Rarum

Another top 1. Maybe a bit more expensive than I'd have liked (not the worst it could have been still), but that might have been because I expected Xiarros to do better than that. A final average of 23.23, really? I know you can do better than that, and I thought the scores I tracked for you suggested a better average. Did you leave some fights unfought? Okay, I suppose Xiarros still did end up in the fourth place. No use fighting wars you can't win. Also while I kind of hyped the master versus pupil aspect of this week, @zoopokemon didn't really want to fight me for the scraps that are the top 1 rewards nowadays. In fact, they told me: The best I think I could have done in this league is 7183 points. Maybe there was a bit of misplaying, but also not that many opponents.

I would be okay with buffing the weaker elements. Making the playing field more complicated gives advantage to those with the computational power to churn through it all... Wait, hmm, for the benefit of the playerbase who wants to see more girls see some action, I mean. No ulterior motives here at all. But yeah, the playful element for instance is so weak it's all too often beneficial not to use it even if it's countering a water team. And that's like the one concrete thing that players would normally be able to rely on, that countering is beneficial.

This seems to be now a talking point, so I'll also give my current thoughts on how good the different elements are. Almost a month ago, I gave a list to @zoopokemon about how many times different elements have appeared in my team recommendations. Sometimes, less so nowadays, teams are just listed as "overall best", so it's not a quite correct list, but we can still gain some knowledge from it. Also note that for the most part, I don't consider teams that only have one girl of an element. Anyway: Water: 1577 Darkness: 904 (we had an elemental blessing) Nature: 463 (we had an elemental blessing) Stone: 433 Fire: 425 Sun: 268 (we had an elemental blessing) Psychic: 153 (we had an elemental blessing) Light: 96 We had the following elemental blessings during the observational period: Dominatrix on 11/1/2021 until 11/8/2021 Playful on 11/29/2021 until 12/6/2021 Voyeur on 12/13/2021 until 12/20/2021 Exhibitionist on 1/17/2022 until 1/24/2022. I went through the trouble of subtracting the affected data points, and got this: Water: 1136 Darkness: 613 Nature: 371 (still do note I miss nature mythics) Fire: 342 (also note I haven't had access to a fully upgraded Alexa) Stone: 328 Sun: 186 Psychic: 34 Light: 29 Sun still is maybe over-represented, and it looks like on the week 12/6/2021 until 12/13/2021, two sun girls were the very best raw power ones even though it wasn't an elemental blessing. If we discount that week as well, we get: Water: 967 Darkness: 595 Nature: 342 Stone: 298 Fire: 291 Sun: 35 Psychic: 29 Light: 29 So as you can see, the outcome depends quite a bit also on which girls happen to be the most powerful in a given week, when we have only this few weeks in our comparison. I wouldn't be surprised if nature also had something similar as sun going on. So really the most I can get from this is that roughly speaking there are three "power categories" for the elements: First is the water and darkness, situational which one you want to prioritize more (and water is near useless in something like the Pantheon xx00 floors). The second power block encompasses nature, stone and fire. There's too little data to meaningfully put these into any order. Last, there are the trash elements that are only good if those girls happen to be among the most powerful ones for the week: sun, psychic and light. Or *sometimes* if you can counter an element with them, but even then it's not a guarantee that they should be used.

The simulators (that I know of) are agnostic to the current blessings. Furthermore, Numbers' simulator doesn't even acknowledge the team's total power. It's not necessary in the simple case where you just have a set of known stats; all it already is part of the more regular stats. I do have to take the total power into account in my simulations because it goes beyond the simple case of known stats. I actually have to simulate new stats for all the hypothetical scenarios. Umm, maybe not the best place to make comparisons, but maybe not atrocious when only worried about player synergies. According to Numbers: Also, what Liliat said about how sometimes a girl with a weaker active bonus can still have an advantage in edge cases.

