Patch 2.6.4 LoN Fire Condemn for Botting?

Get your trial key

Live community: https://discord.gg/s39XdMV

New user ? Check guides​ or ask the discord bot

hercle
Joined: 04/01/2017 - 08:15
Mon, 11/26/2018 - 09:06 #1

I have a test build published here:

https://ptr.d3planner.com/106256436

With the 65 times LoN boost in Patch 2.6.4, as compared to the previous Akkhan Fire Condemn build, this new build will improve Comdemn damage by 35.34% for elites and 78.65% for whites.

Note that In-Geom is cubed for maximizing the up-time of Akarat's Champion (Akarat's Champion will often reset upon an elite pack is killed so the actual up-time should be close to 100%). It also guarantees 100% up-time for Speed Charge, overcoming the removal of LoH: Wings of Angels for movement speed enhancement. Provoke: Hit Me also has 100% up-time, which will benefit resource generation and, in coupled with Punish: Fury, provide increased block (52.5% total; for toughness) and critical (15% in the next attack; for damage) chances.

As for resource managment, LoV: Unstoppable Force is used instead to compensate the loss of 50% resource cost reduction provided by the 2pcs Akkhan. The combination of this law with Provoke and the Righteousness passive in the Hellfire Amulet, as well as the legenadary power of Cindercoat and Hexing Pants, will likely maintain the resource level close to full. This will meet the 90% resouce level requirement for the damage reduction provided by the cubed Aquila Cuirass. The overall toughness level should surpass the previous Akkhan build.

Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Mon, 11/26/2018 - 10:23

Two glaring issues are: Our standard condemn relies on the residual whites to produce the numbers it has.
With Steed Charge we are not able to run and gun as Steed Charge will be cancelled upon casting condemn.
It would leave us with Unleashed rune. That's not ideal, but can work as it works with our "normal" way of running Condemn.

Other issue being - no guaranteed 100% uptime. Biggest issue at RGs or any extended period of no-elites.

As soon as we lose Champion - we can be killed, unless we find viable options for damage reduction.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

burgerking
DONATOR
Joined: 05/14/2014 - 01:14
Mon, 11/26/2018 - 10:31

Cyp already said it in his last 2 lines...

RG will definitely kill it easily. BOT would then have to do an EXP Pool search afterwards and instead of doing 1R2GR, it would be doing 1R1GR or even 2R1GR for a huge EXP/Hr loss.



 

hercle
Joined: 04/01/2017 - 08:15
Mon, 11/26/2018 - 10:31

My current Akhhan Fire Condemn build does not kill whites (i.e., elite only). Killing whites should slow down the progress for GR levels > 90 since you cannot kill them in one shot. Below are the botting performance of my current Akkhan builds (2.6.1) at Paragon level 3.5K and GR level 99:

Furnace+BofP/Cube+Steed Charge: 570KB/Hr (~9.6 games/hr)
Furnace+BofP/Cube+Provoke: 520KB/Hr (~8.9 games/hr)
BofP+In-Geom/Cube+Steed Charge: 530KB/Hr (~9.0 games/hr)

RG killing time is around 30sec right bow. With In-geom, at least the first 20sec will have Akarat's Champion on for the most of the time. For the LoN Build 2.6.4, I expect it won't make it longer with even 60% up-time for Akarat's Champion. Toughness should not be an issue during the AC cooldown time with the cubed Aquila Cuirass. The Pauldrons of Skeleton King has been replaced by Spaulders of Zakara so occasional deaths should not affect the botting performance much. I also have adjusted Paragon in the test build to 3.5K (2.2K for Strength; 600 for Vitality) for a further comparison: 

For Paragon 3.5K:

LoN 2.6.4 with AC on:
Toughness = 858,728,335
Condemn Elite Damage (at CoE Fire) = 6,872.6B

LoN 2.6.4 with AC off:
Toughness = 407,513,779
Condemn Elite Damage (at CoE Fire) = 5,090.8B

Akkhan:
Toughness = 461,307,555
Condemn Elite Damage (at CoE Fire) = 5,261.4B

Most importantly, the builds under discussion are not for GR level pushing, but for Paragon/XP farming. For this purpose, 4-5 minutes per rift run should be optimal. The XP gained at each rift close contributes significantly to the overall gaining. Elite killing is essential thus the Nemesis Bracers must be equiped. This will exclude some builds that have been boosted significantly in 2.6.4 but cannot incorporate the Nemesis Bracers (e.g., Inna WOL Monk).


