Welcome Guest! To enable all features please Login or Register.

Notification

Icon
Error
84 Pages«<4344454647>»
Sepefeets Silverfish update (OG, Kara, MSG)
Options
Previous Topic Next Topic
sepefeets
#890 Posted : Thursday, September 1, 2016 9:53:46 PM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 899

Thanks: 1028 times
Was thanked: 584 time(s) in 238 post(s)
p0p0;40859 wrote:
Hello, reporting a play that might be incorrect.

Log file: http://hastebin.com/ayexebizen.coffee

Screenshot of the situation: http://i.imgur.com/oA7IpnS.jpg

Situation: The botplayer should have attacked the enemy minion (6/6 Ancient Shieldbearer) with Tirion, killing it breaking only tirion's divine shield. The botplayer ended up playing coghammer and buffing Tirion with nothing since he hadn't used his divine shield. After that, I think the botplayer should have traded a minion 1/1 with the 6/1 ancient shieldbearer (the 6/1 ancient shieldbearer ended up attacking the 6/6 Tirion with divine shield).


WOW! There is a lot going on there. It had 10 moves to do and looks like it was trying to use juggles to get the job done while preserving the shield and the reason the shieldbearer lived at all is because it roped out the turn. I'm not sure if either play is better than the other in a bubble though.

The risk of wasting coghammer buff on tirion with that many minions on board seems worth the risk to get 2x divine shield taunts.

I think I would need a lot of feedback on when it should and shouldn't attack divine shield into minions to do anything about it.
Back to top
WWW
1 user thanked sepefeets for this useful post.
p0p0 on 9/2/2016(UTC)
p0p0
#891 Posted : Friday, September 2, 2016 2:32:06 AM(UTC)
Rank: Member

Groups: Registered
Posts: 29
Location: Paris

Thanks: 7 times
Was thanked: 11 time(s) in 11 post(s)
sepefeets;40862 wrote:
WOW! There is a lot going on there. It had 10 moves to do and looks like it was trying to use juggles to get the job done while preserving the shield and the reason the shieldbearer lived at all is because it roped out the turn. I'm not sure if either play is better than the other in a bubble though.

The risk of wasting coghammer buff on tirion with that many minions on board seems worth the risk to get 2x divine shield taunts.

I think I would need a lot of feedback on when it should and shouldn't attack divine shield into minions to do anything about it.


I had 30 health and card draw, meaning N'zoth soon in my hand. I think Tirion's Divine shield absorbind 6 damage would be worth it .. :o
Back to top
WWW
1 user thanked p0p0 for this useful post.
sepefeets on 9/2/2016(UTC)
sepefeets
#892 Posted : Friday, September 2, 2016 4:31:25 AM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 899

Thanks: 1028 times
Was thanked: 584 time(s) in 238 post(s)
p0p0;40866 wrote:
I had 30 health and card draw, meaning N'zoth soon in my hand. I think Tirion's Divine shield absorbind 6 damage would be worth it .. :o

It looked like you were so far ahead that nothing really mattered, not even roping out half the turn... But still, even if I did change it, I have no idea what specifically to change and doing anything would most likely make more problems without knowing exactly what to do about it. For instance, do I just make tirion trade his divine shield into a big minion? Should other divine shield minions make the trades too? Does the trade need to kill the minion? What if attacking face with tirion sets up lethal next turn if he survives because we kept the shield and it keeps him alive? On a separate note, should we try to get rid of divine shields before doing something that applies one randomly? Should we still get rid of them if we also have something that gets more value per shield?

In a complex decision like this I need to know every reason (many that you don't consciously think of) for why it is the correct move in order to program it correctly. If you or anyone else can come up with a list of exactly when to do it then I'll change it.
Back to top
WWW
1 user thanked sepefeets for this useful post.
p0p0 on 9/2/2016(UTC)
sepefeets
#894 Posted : Friday, September 2, 2016 8:11:01 AM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 899

