Podbot Beta 3.0 Updates? ^^

[CS1.6]

Extensive report on the new 'Self-Protection instinct' / 'revenge AI' feature:

The "revenge" or "self-protect" instinct ( KWo ..what is the name for your invention? ) is very advanced.

• You realize very soon the aim deviation of a BOT in "revenge" mode is very high in a totally obvious fashion. He clearly shoots 85% at walls, objects, etc. around you ..but unmistakenly in your direction ..as a warning. While if he really wanted he could headshot you easily. The team-attacker unmistakenly realizes the Player/Bot is shooting at him ..even before the bot has actually hit him. This can be observed often. So the visual BOT feedback ("stop shooting at me") is working & efficient.

• Actually opposed to false reports..the BOTs do not take revenge the next round. New round, new game, new beginning, peace! (Have not tested for CSDM though. Since in CSDM there are no rounds but rather new spawns. I don't know if BOTs make peace again at new round or new spawn. The latter would be more ideal and cover both game modes (cs, csdm)

• One thing i noticed: while at the beginning (after installation of the new build with this feature) the BOTs (#466) reacted to team attacks by a (very intelligent &) more subtle gesture (of initially just aiming at you and watching your next move) ..the BOTs do not take these precautious steps anymore after a while. If you shoot at them they will show no mercy and will react by shooting at you (without waiting for you to shoot at them a second time). I am pretty sure this has something to do with podbots evolving AI after a while (through the .pxp files i guess). Podbot generally allways tends to become merciless as it gathers experience. Put the lowest skill bots on a server and let it play against aimbotters and after a while the bots will play themselves like aimbotters with an insane aiming. It would be nice to preserve the BOTs initial state of precautionary approach policy after the first team attack incident. If after a while (10 seconds) no repeated team attack occures..the BOT should consider it a non-malicious team-attack and stop aiming at the offender ..and let go ..and move on to his bot goals.

• As a user of podbot since 2006 (?) i can say that this feature was present (in its weakest form) in podbot in (some?) of the previous builds. I recall strongly that after allmost abusing the team-bots by unloading dozens & dozens of bullets on them, they would react (although rarely) by shooting one or two bullets towards you. Now if that was a malfunction or bug or really an allready existent primitive form of this AI feature.. i don't know.

This feature really keeps me entertained. ..the new bots make you fear & respect them. It's a total new face of podbot. They used to be dull ragdolls with whom you could do all kind of stupid shizt. Now they have climbed up the respect ladder in the CS game world.

[senecas]

Kwo //

command
./hlds_run -game cstrike -port 27037 -pingboost 2 +map de_dust2_cz +maxplayers 24 +ip myip +exec server.cfg

meta list
Currently loaded plugins:
description stat pend file vers src load unlod
[ 1] POD-Bot mm RUN - podbot_mm_i386.s vV3B21 ini Chlvl ANY
[ 2] AMX Mod X RUN - amxmodx_mm_i386. v1.8.1.3 ini Start ANY
[ 3] FakeMeta RUN - fakemeta_amxx_i3 v1.8.1.3 pl2 ANY ANY
[ 4] Engine RUN - engine_amxx_i386 v1.8.1.3 pl2 ANY ANY
[ 5] MySQL RUN - mysql_amxx_i386. v1.8.1.3 pl2 ANY ANY
[ 6] CSX RUN - csx_amxx_i386.so v1.8.1.3 pl2 ANY ANY
[ 7] CStrike RUN - cstrike_amxx_i38 v1.8.1.3 pl2 ANY ANY
[ 8] Ham Sandwich RUN - hamsandwich_amxx v1.8.1.3 pl2 ANY ANY
[ 9] CSDM2 RUN - csdm_amxx_i386.s v2.1.2k pl2 ANY ANY
9 plugins, 9 running

hum...last 3 days was stable.
but yesterday and today -> crashed 1 time

[wickedd]

I run a few zombie swarm servers (2.4) and I just updated to podbot v3B21. Since the update I notice that the bots are not as smart then before. The bots now attack each other haha T's (remember it's a zs server) they seem to just stand around and if a CT stand still the bot will not attack. I don't know if it's the waypoints (I doubt it ) or maybe the bots themselves. All I know is my players hate them, they complain all the time about the bots. In the podbot.cfg I have pb_minbotskill 99 and pb_maxbotskill 100 and the bots still suck. Kwo I know you put alot of time and effort into this and I appreciate it, please don't think I'm bashing your bots. I love podbots that's why I use them. If you (KWo or Ancient lazy ass ) pm me I will give you my server ip so you can checkout and see if its the waypoints or the bots. Also if you guys need a server for testing let let me I'll give you one.