On a league ending on 13th of January 2022, Lgm123 reached an average of 24.32 with 7443 points. That's the record average currently, if we discount the 15/25 system scores. With the 15/25 system, the highest I have records of is 24.387 by MNEFC on the league ending on 5th of August 2021. Another high one was 24.387 by Agent009, also on 5th of August 2021. But it's not really saying much since that system was so short-lived and data gathering wasn't that large-scale back then, that I have four data points in total from it. We don't know who used AMEs in which leagues. It doesn't show up in the player stats, which themselves aren't kept records of. Ah! But we do know for sure that AMEs weren't used before they were introduced. For those scores, discounting the 15/25 system, sliqguy holds the high score of the records I have, with an average of 23.217 (at 7174 points). It was scored on 3rd of June 2021.

I definitely would like to get my hands on more data for RedEagle. They pretty much must be in the top 10 even if we assume they just got a very weak bracket in that one data point there is. Getting three top 1s out of four "can't" be pure coincidence. But the positioning within the top 10 is quite up in the air still. Maybe a bit more conservative move would have been to at least put them below SJL. I have been thinking of doing this again sometime. The rate of gathering data has increased but there are a lot of gaps in leagues I know the scores for still. So it'll be a few weeks before it makes sense to do another one.

Yeah, I apologize. I don't remember what exactly was said, but I remember hearing that when somebody says something hurtful, it's often because they themselves are feeling negative in some way. But it's no justification; there were clearly more constructive reactions I could have chosen (maybe even leaving up to JVR) even if there was miscommunication in some place or another. I at least didn't even consider the "you in general", since to me the questions don't appear completely rhetorical but concrete examples of problems that they might want to hear some thought process on for solving, to bring concreteness into the discussion. A lot of the heavy lifting of finding out how the combat system works has been done by other people. I'm more of the programmer guy, and even that I don't feel I'm much of a master of despite some amount of formal education and stuff. "I just have a lot of time on my hands to throw stuff at the wall until something sticks." I've been trying to resist saying this but I guess I'm the exception who likes their sensuals at player level 472 now (and I was already using them a decent bit ever since I was in my early 450s or late 440s, but that was even before awakenings so things might very well be different now). The thing with sensual, most notably of the elements, is that it's situationally superb, like you probably have also come to understand. It's just much more often so for very powerful players. But yeah, despite all the simulating experience, I don't actually know what's good in what situations. I have hunches, sure, but it's such a vast problem space that despite attempts at shrinking said problem space, there have been exceptions that cropped up for all but the most generally applicable of heuristics I've tried. And even those are tailored for my specific situation; maybe players of other power levels would have had to develop different heuristics.

I'd like to say it's based on complicated simulations in my case, but really, it's not. I tried doing that at one point, and you really don't want to be running that often. I do have something in place that records how well teams work as defense teams as I switch through different ones, but yeah, it's still for the most part gut feeling. I don't place super high importance on having the best possible defense team; even though it matters, there is a lot of other stuff you can do. I forget to switch to it often enough (and I don't like wasting Kobans on having boosters on all the time) to think a serious opponent won't be attacking me at my strongest regardless. I almost see good defense as a placebo strategy, something that makes you feel you're doing well while not actually mattering all that much.

Some reading comprehension would be nice here before going on this long of a rant. They specifically talk about having 6 of the top girls instead of all 7 (and later on also talk about having just 5). If they were all already upgraded for whatever reason, I'd say Venus is the worst choice of all the four that are being discussed. Path is a 3-star legendary; an absolutely HORRIBLE idea to upgrade even if they were a game-changer on some week. So really your choice comes down to Stacy and Himari. I'd say it depends on your relative power in the league which one is better.