Thanks: 1028 times
Was thanked: 584 time(s) in 238 post(s)
If anyone has played Steward of Darkshire recently or would like to help by running a steward deck for a day then please check logs for turns where it played 1hp minions before the steward (search for "playcard: Steward of Darkshire" to find turns where it is played, then look for things played before it). I can't reproduce it in sims with the logs I have even with multiple older versions of SF.
Back to top
WWW
 |  Edit by user
1 user thanked sepefeets for this useful post.
Somerandombot007 on 9/2/2016(UTC)
sepefeets
#895 Posted : Friday, September 2, 2016 10:31:13 PM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 899

Thanks: 1028 times
Was thanked: 584 time(s) in 238 post(s)
I'm grinding through lots of old logs but it's like 1/3 already fixed through side effects of other fixes, 1/3 can't reproduce even with older versions of SF, and 1/3 data sync that I didn't notice before. It's getting incredibly boring not finding anything to actually code.
Back to top
WWW
sepefeets
#896 Posted : Saturday, September 3, 2016 9:23:32 AM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 899

Thanks: 1028 times
Was thanked: 584 time(s) in 238 post(s)
V123.3SE released.

allow enemy sim to target heroes with hero powers so that priests actually heal themselves instead of causing dumb 2 turn lethal misplays
output all minion attributes for test.txt sims to see things like divine shield too
penalize buffing attack after attacking with all minions too
change turn 1 bonuses to keep evaluating other penalties
[883,885.2] slightly penalize sir finley turn 1-2
simplify 1hp minion fireblast next turn board value bonus
use a background thread to launch silver.exe so that HR isn't delayed by it
log win/loss in silverlogs
Back to top
WWW
3 users thanked sepefeets for this useful post.
Dre on 9/3/2016(UTC), Somerandombot007 on 9/3/2016(UTC), mary7 on 9/4/2016(UTC)
Dre
#898 Posted : Saturday, September 3, 2016 11:31:16 AM(UTC)

Rank: Advanced Member

Groups: HSModerator, Registered
Posts: 888
Man
Location: The Netherlands

Thanks: 216 times
Was thanked: 480 time(s) in 315 post(s)
http://pastebin.com/AJgyti2v

See last turn, why would it hit the Barnes with a 1 AT weapon? Oponent had leathal when i took the facedamage. No reason at all to hit it there.
Back to top
WWW
1 user thanked Dre for this useful post.
sepefeets on 9/3/2016(UTC)
Cyberpolice
#899 Posted : Saturday, September 3, 2016 4:03:02 PM(UTC)
Rank: Newbie

Groups: Registered
Posts: 3

Thanks: 2 times
Was thanked: 2 time(s) in 2 post(s)
Has anyone else found that quite often the bot will try and attack twice with each minion. I've noticed this with my mid-range hunter. It's causing me to lose games simply because the bot ran out of time and did not perform all attacks when I have a few minions on the field. Any fixes?
Back to top
WWW
1 user thanked Cyberpolice for this useful post.
sepefeets on 9/3/2016(UTC)
Jadax
#901 Posted : Saturday, September 3, 2016 5:47:57 PM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 42

Thanks: 2 times
Was thanked: 7 time(s) in 7 post(s)
Cyberpolice;40920 wrote:
Has anyone else found that quite often the bot will try and attack twice with each minion. I've noticed this with my mid-range hunter. It's causing me to lose games simply because the bot ran out of time and did not perform all attacks when I have a few minions on the field. Any fixes?


Yep, happens quite a lot for me too - I initially assuméd it was my connection, but seems to be the same using the uni connection too.
Back to top
WWW
1 user thanked Jadax for this useful post.
sepefeets on 9/3/2016(UTC)
sepefeets
#900 Posted : Saturday, September 3, 2016 7:53:50 PM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 899

Thanks: 1028 times
Was thanked: 584 time(s) in 238 post(s)
Cyberpolice;40920 wrote:
Has anyone else found that quite often the bot will try and attack twice with each minion. I've noticed this with my mid-range hunter. It's causing me to lose games simply because the bot ran out of time and did not perform all attacks when I have a few minions on the field. Any fixes?

