Nd6 Bd3 Re5 Bxc4 Nxc4 Rcd1 a6
00:02:03.2 0.46 27 2147483647 b4 d5 Nc3 Bc4 Bf4 Qg8 Rad1 Re8 Bxh6 gxh6 Qd4+ Kh7 Ne4 Nd7 Bh3 Qg7 Bxd7 Qxd4 Rxd4 Rxe4 Bf5+ Kg7 Bxe4 Re8 Rc1 Nd6 Bd3 Re5 Bxc4 Nxc4 Rcd1 a6
00:02:11.2 0.55 28 2147483647 b4
00:02:14.6 0.46 28 2147483647 b4 d5
00:02:24.4 0.55 28 2147483647 b4
00:02:29.1 0.50 28 2147483647 b4 d5 Nd2 Bd3 Bb2 Nd8 Re7 Bf5 Qe2 Nd7 Nb3 a5 g4 Be6 f4 axb4 f5 Nc5 fxe6 Ndxe6 Nd4 Nxd4 Bxd4 b3 axb3 Rxa1+ Bxa1 Qb1+ Qf1 Qxb3 Bd4 Qa4
00:02:29.3 0.50 28 2147483647 b4 d5 Nd2 Bd3 Bb2 Nd8 Re7 Bf5 Qe2 Nd7 Nb3 a5 g4 Be6 f4 axb4 f5 Nc5 fxe6 Ndxe6 Nd4 Nxd4 Bxd4 b3 axb3 Rxa1+ Bxa1 Qb1+ Qf1 Qxb3 Bd4 Qa4
00:02:38.8 0.59 29 2147483647 b4
00:02:45.1 0.50 29 2147483647 b4 d5
00:02:53.5 0.59 29 2147483647 b4
00:03:01.7 0.48 29 2147483647 b4 d5
00:04:11.6 0.46 29 2147483647 b4 d5 Nd2 Bd3 Bb2 a5 Re7 Nd8 b5 Bxb5 Nf3 Bd3 Nh4 Na6 Ng6+ fxg6 Rxg7 Qxg7 Bxg7+ Kxg7 Qd4+ Rf6 Qxd3 Nc5 Qb1 Rb8 hxg6 Rb7 Qb2 Kxg6 f4 Nf7
00:04:13.1 0.46 29 2147483647 b4 d5 Nd2 Bd3 Bb2 a5 Re7 Nd8 b5 Bxb5 Nf3 Bd3 Nh4 Na6 Ng6+ fxg6 Rxg7 Qxg7 Bxg7+ Kxg7 Qd4+ Rf6 Qxd3 Nc5 Qb1 Rb8 hxg6 Rb7 Qb2 Kxg6 f4 Nf7
00:04:14.4 0.55 30 2147483647 b4
00:04:24.6 0.62 30 2147483647 b4 d5 Nd2 Bd3 Bb2 Nd8 Nf3 Bf5 Qh4 Ne6 Ne5 Ng5 f4 Ne4 Qe7 Nd7 Bxe4 dxe4 Nxd7 Rfe8 Qxf7 Be6 Bxg7+ Qxg7 Qxg7+ Kxg7 Ne5 a5 a3 axb4 axb4 c5
00:04:25.4 0.62 30 2147483647 b4 d5 Nd2 Bd3 Bb2 Nd8 Nf3 Bf5 Qh4 Ne6 Ne5 Ng5 f4 Ne4 Qe7 Nd7 Bxe4 dxe4 Nxd7 Rfe8 Qxf7 Be6 Bxg7+ Qxg7 Qxg7+ Kxg7 Ne5 a5 a3 axb4 axb4 c5
00:04:43.9 0.61 30 2147483647 b4 d5 Nd2 Bd3 Bb2 Nd8 Nf3 Bf5 Qh4 Ne6 Ne5 Ng5 f4 Ne4 Qe7 Nd7 Bxe4 dxe4 Nxd7 Rfe8 Qxf7 Be6 Bxg7+ Qxg7 Qxg7+ Kxg7 Ne5 a5 a3 Bd5 Kf2 b5
00:04:45.4 0.61 30 2147483647 b4 d5 Nd2 Bd3 Bb2 Nd8 Nf3 Bf5 Qh4 Ne6 Ne5 Ng5 f4 Ne4 Qe7 Nd7 Bxe4 dxe4 Nxd7 Rfe8 Qxf7 Be6 Bxg7+ Qxg7 Qxg7+ Kxg7 Ne5 a5 a3 Bd5 Kf2 b5
00:05:16.4 0.68 31 2147483647 b4
00:05:26.5 0.61 31 2147483647 b4 d5
00:06:30.5 0.68 31 2147483647 b4
00:07:08.7 0.58 31 2147483647 b4 d5
00:10:57.0 0.37 31 2147483647 b4 d5 Nd2 Bd3 Bb2 Nd8 Nf3 Ne6 Ne5 Bf5 Qh4 Ng5 f4 Ne4 Qe7 Nd7 Nxf7+ Kg8 Nd6 Ndf6 Bxf6 Nxf6 Qe5 Bd7 f5 Rae8 Nxe8 Rxe8 Qc7 Qxf5 Rxe8+ Bxe8
00:13:31.4 0.37 31 2147483647 b4 d5 Nd2 Bd3 Bb2 Nd8 Nf3 Ne6 Ne5 Bf5 Qh4 Ng5 f4 Ne4 Qe7 Nd7 Nxf7+ Kg8 Nd6 Ndf6 Bxf6 Nxf6 Qe5 Bd7 f5 Rae8 Nxe8 Rxe8 Qc7 Qxf5 Rxe8+ Bxe8
00:13:31.7 0.37 31 2147483647 b4 d5 Nd2 Bd3 Bb2 Nd8 Nf3 Ne6 Ne5 Bf5 Qh4 Ng5 f4 Ne4 Qe7 Nd7 Nxf7+ Kg8 Nd6 Ndf6 Bxf6 Nxf6 Qe5 Bd7 f5 Rae8 Nxe8 Rxe8 Qc7 Qxf5 Rxe8+ Bxe8
00:14:22.2 0.99 31 2147483647 Bg5
|