Specialty breakdown

[Chthugha]

Intro:
After the change in the league point scoring system, I felt that it became even worse for charm players.
We all know that charm was never going to do great compared to hardcore and know-how.
I also had the feeling that know-how was worse than hardcore, but with the new system, it actually has an advantage over hardcore, even if you won't actually win more battles than hardcore.
So to check that I decided to do some simulations.

Explanation of the simulation:
So this simulation is obviously limited. I am not going to create accounts indefinitely with different level and different setups and girls.
So to make it easier for myself, I decided to use a couple of restraints:

1. All players are off the same level (so have the same base stats), bought all stat points available in the shop, have access to the same equipment sets, are not part of a club (forgot to include that), and have the same harem and harem level.

2. Every player in the simulation is unique. This gives me 3 x 7 x 35 different players (3 specialties, 7 different ways to equip mono legendary equips and rainbow equips, 35 different ways to equip 4 different epic boosters on 4 slots).

3. Every player plays at least 100 games as attacker against all other players, including a copy of himself/herself.

4. Judges are not included in the simulation, specialties and orgasms are.

5. Every player uses the 3 current best girls in their team for their specialty.

EDIT: As addendum, I have to say this obviously doesn't represent reality. I simply do this so you can see that even if you give players the same things, the specialties are not balanced.

Results of the first simulation:
In the first simulation, everyone did 100 fights against everyone. I wanted to do 1000 originally, but that would take too long for just a simply simulation.
Because it's annoying to dig through countless amounts of data, I decided to look at two things:
1. Percentage of overall won fights
2. Average amount of points scored, according to the current point system in the league.

There were a couple of clear results immediately.
First of all: the best setup for best win percentage was different from the best setup for best point score for all three specialties.
For hardcore 3 high-end rainbow equips and 3 high-end mono-stat equips + 4 cordiceps gave the highest win rate (99.884%, 18.43 points/fight) whereas 6 high-end rainbow equips and 4 cordiceps gave the highest point score (99.882%, 18.63 points/fight). Note that the percentage difference is not that big, and can be due to RNG.
For charm the same thing happens, but it's 2 rainbows + 4 monos at the best percentage (99.64%, 17.72 points/fight). The same setup does give the highest point tally again though (99.37%, 18.04 points/fight).
For know-how we see the same setups doing the best as for charm: (99.83%, 19.36 points/fight) and (99.72%, 20.35 points/fight).
The other thing that is noticeable, is that jujubes are completely useless (I think we already knew that).
Some charm accounts with 4 jujubes got as low a score as 15% and ~5 points/fight on average, so that's really bad when you have the same base.
Lastly, I took the average of all win percentages for each specialty: HC = 84.27%, CH = 71.62%, KH = 92.15%.

So, conclusions:
1. know-how is outscoring hardcore and charm massively.
2. jujubes are useless, better remove them from the simulation.
3. hardcore still is the best, when it comes to just winning the fight.
4. because jujubes are so useless, you could see these accounts as boosterless accounts. it's clear to see that without boosters you will be grossly outscored by others of the same strengthlevel as you using boosters.
5. charm doesn't seem that bad at first sight, compared to hardcore.
6. judging from the overall percentages of all accounts of one specialty, it seems that know-how is the least hampered by a bad setup, whereas charm is the most hampered by it. which means that as charm player you have less options to play optimally.

Results of the second simulation:
Like I suggested in the conclusion section of the first simulation, I should remove the jujubes from the equation, which means no more really easy victims to get lots of wins and points from.
That means that the amount of accounts was reduced to 3 x 7 x 15 = 315 accounts.
The amount of runs went up to 250 for this one.

results:
HC: 6 rainbow equips + 4 cordiceps --> 19.28 points/fight at a win rate of 99.73% versus 4 rainbows, 2 monos + 4 cordiceps --> 18.10 points/fight at a win rate of 99.77%.
CH: 4 rainbows, 2 monos + 4 cordiceps --> 17.48 points/fight at a win rate of 98.94% versus 2 rainbows, 4 monos + 4 cordiceps --> 17.30 points/fight at a win rate of 99.14%
KH: 6 rainbows + 4 cordiceps --> 19.96 points/fight at a win rate of 99.44% versus 2 rainbows, 4 monos + 4 cordiceps --> 19.13 points/fight at a win rate of 99.67%
But the biggest thing I got from this was that chlorella are also inferior to the other two boosters. 4 ginseng root outperformed 4 chlorella. Basically chlorella had taken the place of jujubes in the first simulation.
Apart from that: charm starts to look worse and worse. The gap between hardcore and know-how seems to close a bit when the accounts are closer in strength to eachother.