It's data sync, HR says the minion can still attack after it already did.
Back to top
WWW
1 user thanked sepefeets for this useful post.
Cyberpolice on 9/5/2016(UTC)
clee
#902 Posted : Sunday, September 4, 2016 12:43:44 AM(UTC)
Rank: Newbie

Groups: Registered
Posts: 8

Thanks: 1 times
Was thanked: 9 time(s) in 6 post(s)
@sepefeets : I have observed how the log window in HR does not stop when Silverfish is called, this means that HR does not block when calling the external AI module. It is probably an event call to the external functions. In this case if game state update events are called when the silverfish dll is processing certain things (in OnQueryNextMove), it might miss that event call for game state update. Now that the dll is writing even more things to the external file/Network (IO events are the slowest), more game state events might be missed. I would suggest generating less file output. The best would be to send a simple JSON package of the state variables to the external process and let the external process update the internal variables. I can see that there are a lot of logic embedded in the updateeverything call which should not be the case.
Back to top
WWW
1 user thanked clee for this useful post.
sepefeets on 9/4/2016(UTC)
sepefeets
#903 Posted : Sunday, September 4, 2016 2:38:07 AM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 899

Thanks: 1028 times
Was thanked: 584 time(s) in 238 post(s)
clee;40927 wrote:
@sepefeets : I have observed how the log window in HR does not stop when Silverfish is called, this means that HR does not block when calling the external AI module. It is probably an event call to the external functions. In this case if game state update events are called when the silverfish dll is processing certain things (in OnQueryNextMove), it might miss that event call for game state update. Now that the dll is writing even more things to the external file/Network (IO events are the slowest), more game state events might be missed. I would suggest generating less file output. The best would be to send a simple JSON package of the state variables to the external process and let the external process update the internal variables. I can see that there are a lot of logic embedded in the updateeverything call which should not be the case.

I don't know about your reasoning for it and JSON is more data not less, but I've known for a while that playfields should be sent serialized in some way, see PM.

I need code more than I need suggestions though.
Back to top
WWW
Dre
#897 Posted : Sunday, September 4, 2016 4:03:14 AM(UTC)

Rank: Advanced Member

Groups: HSModerator, Registered
Posts: 888
Man
Location: The Netherlands

Thanks: 216 times
Was thanked: 480 time(s) in 315 post(s)
sepefeets;40905 wrote:
V123.3SE released.

allow enemy sim to target heroes with hero powers so that priests actually heal themselves instead of causing dumb 2 turn lethal misplays
output all minion attributes for test.txt sims to see things like divine shield too
penalize buffing attack after attacking with all minions too
change turn 1 bonuses to keep evaluating other penalties
[883,885.2] slightly penalize sir finley turn 1-2
simplify 1hp minion fireblast next turn board value bonus
use a background thread to launch silver.exe so that HR isn't delayed by it
log win/loss in silverlogs


Thnx btw! and damn the win/loss report is nice, now i just filter out the logs with the text "Match lost :(" in it with Agent Ransack software so i just look at the logs where i could turn a loss into a win instead of going trough it all :)
Back to top
WWW
1 user thanked Dre for this useful post.
sepefeets on 9/4/2016(UTC)
charis2324
#904 Posted : Sunday, September 4, 2016 4:04:42 AM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 76
Location: Hong Kong

Thanks: 6 times
Was thanked: 22 time(s) in 20 post(s)
SF use 2 Rockbiter Weapon in a role and kill a 0/1 frog.

Log(hope I get it right)
http://pastebin.com/dfiPmK3R
Back to top
WWW
1 user thanked charis2324 for this useful post.
sepefeets on 9/4/2016(UTC)
Dre
#905 Posted : Sunday, September 4, 2016 4:14:04 AM(UTC)

Rank: Advanced Member

Groups: HSModerator, Registered
Posts: 888
Man
Location: The Netherlands

Thanks: 216 times
Was thanked: 480 time(s) in 315 post(s)
charis2324;40936 wrote:
SF use 2 Rockbiter Weapon in a role and kill a 0/1 frog.