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Wed, 11/28/2018 - 11:49

I should have worded it better. We rely on residual whites to produce best experience per hour.
Your presented numbers are god awful for a paragon 3.5k toon - no offense. Those eph numbers that Paragon 1800-2200 users find good at 1r2gr.

There are many ways to bot a build, but I assumed we are trying to maximise results. As that is the point of running Condemn.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

hercle
Joined: 04/01/2017 - 08:15
Sat, 12/01/2018 - 10:02 (Reply to #5)

XP/Hr is affected by botting environment and gear setup in addition to Paragon levels. It doesn't make sense to compare different builds if these factors are not controlled. My results are based on the testing of over 100 runs in each condition, not just a few runs that can give 650+KB XP transiently.

Botting envronment:
Guest Specs: VMWare Workstation Pro 15, 4 virtual processors, 4GB RAM, Windows 10 Pro 64 bit, 800x600 Windowed FullScreen. Peak FPS ~ 50
Host Specs: Z170,  i7-6700K (no overclock) , 16GB RAM, GTX 1080, Windows 10 Pro 64 bit
Gear setup: https://www.d3planner.com/933047542

Botting GR level = 99
Furnace+BofP/Cube+Steed Charge (Kill Elites Only): 570KB/Hr (~9.6 games/hr)
Furnace+BofP/Cube+Provoke (Kill Elites Only): 520KB/Hr (~8.9 games/hr)
Furnace+BofP/Cube+Provoke (Kill Elites & Dense Mobs) : 570KB/Hr (~9.2 games/hr)

FastMode Settings:
Kill Elites Only:
Density limit: 20
Scan range: 40
Elite weight: 20
Minion weight: 0
Goblin weight: 20
Normal monster weight: 0
Warden weight: 20
Life limit: 30
Minimum duration: 500ms
Tolerance: 0
Stuck time detection: 500ms
Kill Elites & Dense Mobs:
Density limit: 20
Scan range: 40
Elite weight: 20
Minion weight: 20
Goblin weight: 20
Normal monster weight: 2
Warden weight: 20
Life limit: 60
Minimum duration: 2000ms
Tolerance: 0
Stuck time detection: 500ms

Scripts:
Kill Elites Only:
https://www.ros-bot.com/custom-script/killelite-1402089
Kill Elites & Dense Mobs:
https://www.ros-bot.com/custom-script/killall-1401677

The main difference between the two scripts is the density setting of the Provoke and Condemn buffs (Normal monster weight = 0 or 1). With normal monster killing enabled (more than 80% killed), the Provoke build does have improved performance. Games/Hr is higher than killing elites only but still lower than the Steed Charge build. XP gained by killing normal monsters approximately compensates the lowered Games/Hr. Therefore, the two builds basically give similar XP/hr. But the Steed Charge build has higher legendary gaining due to the increased Games/Hr.

The two builds have toughness = 426M with AC activated. At GR = 99, death rate is about 1 per 4 runs. More than 50% of deaths are caused by the stuttering/freezes of D3. Botting at GR >= 100 will have significantly lower Games/Hr due to increased monster HP as well as higher death rates, thus giving lowered performance.


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Sun, 12/02/2018 - 05:45

You are assuming that I talk of a few runs.
I have ran over a year of condemn sader. 8+ months were spent testing and perfecting VSSWINTA.
During which I have gotten feedback and results from different paragon levels and gear levels.
I do feel qualified to say that your elite only idea is fruitless IF best experience is in mind.
Now, but you seem adamant on your results so I assume you are just talking of this one specific scenario and are not actually after best results.
In this case - carry on.

Tldr: You can bot a non-optimal build.
You could bot a WD if you like.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

hercle
Joined: 04/01/2017 - 08:15
Sun, 12/02/2018 - 12:08 (Reply to #7)

I'm a statistician who knows well about sampling and testing. Again, the comparison is based on large samples of over 100 runs per condition. The results are fairly stable. The reason for an elite only rift is pretty easy to understand given that the Xp earned at the rift closing is actually higher than that earned by monster killing. For a 4-5 minute runs, focusing on elite-killing will always speed up the rift progress. For GR level pushing, it will be a completely different story: killing yellow elites often slow down the progress. This is why Nemesis Bracers are not recommended for that situation.