[KWo]

Update the zombie swarm to V3 beta 7d. It's called beta, but only because of still added functions they maybe aren't fully checked, but it works stable (without any crash) on a lot of servers.
http://forums.alliedmods.net/showthread.php?t=29626

[Julian de L. Lopes]

quote: "Put the lowest skill bots on a server and let it play against aimbotters and after a while the bots will play themselves like aimbotters with an insane aiming"

someone can explain it for me?
Why Lowest skilled ones?
why development of the bot aiming?

[KWo]

Update - V3B21a:

+Added the code to prevent bots looking up or down for no reason (need more testing to let me write here it's fixed).
+Added the new cvar - pb_firsthumanrestart (0 stops the function added in V3B21 - when the first human joins the server and there are some bots already playing there - the round is restarted). Now this function is off by default. You can switch it to on by changing the value from 0 to 1.
+Added code to let the bot reload the weapon at round end (if the bot doesn't see any enemy that time).
+Fixed - the vip bot shouldn't hunt the enemy anymore.

As usually, please, leave a feedback.

[Serious Sam]

I spectated a 5v5 bot match on de_dust2 with difficulty 60-80 for about 30 minutes (I'll do more later) and I didn't see them shoot in random direction before they see an enemy, which is good, but they do it after they see the enemy when they stop seeing it, for example at the doors, a bot passing by the door sees an enemy through the hole, starts shooting at it, moves some more and cant see it anymore, then I guess he tries to shoot it through the door but shoots at an almost random dirrection - floor, ceiling, etc... also, for all these 30 minutes of testing the terrorists planted over 8 bombs, and the CTs did not try to defuse any of them...

EDIT: They defused it once just now, right before it was going to blow. So they do try but they are not in much of a hurry ^^

[Serious Sam]

For some reason the edit button doesn't show up right now, so I have to bump.

While looking through the code, in bot_globals.cpp, weapon_maxClip array, there is ... 10, // tmp ... - I believe it should be 30 ^^

I also want to ask about this, every time I load a map, I get this message in the console

Visibility Table file (pvi) exists and is newer than waypoint file (pwf).
No Visibility Table File or old one - starting new

for some reason the bot starts rebuilding the visibility table each time, although it approves the current one... weird...

[KWo]

Originally Posted by Serious Sam

While looking through the code, in bot_globals.cpp, weapon_maxClip array, there is ... 10, // tmp ... - I believe it should be 30 ^^

It is so since the beginning (since Count Floyd wrote it so). I didn't check those values, but it looks like it needs to be re-checked.

Originally Posted by Serious Sam

I also want to ask about this, every time I load a map, I get this message in the console

Visibility Table file (pvi) exists and is newer than waypoint file (pwf).
No Visibility Table File or old one - starting new

for some reason the bot starts rebuilding the visibility table each time, although it approves the current one... weird...

Do You mean - does it happen for all maps all the time? Normally - when the round is ended - and the new visibility table was created (You could see the messages about building the visibility table) - the pvi file is saved and the next time - when You load this map again - it shouldn't try to create the new visibility table - unless You changed waypoints (pwf file).

[EDIT]
I'm adding into the code more diagnostic about that issue with pvi files, so it would tell why - even if the file is pre-approved - then - is not loaded. That will be in V3B21b (but for this version I have to add something yet for CTs - to make them more hurry in defusing the bomb).
[/EDIT]

[Serious Sam]

So, it looks like the pvi file is generated immediately after analyzing to 100%, but data is compressed and stored in it a few minutes later, so if I quit the game before that, the file exists showing the first message, but it still does show the second and start building it again, as a quick fix I wont quit before I see that the data is compressed and stored, but I see no reason for these few minutes of delay, I think that should be done right after analyzing is finished.