Log(hope I get it right)
http://pastebin.com/dfiPmK3R


Sync-problem - it didnt register that it put rockbiter on itself - Its a known HR problem
Back to top
WWW
1 user thanked Dre for this useful post.
sepefeets on 9/4/2016(UTC)
secret3
#906 Posted : Sunday, September 4, 2016 11:36:32 AM(UTC)
Rank: Newbie

Groups: Registered
Posts: 6

Thanks: 1 times
Was thanked: 1 time(s) in 1 post(s)
Hello, i'v updated to the xxx.3SE version and i'm currently using the shaman deck by dre,

i was monitoring the bot today and i saw some mistake he did

here a screen shot and the log from the match

nasty bot


tldr, i was controling the board after using doom hamer, only blood mage thanos was on the board from opponent and my only card who didnt attacked yet was my argent squire girl, but insteat of killing blood mage thanos who is +1 spell she just faced the mage,

the next turn they(all my card) keep attacking face, insteat of killing the +1 spell monster, that could have cost me the game if the mage used arcane epxlosion , etc

here the log log here

i'm using the new muli/combo from dre, the one he updated the 03.09, so i dont know if the issue is from the A.I or from the settings.txt
Back to top
WWW
 |  Edit by user
1 user thanked secret3 for this useful post.
sepefeets on 9/4/2016(UTC)
rffc
#909 Posted : Sunday, September 4, 2016 12:14:41 PM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 33

Thanks: 5 times
Was thanked: 10 time(s) in 8 post(s)
Hy,
here are some misplays of my last session,
im running settings on behaviour=control and playaroundsecrets=true

---------- 1 ----------

http://pastebin.com/37kG0mNq



---------- 2 ----------

http://pastebin.com/4hJNzdSq



---------- 3 ----------

http://pastebin.com/NQ2FLYMR



---------- 4 ----------

http://pastebin.com/H6MRVF6V



---------- 5 ----------

http://pastebin.com/kHXkTUyf
Back to top
WWW
 |  Edit by user
1 user thanked rffc for this useful post.
sepefeets on 9/4/2016(UTC)
sepefeets
#910 Posted : Sunday, September 4, 2016 6:02:16 PM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 899

Thanks: 1028 times
Was thanked: 584 time(s) in 238 post(s)
@secret3 - Aggro shaman is an aggro deck, it's supposed to go face most of the time. You win by bursting them down, not by controlling the board.

@rffc - I need some idea of what you think is wrong in those logs.
Back to top
WWW
rffc
#911 Posted : Monday, September 5, 2016 2:07:25 AM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 33

Thanks: 5 times
Was thanked: 10 time(s) in 8 post(s)
sepefeets;40957 wrote:
@secret3 - Aggro shaman is an aggro deck, it's supposed to go face most of the time. You win by bursting them down, not by controlling the board.

@rffc - I need some idea of what you think is wrong in those logs.



on the logs ? i rly have no idea of the logs to be honest.

I daily watching the replays of my new decks and that were the misplays i could see, so i did copy the logs on the turn the plays have done, in the hope that it's helpful for you guys.

well,

so is there a chance to change this plays by myself ? or is there no chance to improve this...

Kinda like on my first picture, the bot played the huffer before testing the explosion trap and so also got killed ...

best regards
Back to top
WWW
 |  Edit by user
1 user thanked rffc for this useful post.
sepefeets on 9/5/2016(UTC)
sepefeets
#912 Posted : Monday, September 5, 2016 4:10:36 AM(UTC)

Rank: Advanced Member

Groups: Registered
Posts: 899

Thanks: 1028 times
Was thanked: 584 time(s) in 238 post(s)
rffc;40964 wrote:
on the logs ? i rly have no idea of the logs to be honest.

I daily watching the replays of my new decks and that were the misplays i could see, so i did copy the logs on the turn the plays have done, in the hope that it's helpful for you guys.

well,

so is there a chance to change this plays by myself ? or is there no chance to improve this...

Kinda like on my first picture, the bot played the huffer before testing the explosion trap and so also got killed ...

best regards

Yeah I even say in the OP you should rename the log to include a brief description so that you can remember and tell me what happened when you do post it. If I know what to look for then I can try to do something about it but I'm not going to stare at logs and guess what might be a problem when there might not even be one. Since you saw it happen and even edited the logs down to the turn (which I don't care about), then you can probably remember what happened by looking through them.
Back to top
WWW
1 user thanked sepefeets for this useful post.
rffc on 9/5/2016(UTC)
84 Pages«<4344454647>»
Forum Jump  
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.

Powered by YAF | YAF © 2003-2011, Yet Another Forum.NET
This page was generated in 0.172 seconds.