This thread is for trying to keep track of how tough top Dicktator III players are in PvP. My efforts to this effect could be of general interest, hence this thread, but of course there are no clear-cut answers, so I encourage discussion about it, too. The following is the very first ranking I produced about Nutaku's top players, and more up-to-date ones can be found later in the thread. I tried to produce a ranking of whom I think are The Roughest Toughest Fighters in the Land at the moment (only data after awakenings were introduced is used in the evaluation). Take it with a grain of salt; rarely there are direct confrontations between these people (and even if there are, I might not have the data of it). I also have indirect evidence for D3 first places on leagues that weren't directly observed. Lgm123. Reasoning: 4/4 D3 most recent victories, highest recorded average during the observation period, on all three weeks I have data for has gotten the highest average. Also, while not strictly evidence for PvP success, tied with Hugh Jerexion for highest Pantheon floor. La Kajira. Reasoning: 4/4 D3 most recent victories. Gets second place on both "expected score" and "highest personal score" rankings. The only weeks where somebody has gotten a higher score have been from Lgm123 (twice) and karistea (once), and the highest once. RedEagle (very little data). Reasoning: 3/4 D3 most recent victories. The only week I have data for them, they have the highest average. Highest expected score ranking and third highest personal best score. SJL996 (very little data). Reasoning: 4/4 D3 most recent victories. The only week I have data for them, only La Kajira and karistea got a better average, and was followed by people like sliqguy and bla. The third best expected score. sliqguy. Reasoning: 4/4 D3 most recent victories. Sixth on the "expected score" ranking and fourth on the "highest personal score" ranking. Beat people like bla and Hugh Jerexion in average on one week, lost to karistea and bla on another. karistea. Reasoning: 4/4 D3 most recent victories. Fifth on the "expected score" ranking and fifth on the "higherst personal score" ranking. On one week, had the best average, but on another Thulsa Doom had a higher one. Thulsa Doom. Reasoning: 3/4 D3 most recent victories. Third highest average on one week and second highest on another. Fares okay in predicted score. Hugh Jerexion. Reasoning: 3/4 D3 most recent victories. While seems to possess a lot of raw power as evidenced by the tie in highest Pantheon floor, has fared not that well in weekly rankings or estimated score ranking (11th). bla. Reasoning: 2/4 D3 most recent victories. 8th in expected score. On some weeks seems to score well and on others not so well. Rarum. Reasoning: 2/4 D3 most recent victories. Only 13th on the expected score front, but has beaten Hugh Jerexion, and gotten mostly higher weekly averages than other candidates I had for the tenth place. There are other people who have gotten 3/4 D3 most recent victories, but their expected averages are even worse.

Yes. I'm talking about a possible bug in how the playful element gets handled in the hypothetical scenario modelling that's going to be really annoying to find out if it exists or not. And yeah, because it's so bad as an element I didn't think this Schrödinger's bug affected the findings about the awakening strategy all that much, and thus didn't bother attempting to pin it down and fix for those. But if we are comparing elements, well, then it might be kind of important that even one of those is busted.

Some form of an idea of what works when fighting for the very top ranks can be gleaned from my replies on this topic: Well, it depends a bit even then if an elemental team is recommended for defense. Attempting to make something more rigorous for comparison of the elements (different situations should be accounted for too, since it seems dominatrix is better when the player is low-power and stuff) would certainly be interesting but with me getting nothing out of it while helping enemies, ehh, maybe not right now. Also because I don't yet know how I would set something like that up; there are so many situations to account for when the player level can change and the girl level can change for both the player and the opponent. And how do you select which girls the opponent should have in the comparisons? Each situation would of course have to be modelled across each blessing which becomes unworkably much computation really fast besides not being easily presentable. So it's kind of difficult to make a comparison that really holds water. And I think the playful element might still be a bit busted when modelling hypothetical players, even though now it works properly in normal usage. This is something I do kind of dislike about BDSM, it's very difficult to work with in simulations.

Basically, don't trust me. Someone called me out on that the gem costs for the strategies I simulated were not, in fact, equal. So I made a bit more proper calculator for gem costs, and came up with a new set of strategies to compare. Also, of course there would be some pretty major problem in how the simulations are set up. Yeah, there's always something, and this is not an easy thing to debug. However, I did think to reverse the players' levels and teams. If the simulation set-up is working properly, the predictions should stay the same but shown on the other player. And they weren't. Girl objects weren't even shallow copied, and the order of execution meant this became a problem. Anyway, this problem at least is fixed. Some differences to the parameters of the simulations were made, too. The boosters on the attacker are 2 cordyceps, 1 ginseng and no AME. The opponent remains unboosted. Both players have access to all 5-star common and starting girls at 50 levels below their cap, but no other commons or starting girls. I'm maybe not listing the mythics and legendaries assumed. Both players use the team that has the 7 highest raw strength girls in each blessing. And I do go through all of the blessings and the different percent modifiers too, so you can bet it took a while to compile this table. Analysis: It seems that the previous conclusion has almost been turned on its head. Girl level is more powerful than the previous simulations would have had you believe. But a moderate amount of good legendaries can be just as effective; compare the level 700 12 mythics 12 legendaries roster to the level 750 one. Also going overboard with legendaries might be the reason why level 650 suffers in this comparison, but also it was kind of necessary since there aren't any more mythics to choose from.