Since I've shared scripts and FastMode settings, just use them to repeat 100 runs per condition on your computer and character. Let me know what you find out. This will tell whether the result is only for my "specific scenario" or not (how "specific" is it just being curious?).

I'm currently on Paragon 3600+ now. With 600 points allocated to vitality, bot can now comfortably run GR = 100 with 1 death over 4 runs. Just finished 100 runs of Kill Elites & Dense Mobs condition. The new result is 9.0 Games/Hr and 600kB Xp/Hr. I'm running the Kill Elite Only condition. The extra Paraon points gained during testing were not used in order to make both condition Paraon = 3610. Will publish what I will find out.
 


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Sun, 12/02/2018 - 23:55

Mayhaps I was not clear: this idea has been tested.
Planner you shared was in Compendium for a long time, but i scrapped it cause it underperformed.
I have 4 or 5 different test scripts for it.

Which is no surprise if you take emotions out of it.
100 games is a small sample, thus i used it myself over long periods and had others test it.
Overall it was a net loss. Elite hunting only is a net loss.

You as a statician lack any results to compare with. Run the usual fire condemn. Vsswinta is public. Use it. Compare results. Then argue that your statician status has any bearing or meaning at all

LoN fire condemn, that you originally suggested is yet to be tested, until new patch comes so i'll hold off on any definite decisions on that.

To add: Steed charge use is not new. I have a script called Holy Pony for rifts and T13.
Can be used for GRs. Use Unleashed rune and Furnace/botp or Botp/addendum for more uptime of pony.

But, also run the normal variations.
We have build guides and recommended scripts so you can understand what is considered best/good and then you can make a more educated comment and claim.

I took a look at your script and as a newcomer its fine, but I wouldn't go writing home about it.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

hercle
Joined: 04/01/2017 - 08:15
Mon, 12/03/2018 - 09:21 (Reply to #9)

Thanks for pointing to the Condemn script you shared in public. I will give it a try. Any hint for what VSSWINTA means?

A few suggestions for your VSSWINTA script after a quick glance:
1) The damage increases of Condemn sword and shield are multiplicative so the combined increase is 8000% (9*9-1) not 1600% (8+8). This can be find in the skill damage of d3planner.com, which gives a result very close to the damage ticks shown in game.
2) I would suggest to remove irrelevant skills such as Bombardment, Heavens' Fury, Blessed Shield, etc. to make the script overall cleaner.
3) There are 3 Condemn actions with different attack distances. I assume they are for different runes (Unleashed, Reciprocate, Vacuum, Shattering Explosion, etc.)? Will the one the the maximum distance always be a catch-all so that it will fire even out of range with the Unleashed rune?
4) I didn't find a "move around" Punish action for procing the Hexing Pants. Did I miss something?

One quick finding that may explain the discrepancy between our testing results. I did finish the testing of the Steed Charge build on GR = 100. It turned out that at this cut-off level, the Kill Elite Only method has significantly higher death rate (1 over 2 runs) than Kill Elites & Dense Mobs even with the same toughness level. This is probably due to the clearance of normal monsters before approaching to elites in the latter method. As a result, the performance of Kill Elite Only with Steed Charge is significantly lower at this level (Games/Hr = 9.2; Xp/Hr = 560 kB/Hr, which is even lower than GR = 99). In addition, there are also 2 failed runs due to "no CPs" among 100 in this testing, which have affected the performance. I agree that 100 samples may still be lacking for the reason of failed runs caused by CPs or sometimes chained deaths (portal, Voracity, etc.). But the two builds are still pretty close given all these factors. My current conclusion is the Kill Elites & Dense Mobs is better at +1 GR level due to the less stringent requirment for toughness; whereas the Kill Elites Only method gives equal Xp performance and better legendary gaining at GR levels that toughness is not a limiting factor.


KatziKasimir
MODERATOR
Joined: 04/16/2015 - 19:28
Mon, 12/03/2018 - 11:04 (Reply to #10)

Very Secret Script Which I'll Not Talk About

1) Not sure how that is a suggestions 
2) ppl ask for these skills so he adds them, if you dont want them in the script you can feel free to remove them out of it.
3) Not sure what this questions means "catch-all" ?
4) It is there, as you see from the bot moving around.

Damn now I talked about it. (It was to mock ppl who talked about secret scripts they had)


I like cats come to Discord if you like them too.

Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Mon, 12/03/2018 - 11:26 (Reply to #11)

You are correct on number 1) The damage is multiplicative instead of additive. I will rectify that.
All the rest of your points are self explanatory and warrant no response.