Results from the third simulation:
This simulation has only 3 x 7 x 5 = 105 accounts left, so I upped the amount of runs to 1000.
Results where very similar to the second run.

results:
HC: same setups have the best scores in the two different departments again. this time in the elite league with the following scores: (17.72 points/fight, 99.23%) and (17.59 points/fight, 99.33%).
CH: same setups. this time: (16.66 points/fight, 96.91%) and (16.63 points/fight, 97.77%).
KH: same setups. this time: (19.19 points/fight, 98.78%) and (18.48 points/fight, 98.98%).

Conclusions:
1. The right setup (6 up to date high-end rainbow equips + 4 epic cordiceps) leads for a massive league advantage for know-how over the other specialties, event though hardcore accounts might win more fights. The difference is over a point.
2. Charm is by far the worst specialty to have if you are in a league with a lot of players with about the same strength as you.
3. Hardcore with the right setup was indeed the best setup for the previous league system, and still is for arena.
4. Cordiceps are overpowered as boosters. The more you have of them, the better. Unfortunately they are the third rarest boosters in the shop (after the useless jujubes). So it won't be possible to keep 4 on all of the time. If you need to replace them, go for ginseng roots first, chlorella second. In fairness, rare cordiceps might even be better than those, but I didn't include those in the simulation.
5. Rainbow equipments indeed have become way better after their update. So much so that for both hardcore and know-how, it is best to use 6 of them. Take in mind that the need to be up to date, and with stats in the high end of the range they can be for basically all 5 stats.
6. Know-how is the specialty that is the least affected by a bad setup. Charm is affected the most. Which is another strike against charm (and it's useless specialty move) and another favourable trait of the know-how specialty.

Lastly:
The league advantage of know-how isn't that easy to solve. Hardcore is still winning more (with a good setup), so simply tweaking know-how's specialty is going to put them in the same trouble as charm. Charm's problem might actually be solved by tweaking that skill. In a future simulation I'll try to match hardcore in points by making the charm specialty a bit stronger.
If anybody is interested in them; I have some files of data. The three average files for the three different simulations. The python codes used. The actual result files of the third run for every setup. If you want them, just ask for them.

Edited August 25, 2019 by Chthugha

[Habi]

cool thing!

i have 3 questions though:

1. did your sim account for dominance scheme? results might be shifted a little if rng for harmony procs is shifted aswell
2. can you include legendary ginseng roots in the sim? at high levels using some of those is far better than epic cordyceps only. they're the only viable legendary boosters working almost properly at the moment, increasing all your stats by 6% (calculated from the hero base values + equipment before club bonuses are applied, however they seem to boost used epic ginseng roots aswell)
3. can you include club bonuses? one might think +10% on everything doesn't change anything but depending on how close some fights were won/lost the results could indeed vary.

[Chthugha]

48 minutes ago, Habi said:

cool thing!

i have 3 questions though:

1. did your sim account for dominance scheme? results might be shifted a little if rng for harmony procs is shifted aswell
2. can you include legendary ginseng roots in the sim? at high levels using some of those is far better than epic cordyceps only. they're the only viable legendary boosters working almost properly at the moment, increasing all your stats by 6% (calculated from the hero base values + equipment before club bonuses are applied, however they seem to boost used epic ginseng roots aswell)
3. can you include club bonuses? one might think +10% on everything doesn't change anything but depending on how close some fights were won/lost the results could indeed vary.

1. Yes
2. I don't know how the legendary boosters work. I haven't checked it yet, but last time they didn't do what they said they should do. That's why I didn't include them to begin with.
3. They influence the boosters as well, don't they? God, that's going to be annoying to add.

[Zteev]

I'm wondering if you've considered the trying of a mix of booster types? For example, 2 cordyceps & 2 chlorella, 3 cordyceps & 1 ginseng, etc.

Edited August 26, 2019 by Zteev

[GeorgeMTO]

11 minutes ago, Zteev said:

I'm wondering if you've considered the trying of a mix of booster types? For example, 2 cordyceps & 2 chlorella, 3 cordyceps & 1 ginseng, etc.

I think you've missed this section

12 hours ago, Chthugha said:

2. Every player in the simulation is unique. This gives me 3 x 7 x 35 different players (3 specialties, 7 different ways to equip mono legendary equips and rainbow equips, 35 different ways to equip 4 different epic boosters on 4 slots).

 

[Zteev]

1 hour ago, GeorgeMTO said:

I think you've missed this section

 

You're right, I did. thanks.

[Dmytro]

Can you post the code for the simulation somewhere?

It is also interesting to see what the simulations say about the relative value of the attack, defense, endurance, harmony, and also how these depend on class and your opponent's strength.

Even without simulation, it is easy to see that charm is the weakest class since defense is significantly below 50% of the attack or 10% of the ego (and Narcissism cannot restore lost ego, which caps its effectiveness).  Still, against weak opponents, HC < CH < KH.  Against strong opponents, CH < KH < HC.

I guess one reason that CH remained weaker is that the differences are small relative to differences in player level, and the developers like round numbers.