I do apologize for this week's lackluster update. Uncharacteristically I was away at reset time so the leaderboard is not very up-to-date. I still won. Also, I clearly overestimated how problem-free my data recording code now would be; I don't think even the previous problem of incorrect attribution of some fights was actually solved, and there's also a gap in data, which noticeably shows up in one of the graphs as a jump in average. To be clear, these problems only affect the stats and team composition graphs. Well, the main thing is that at least I got a cheap first place even though the presentation of that went a bit wrong. This starting week I'll be facing off against my "mentor", @zoopokemon. And even though there certainly are other noteworthy opponents, it might finally be time for that master versus pupil showdown. We even are currently tied in the number of D3 victories, so if either of us wins this league, it'll be a bit of a double-whammy.

Finale for my situation: Main account: 1st stage - shards (and got the girl) Test server: 1st stage - no shard 2nd stage - shards 3rd stage - shards and the girl

Hopefully these are correct; I have to do it from memory now since this recording wasn't started before. Stage 2: 35 shards Stage 3: No shards Stage 4: 35 shards On the test server I know I have 35 shards, but I didn't keep track of if it was in three or two wins.

What you describe is not totally useless, but the boost detection already goes a long ways in achieving that. What your idea does give and what just looking at unboost status doesn't, is the amount of boostedness. But with the boost detection approach you might already be making mental notes of the strong players you would want to prioritize. This simplistic of a model you present will undoubtedly confuse people if it's put into the script since they'll see big changes when they change their teams or whatnot. So it wouldn't be useless, but hardly suited for the general public. Proper analysis of when to fight opponents is a super complicated task, hardly something that's easily implemented within a script (that is, without also maintaining a separate spreadsheet for it). Also it might defeat the fun if there's no place for human decision making any longer. Well, fortunately it's so difficult I manage to get my fun from having tried to solve that problem at least since September 2021, and I still find places for human decision making. After including the blessings change into consideration, making a precise formula started seeming once again pretty futile.

You know what? I might post a bit of something else, too. Rankings I do of the "raw power" of my opponents. Lower means a tougher opponent. That is, if they fought the same opponents with the same AMEs as each other but always with the toughest regular boosters each used throughout the league, this is how I think they'd rank up against each other. But don't take these numbers too seriously; they are just a quick gauge that I produce while making other predictions. The league that ended just now: Mr.Wright: 17.48 (in actuality came in 5th) SJL996: 17.55 (1st) Rarum: 17.85 (not even that far off from SJL, but far enough that I didn't want to risk it), 2nd place Socrates: 19.23 (3rd) League ending on 27nd of January: Rarum: 18.28 (3rd) La Kajira: 18.39 (1st) beraldo: 18.44 (4th) Enpatsu: 18.53 (8th) ... Thulsa Doom: 19.71 (2nd, maybe due to heavy AME usage?) League ending on 20th of January: Socrates: 16.74 (2nd) aizen: 18.15 (3rd) Rarum: 18.28 (1st) hft111: 18.39 (4th) League ending on 13th of January: La Kajira: 16.07 (1st) bla: 17.68 (4th) phyzel: 17.79 (3rd) Rarum: 18.02 (2nd) Before those leagues I don't think the system was mature enough to present. This week I have somebody at 16.68, while I'm only at 17.99. It'll be an interesting match-up since from one other form of measuring opponents I wouldn't expect them to actually play well. We'll see.

Congratulations to SJL996 for their 101st (in-game record-wise) D3 victory. I couldn't have beat you if I tried. My best would have been more or less 7196 points. I did have to do some data reconstruction since some data points weren't recorded properly for some reason. The graphs are going to be slightly misleading for a handful of data points, too. But in the future I think they might be more reliable than before since I tried fixing a long-standing issue. I did see SJL996 getting very good amounts of points and decided against fighting them. It was also a factor that they are one of the very top players. In-game statistics wise I think there are just sygfried01, sliqguy, Kurt and SJL996 who have over 100 D3 first places (but there was a data wipe at one point, I heard). In the Pantheon Stairway SJL is second, only behind Hugh Jerexion (whom I already managed to beat once but I'd be a little surprised if Hugh will allow that to happen a second time). This week I was also further developing a strategy for how I want to deal with blessings changes. I don't think it was a huge benefit this week and it's a bit difficult to even quantify, but every little bit counts.