To recap: I want you to understand that your baseline where you are making the claims and assumptions are based on a flawed testing method. You are using your own script designed for only one thing. 
Try a community tested, proven as a clean and working baseline. 
1) Run Steed charge with properly tuned script.
For Holy - Unleashed rune:
https://www.ros-bot.com/custom-script/vsswinta-holy-pony-1153265
For Fire/Holy/Phys - any rune other than Unleashed:
https://www.ros-bot.com/custom-script/vsswinta-rifts-1149391 (rifts)
https://www.ros-bot.com/custom-script/vsswinta-crusader-condemn-compendi... (greater rifts)

Write down the numbers, results etc.

2) Run Fire Condemn build as shown in the planner ( https://www.d3planner.com/225012527 )
3) Compare the numbers and be amazed that Steed Charge is actually rather lackluster if used in a real baseline.

Just as a side note: I recently ran Papa (the moderator) account since I have better connection. He was some 3.6k paragon. Decently geared. Fire condemn. 1R2GR. 
17 hrs and it was 680B/hr average. Some 220 games or so. In a host environment.
For a VM it would have been certainly less, but above 600 overall.

So that should get you interested.

4) After you have done all that - maybe you will have a better understanding of the class in general and why most users bot one build and one build only.

Good luck.

 


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

dothepe
Joined: 11/05/2017 - 04:08
Mon, 12/03/2018 - 13:13

"I'm a statistician who knows well about sampling and testing."

Just in case you missed it, Good Will Hunting is playing Diablo now. Can someone get this nibba a chalkboard?


hercle
Joined: 04/01/2017 - 08:15
Mon, 12/03/2018 - 15:48 (Reply to #13)

Any suggestion to make the bot run 1R2GR? My bot is actually running 1R2GR+1R1GR if there is no need for repair. The 3rd run is always single because the shards will reach the limit for gambling. Maybe this is why I have lower performance?

p.s. Found the ahk script for 1R2GR from a link in your public script. I think that is probably the key thing for improving the botting performance. I will give a test later today. For a comparison with your build and script, do I need to use Slash as primary or keep Punish as it is in my own build?


blueray_master
Joined: 12/22/2015 - 01:59
Mon, 12/03/2018 - 16:36

You don't get "the 65 times LoN boost in Patch 2.6.4". You get a 5x boost. The LON bonus goes from 1300 (13 ancients x 100)  to 6500 (13 ancients x 500).

 

Lower the shard gamble limit so it gambles after every GRift.


hercle
Joined: 04/01/2017 - 08:15
Mon, 12/03/2018 - 18:09 (Reply to #15)

The 65 times damage boost is for the base damage. The original boost is 13 times. Your 5x boost is the boost by the patch as compared to the original. Both are correct. :)

If it gambles every GR, it will be 1R1GR. The performance will be certainly lower. The folks are recommending 1R2GR.


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Mon, 12/03/2018 - 21:42

The way I have set my master profile is presented in the master profile.
Set gambling so all shards are used up and make sure that whatever you gamble will go over the inventory limit - so it gets salvaged/kept as well.
1R2GR can be done in two viable ways.
Run T11 so it's always 2 keys that drop or opt for T13 for the chance of a 3rd key. I don't bother with T11.

Ideally: You want to aim for bot gambling only in rifts.

So this can be achieved by knowing how much your GR drops shards and add to that number 150-200 for good measure.
Start with 2 keys/2 GRs and run T11.
But this method relies on bot not failing a game so not the most reliable..

Whatever way you go for: We do large samples for the average. And we repeat our runs over and over to see the average expected gains.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

hercle
Joined: 04/01/2017 - 08:15
Mon, 12/03/2018 - 22:16 (Reply to #17)

So I must have misunderstood the meaning of 1R2GR. If you run 1 normal rift then 2 greater rifts in one sequence and can still get 680kB Xp/hr, that's really amazing. Which GR level are you botting at? Is this done by the tuhz0r script?

I always run normal and greater rifts separately because they use different gears (weapon, amulet, rings, chest, belt, bracer, Sage sets, which are a whole lot of changes). Does Ros-Bot or any AHK script can change gears in the Armory?

What I'm thinking of a way to improve performance in a GR only sequence is to go through the vendor loop only after every two GR runs when the inventory is close to full. Currently there are always some runs in which bot will go through the vendor loop with the inventory occupied by less than half. This cannot be avoided even if gambling belts or jeweries. I am still waiting for a future version of Ros Bot to implement such possibility.