A possible fix for most of the difference is for Narcissism not only to double the defense, but also deflect 25% of the remaining attack (alternative, 20%).   With the right numbers, this setup would create the circular dominance scheme even if there was no 20% harmony bonus.

Other comments

99% winning probability is unrealistically high for most players (but admittedly, it reflects the current state for top players in Dicktator leagues).

Epic Cordyceps increases the gap between HC and CH players by increasing the attack over defense ratio; most players will not be spending 180 Kobans/day on boosters (60*4*3/4 to complete all league fights).  And among legendary boosters (relevant for players with level >≈300), Ginseng Root is comparable to Cordyceps (or better if Cordyceps is bugged or does not boost the alpha's attack; I did not test it).

On 8/25/2019 at 10:39 AM, Chthugha said:

Some charm accounts with 4 jujubes got as low a score as 15% and ~5 points/fight on average, so that's really bad when you have the same base.

This does not sound right (unless you mean 3+5 points) since a level one player would get 4 points/fight.

21 hours ago, Habi said:

can you include club bonuses? one might think +10% on everything doesn't change anything

Clubs should give a small advantage to KH since by increasing both your three stats and endurance, life over attack ratio is increased slightly.

[Habi]

23 hours ago, Chthugha said:

1. Yes
2. I don't know how the legendary boosters work. I haven't checked it yet, but last time they didn't do what they said they should do. That's why I didn't include them to begin with.
3. They influence the boosters as well, don't they? God, that's going to be annoying to add.

from what i found out with my own tests (credits also go to @Shal, he's a geat help in finding out stuff like that) the scheme is as follows:

1. everything that gives you absolute boost is applied first (like all boosters but the legendaries)
2. everything else that gives you relative boost is applied on top of that independently (like legendary boosters and club boosts)

so that means that e.g. 3 legendary ginseng roots and the club bonuses increase the effect of 1 epic ginseng root by 28% (10+3*6) or 2 legendary cordyceps increase the effect of 2 epic cordyceps by 20% each. i didn't test legendary chlorella/jujubees because of their uselessness but that scheme probably applies aswell. since the legendary cordyceps are completely bogus and far away from what they should provide (~3200 bonus damage at ~lvl400) the legendary ginseng roots would be the only legendary booster that would make sense to include for the moment.

hard to tell the outcome of that situation since reassurance's and wild burst's effects both grow with 0.5*mainstat. my guess is though that know-how gets even stronger, especially with rainbow equipment. ego suffers less than damage does when shifting from monos to rainbows. that's probably the reason why nonoptimal kh setups are more forgiving than nonoptimal hc/ch setups.

Edited August 26, 2019 by Habi

[linkdream]

Chthugha does what kinkoidon't. Thank you so much!

Seriously, how come the dev team didn't use this kind of simulation to balance things? They certainly have the tools to do it.

I don't wanna pry but seriously, is this incompetence or couldn't-give-a-damn approach?

[Observer_X]

Come on, i think it's quite obvious. It's how-we-mug-the-players-that-are-too-lazy-too-use-math (which probably are the largest slice)

[_shal_]

7 hours ago, Dmytro said:

This does not sound right (unless you mean 3+5 points) since a level one player would get 4 points/fight.

I agree that Chthugha probably meant 5 bonus points rather than 5 in total, but for reference, the minimum possible score is in fact 3 points if you're not doing any damage to the opponent at all.

Edited August 26, 2019 by _shal_

[Chthugha]

No, I meant, 5 points in total, but that just means there was a mistake somewhere in the code.
And indeed there was a mistake somewhere in the code (a minimum instead of a maximum), so I'll do a rerun and see if it changes anything.
This doesn't change the percentages, it does change the average point tallies.

Edited August 27, 2019 by Chthugha

[Chthugha]

I did the simulation again after fixing the small mistake in the code, and the results for the first run are pretty much the same as before for the best results, but with some minor increases here and there. Which is mostly because of the wrongly calculated amount of point for a loss. The lowest average point total obviously changed a lot though. It's at 12.81 points/fight with indeed about a 15% win rate. This does seem more likely.

I didn't do the second simulation again, and for the third simulation the percentages are about the same, so that's good for reproducing results, and the point tallies went up a bit (as expected) by about 0.2.

---------------------------------------------------------------

Adding the 10% bonus from clubs made it a little bit different.
For HC and CH the most succesful setup was 6 RB + 4 cordiceps for both win rate and total points.
For KH the most succesful setup for wins was 6 RB + 4 cordiceps as well, but this was a very close second in the points department to 3 cordiceps and 1 ginseng root.

I think it's fair to say that the difference displayed are probably down to some RNG involvement. The differences aren't that great overall.

@Habi I have decided not to do legendary ginseng roots. At the moment that would mean reworking the code a bit too much, and right now I'm not in the mood for that.

Edited August 27, 2019 by Chthugha