I managed to squeeze out more numbers from my simulations. I simulated a level 500 player using the level 650 girls with the strategy of always selecting best raw power ones, and got an average of 23.157 against the 7-mythic player across every blessing. So about 30 levels seems to be about how much advantage I can close with careful team and equipment selection. Actually somewhat more, since we've thus far assumed a balanced enemy team. If I fight without an AME, I get about 22.069 for average (at player level 469). And since otherwise we can't compare easily, I redid the level 500 calculation without an AME too, and got an average of 22.30. Another "fun" number is 19.511, which I got with no boosters at all at player level 469. Any regular D3 player already knows that boosters matter a lot, though.

After around 1600 blessings, my simulation had converged to an average of about 23.142 (the last digit isn't very converged yet) for the situation of using my optimal team finder in the level 650 teams vs. the balanced 7 level 700 mythics team situation. For comparison, I got an average of 22.989 when the highest raw power team was always chosen. Unsurprisingly, this is not nearly as significant advantage as having a good spread of girls to choose from in the first place. However, it's already about a whole point more than when using the 7-mythic team strategy. In 13 out of 31 leagues I have good enough records of after awakenings were introduced, 0.95 more in your average would have been enough to bump you up from the top 15 rewards to top 1. That's huge. For comparison, 0.8 points (the difference without my optimal team finder) would have been enough to bump you up from the top 15 rewards to top 1 only 6 times out of the 31. And in 8 out of 32 records, 0.15 (what you gain from choosing your team carefully) was less than the difference between the first and second place players' averages. A-ha! I actually do that personally. And it only takes into account the girls I do have instead of assuming you have every girl in the game. Great minds think alike, eh? Besides those two types of ranking, I can sort girls of each element separately by their max base sum for the current blessing, next blessing and unblessed. And of course sort all girls by their current damage, damage in the next blessing or unblessed (or the same by the max base sum). And by their champion damage by each pose. And by their level, required gems to upgrade to current cap... Well, you get the idea.

Okay, because I don't have a life, I've managed to make some kind of a simulation already. Here's a description of what happens in it: A level 700 mythic team goes up against a level 650 all-round team, and vice versa. This is done in every blessing combination there is, and the average battle score in each blessing is averaged over the number of blessings. I'm always assuming the player has an AME, two cordyceps and a ginseng on, and that the opponent is unboosted. I chose levels 650 and 700 for comparison because 650 is where I'm currently at, so it makes these reasonable points for a free-to-play player to be at. However, the decision to somewhat utilize my stats also means the player level is 469, which might be a bit unrealistic for a free-to-play player to manage a 7-mythic strategy since it costs so much. Oh, well. The girls selected for the simulations are actual ones. For the 7-mythic strategy, they are Matcha, Level up Red, Undercover Valentina, New Year Estelle, Radka, Royal Housemaid and Alexa. I omitted considering the two mythics that came after the awakenings were introduced. I believe this is a pretty good selection of mythics. For the all-round strategy, the mythics selected were Matcha, Level up Red, Undercover Valentina, New Year Estelle, Radka, Royal Housemaid, Alexa, Golden Lupa, High Mage Arcana, Nike and Bunna. As for the legendaries, they are all 5-star legendaries with max base sum of 25.00, except for Treasure Flynn and Eliana. Also various common and starting girls were selected but modelled as having a max base sum of 19 (I couldn't be bothered to use real stats there). I *guess* there's a slight bit of bias toward this side, since I didn't only select legendaries that were introduced before awakenings. I've been feeling that even outside of this specific simulation there's maybe slight error against the playful element in my calculations currently but that's not a substantial enough effect to invalidate the general finding. But beyond that, if we assume I didn't make mistakes in writing this fairly delicate simulation, here are my findings. All-round team's average across all blessings: 22.989 (when the highest raw power team is always chosen) Mythic team's average across all blessings: 22.192 (when using the same 7 mythics in every blessing) I suppose it does make sense. The all-round team also has mythics, only at 50 levels lower. That's a power difference of just 7.7% more for the level 700 mythic team even in the worst case scenario. It would require a very low amount of blessings that affected other girls than what's in the 7-mythic team for that power difference to be enough to justify the all-in strategy. I'm also trying to figure out just how much team switching can help the all-round player. However, after an hour or so of number crunching, the page crashed (after having consumed a lot of memory, hmm). Clearly I'll have to implement some kind of stochastic selection of blessings to test because brute force doesn't seem very workable.