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Tue, 12/04/2018 - 02:31

680 was with 1r2gr.
You are talking of either stacked runs (keys and pools pre-stacked)
Or
RFP - keys stacked and rifting for pools.

With it stacked its not hard to get 850b+ at that paragon level.
With rfp i would expect around 780-810b+

I once told you we have guides in Discord #tags. That was not a joke nor a jab. Community has been botting condemn over a year now. The amount of testing that has gone into it is rather remarkable.

You are welcome to ignore all sources that are meant to make you excell at crusader, but don't be surprised when there are eyebrows raised at the results you consider...good. Just saying.

*I ran 96 or 97. Can't remember. That was with Papas account.
My own toon is nearly twice the paragon and it has numbers you would expect:
Stacked 1.3 tr+
Rfp 1.150-1.2tr+
1r2gr 900-950b
At GR100-104 for speeds


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

hercle
Joined: 04/01/2017 - 08:15
Tue, 12/04/2018 - 13:45 (Reply to #19)

I will test your script on GR = 100 stacked runs and compare it with my script.

Running your script now. It turned out my original FastMode density limit is too stringent (10 normal monsters). I now also copied your setting with a limit of 7 normal monsters.

I also figured out a reliable way to always gamble after every two runs In the stacked mode:
Items to gamble = Belt
Shards limit = 2*shards/run

At GR = 100, shards per tun = 427. So the ideal shards limit will be 854. Bot will use up all the shards for belts after every two runs. If repair is needed in a vendor loop without gambling, this will reset in the next run.

I just finished 10 runs in VM with zero death. Currently I've got 10 Games/Hr and 950 kB Xp/Hr. So far it is amazing! It seems that the Xp earned at rift close also scales with the Xp earned during the rift, contrary to my original belief of fixed Xp scaled with GR level only. If this is correct, killing normal monsters will be really important.

I noticed that your script use the Town movement of a Law of Hope skill to enable movement. This looks good in normal monster killing but seems not very reliable during boss killing. I saw the "IsMoving" sensor keeps toggling between True and False. I will see how does this method compare with the Punish move around method in my script. I will bot 24 hrs with each method and post screenshots for results.


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Tue, 12/04/2018 - 23:19

The method is reliable and it keeps Hexing up.
I would expect no less than for you to do your due diligence in testing in thoroughly.

Move around was tested and found lacking by me. Although it might allow situationally more movement, it does not keep range all too well and residual mobs were not sufficiently hit.

But you are welcome to edit it.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

hercle
Joined: 04/01/2017 - 08:15
Wed, 12/05/2018 - 07:25 (Reply to #21)

Just completed 175 runs in 17 hrs. The result is converging and quite stable now. With the same FastMode setting and script as yours, perfomance in stacked mode is 10.12 Game/Hr, 680 kB Xp/Hr (screenshot taken at the rift entrance is below). It's about 13.3% improvement as compared to my previous result of 600kB Xp/Hr. Since I've changed three factors: FastMode setting, script, and shards gambling, further testing is needed to find out which ones cause this improvement. I will change the script to my own one and give it another try.

Screenshot


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Wed, 12/05/2018 - 21:41

Fyi - 42 deaths is not considered good.
Mayhaps use indest and some avoidances.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

hercle
Joined: 04/01/2017 - 08:15
Thu, 12/06/2018 - 21:01 (Reply to #23)

The only avoidance setting I have is Ghom's Gas Cloud for fighting the Voracity Rift Guardian. Other than that, my Sader is pretty tough and can fight any Rift Guardians without a single death. Currently 80% of the deaths is caused by stuttering/freeze of D3 in VM. This happens once in about 5 games, during which Bot is still refreshing the overlay but the game competely freezes for 5-10 seconds. This rarely happens if I run the game on the host machine.