Curses that you would "promise" something like that for me. It's far from trivial to get even remotely definitive answers to the question of which strategy was better. It might take days to set up such a simulation, and there are always going to be some disagreement about its specifics. With that said, I'll see what I can do. I am indeed myself also intrigued, and maybe the time is ripe to produce and publish some data, even knowing it might be used against me somehow.

I decided to calculate what's the point where the gem costs are pretty much identical between the all-in on mythics versus a relatively all-round strategy, assuming again a starting point of level 500 for all girls. It turns out you can get 7 mythics to level 700 and 78 common girls to level 650 for pretty much exactly the gem price of getting 11 mythics to level 650, 26 legendaries to level 650 and 33 commons to level 600. Those don't seem even that far off power-wise since the more all-round one can better take advantage of blessings. Girl XP wise we are looking at 14325438 for the level 700 mythic strategy, and 11858924 for the level 650 all-round strategy. Now let's look at how much additional cost that girl XP would mean. Let's say a market refresh gives an average of 12250 XP. With the difference in girl XP for the strategies being 2466514, you'd be looking at 201.3 market restocks, for a cost of 1409 Nubans or 8457 Kobans. Real money wise that could be bought with about 60 euros. All for the convenience of not having to think about the team all that much. Of course you could push that power difference further to actually meaningful territory by buying gem bundles. Another comparison. Let's say the all-in mythic strategist has gotten to level 750 with their mythics and 700 with commons. How far can an all-rounder get with the same amount of gems? It turns out the all-rounder can get 11 mythics and 27 legendaries to level 700, and 47 commons to level 650. This time the girl XP difference is 4231297, which translates to 2418 Nubans or 14507 Kobans.

I did some calculations to have something more concrete. Let's say you are a player who started with their entire harem grandfathered to level 500. How many gems and girl XP would it require to reach level 650 with their battle teams and level 600 with common girls? Let's compare three different strategies: one with 7 level 650 mythics and 63 level 600 commons, one with 12 level 650 mythics and 58 level 600 commons, and one with 10 level 650 mythics, 25 level 650 legendaries and 35 level 600 commons. The 7-mythic strategy will use 21875 gems to upgrade the mythics and 22050 gems to upgrade the commons, totalling 43925 gems. The 12-mythic strategy will use 37500 gems to upgrade the mythics and 20300 gems to upgrade the commons, totalling 57800 gems. The 10 mythic, 25 legendary strategy will use 31250 gems to upgrade the mythics, 62500 gems to upgrade the legendaries and 12250 gems to upgrade the commons, totalling 106000 gems. Yes. Going all-in on a single team of mythics is much, much cheaper gems-wise (2.41 times). The 7-mythic strategy will use 3212615 XP for the mythics and 3887541 XP for the commons, totalling 7.1M XP. The 12-mythic strategy will use 5507340 XP for the mythics and 3579006 XP for the commons, totalling 9.1M XP. The 10-mythic 25 legendary strategy will use 4589450 XP for the mythics, 4590575 XP for the legendaries and 2159745 XP for the commons, totalling 11.3M XP. So again, cheaper. But not by a whole lot, "only" 1.59 times. If XP isn't the limiting factor, the all-in on mythics gets you much further, maybe enough so that it counteracts the loss from not having blessed girls as often. Or at least, it requires a lot more careful selection of teams and so on with which to battle. But if XP is a limiting factor, I'd hazard to say having some amount of variety in the available girls is worth having that somewhat slower progression. I can't really produce an exact formula to tell where the ideal harem composition is for each girl XP and gem availability, though. It gets very complicated beyond some generalizations.