I finally figured out some mechanisms of Condemn scripts after playing around your script and tuning various settings. I also made some changes to your script, which seems to improve the performance a little bit:

1. Force-move mechanism: The "move around" bound to a primary skill in my original script, although works for boss killing, will significantly slow down the normal moster killing. I finally abanded it and only make use of the default "keep exploring" option. I saw many scripts with hexing pants use an attack distance of 5-8 yards for melee primary attacks. I found this is actually not necessary and may decrease the performance on the contrary. The key variable is the delay time of the primary skill. Basically, if a delay time significantly longer than 50ms is used. Bot will have no action to cast during this period since other skills are mostly in their cooldown periods (including condemn, around 500ms). This will effectively enable the "keep exploring" movement. I found the optimal delay time is 220ms for fire condemn. This is derived by the 53 frame breakpoint of primary skill if there is no attack speed bonus on any gears: 53/60/4 = 0.22. In this way, the primary skill will be casted every 4 actions. This will avoid possible mismatch of action cast delay and IAS delay, which may cause a break/pause of the stutter step when attacking with the primary skill.

2. Multiple attack distances of Condemn Buff+Attack. Your script have three tiers of attack distances. I figured out this is probably for the purpose of compensating the 3 sec delay of Reciprocate or Vacuum runes as well as resource management. I made some modifications to your script by removing elite from the mid-tier distance. From the discussions in the forum, B+A does not seem to work with Elite. I suggest to add a B+E action in the first place with 0 resource requirement. Then the three distances tries will have incrementing resource requirment. Since each Condemn costs 12.6 wraths, I would suggest to use 26, 51, 63 for the three tiers. Given the movement speed with Laws of Hope enabled is around 10.5 Yr/Sec (6*175%), I suggest to use 14, 24, 35 as attack distances in the three tiers. In this way, when resource is abundant, Condemn will be able to fire 3 seconds ahead when approaching monsters. When resource is low, it will only fire when at least a monster is within range. When Elite is detected, it will fire always as long as resource is not depleted. In fact, some further fine tuning for the resouce threholds is probably needed for abvoid resource depletion.

3. I also removed the A+E actions for primary skills. As long as the Elite and Big Buy weights are much higher than normal monsters (e.g., I use 40 and 80), A+D will be sufficient for bot approaching the best targets.

Please take a look at the modified script: https://www.ros-bot.com/custom-script/vsswintamod-1406925
Comments are welcome!


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Thu, 12/06/2018 - 21:52

1. Move around was tested and is a net loss overall.
The timings have been tested and I ran force recast on A+E with 200ms for A+D for a long time. This caused more furious twitching and did not allow the bot to wonder off too far in the alotted radius of 5-10 yards that I've ran.
But this also meant that surrounding residual mobs did not get hit.
Net loss overall. Scrapped.
We are running with 49 frames.
Furnace has attack speed of 1.10 and our paragon allocation is giving us 10% IAS, unless you skipped it.
It makes virtually no difference however in real time application, the 4 frame difference.

2. A+B does work on elites, having A+B+D would be finicky.
3 tiers is for resource management.

3. A+D is fine also, but I preffer A+E for guaranteed elite targeting.

Going with the rule of: If it aint broke - don't fix it.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

hercle
Joined: 04/01/2017 - 08:15
Thu, 12/06/2018 - 23:05 (Reply to #25)

1. Furnace actually has attack speed of 1.00. The IAS is 1.10 with the Paragon 10% already taken into consideration. I observed quite a lot of runs for the debug overlay 'IsMoving" sensor. A delay of 220ms gave significantly less "False" than 240ms. The distance of running in "keep explorling" actually does not matter as long as it is still within a certain range. If the delay is too long, bot may run out of 15 yard condemn range occasionally (e.g., when Iron Skin Flash is active).

2. Any idea why B+A (I would rather not denote it as A+B because this is still a BUFF action) is more efficient than B+D for Condemn? My understanding of B+A with an attack distance of x should be equivalent to B+D with a surrounding distance max of x and density limit of 1. But when comparing the two implementation, I found B+A fires significantly faster.

3. In certain senario when there are more than one nearby elites, A+D may found a better target than A+E. A+E will always attack the nearest elite, which may not be the best target. The best target is the monster with two elites both in the circular range of 15y. For example, with a monster distribution of E--10y--N--10y--E, where E are two elites of 20y distance and a normal monster is between them, the best target will be the normal monster, but not either of the two elites.

 


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Fri, 12/07/2018 - 00:16

1. Right. 1.00 speed. In my mind i added 10 ias to it, but thats a mistake.
Does not change 48 frames however.

I seriously doubt that 0.02 is something you noticed nor it showed any improvement, but you can believe what you like without testing it.

2.First and foremost its an attack action. Buff just lets it be cast starting from X range.

ABD starts favoring mobs and has not been as reliable for me in casting.
That is why theres an AEB.

3. We want to target nearest elite and move up in progression. We do not want to run to a farther pack even if it means its beefier. We will get to the next pack eventually.

Your system would have it disregard closest elite, run off, then return, then have bot make its way back again.

Look, Hercle, you are thinking like a human would. Thats a mistake.
Bot acts different, lot of times its just down to making a change and letting it run. Compare the numbers. See how bot acts, how it feels.

And eventually you might be a decent botter.

Btw. One fundamental flaw in your results are the deaths. Thats basic knowledge to drive the amount as close to 0 as possible.
Not only do you lose experience pools, you lose BoTP buffs, will do more town runs for repairs and you lose time on resurrecting.
You get hung up on your maths, but real world application requires some logic and understanding of the end goal.

Tldr: Start by reading guides. Molten explosion avoidance is a necessity.

Good luck.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

hercle
Joined: 04/01/2017 - 08:15
Fri, 12/07/2018 - 12:34 (Reply to #27)

As I explained above. More than 50% deaths are caused by D3 freezes in VM. This cannot be avoided as I have fresh installation of Win10 with only D3 and AHK installed, plus downloaded binaries of Ros-Bot only. The game runs smooth on the host but I have to use the computer to work. The remaining deaths are mostly caused by portal/shines. There is no way to make it close to zero after more than 100 runs even on the host. Send me a screenshot at GR=100 if you can do this.

I haven't seen a single death caused by Molten explosion after waching more than 10 games contrinuously. It just slows down the progress after adding an avoidance because the bot will always wait at a nearby location. Without the avoidance, it will pass the spot quickly in 70% chance, or take a potion or call the Akkrat cheat death in the remaining. The avoidance is only necessary in GR push when toughness is at the border line.

Breakpoint is 53 frames (58/1.1) not 48, again. You need to use some maths this time, not just relying on intuition. Or you can find it on d3planner.com. The 20 ms difference is actually influential because is more than one frame. After 7 rounds of pingpong movement (240*7=1680ms), the attack of the primary skill during the next action (at 1767ms, i.e., 53/60*2000) will cut down the timer delay to only 87ms. This is when the bot may generate a pause movement. I've done some basic counts of of the "IsMoving" sensor. It indeeded reduced the frequency of "False", which only happens when attacking with the primary skill.


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Sun, 12/09/2018 - 23:05

Yes. Yes it can be avoided. Diablo 3 freezing is nothing new. Just take a moment to ask Google about it. Posts from 2012 up until 2018.

There are tons of "fixes". I use quotation marks since there is no one cause for everyone.
I had freezes for months until I got them handled by disabling services that were causing it.
Latencymon app pointed the way.

Hercle - I know this sounds impossible for you to understand, but unless you skipped 10 IAS - its 48 frames. Check with d3planner.
Your maths is busted.

More so - until you get some hard results about whatever you believe to be true - I see no point in your beliefs.
Because thats what they are.
Have you considered in network latency and VMs input lag into your maths? No? Well, you gotta do some maths and not rely on your intuition.

My intuition however has led me to test everything out with numbers to back it out, instead of purely theoretical yapping.

:) Good luck, you seriously need it.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

hercle
Joined: 04/01/2017 - 08:15
Mon, 12/10/2018 - 09:07

This is from the d3planner build you provided (not mine):


Cyp
MODERATOR
Joined: 12/26/2015 - 19:41
Mon, 12/10/2018 - 23:22

Yet I saw 48 frames. I'm quite sure of it. Maybe I imagined it?
In any case - if you lower the delay, you will achieve only that bot will twitch in more restricted area.
That was a version many tried. It helped in the sense of generating tad more wrath/health, but overall EPH loss.
Due to simply bot staying very close to elite - thus not utilizing the 15yrd range.

Elite staying in the epicenter of blast zone vs elite staying on the edge of blast zone doesn't matter for elite. Same damage taken.
Yet residual mobs staying farther away do not get hit.

A+E is needed for prioritizing elites. I mean I have over 100 versions with different tweaks to gain bit more eph out of it, but so far all the results point to one version being the best and I don't feel like doing much with it after 8 months.

You are welcome to test and run it. Change it. Use all the maths you want, but only thing that matters is the hard results from eph numbers.
Make a better version - all glory and power to you.


Need fast help? Join Discord. Guides: http://tiny.cc/s7poaz Common troubleshooting guide:http://tiny.cc/b8poaz

Advertisements

Advertisements