Part 3: Do Crescent Upgrades Screw Over Small Clans?

Welcome back to the ever-expanding description of our incentive overhaul.

Part 1: Activity Tiers and XP
Part 2: Crescents
Part 3: Do Crescent Upgrades Screw Over Small Clans?

Part 1 went pretty smoothly, but after Part 2, a bunch of folks had a Problem. The concern is as follows:

• Because crescent upgrades kick in at 10/15/20/25 participants, for clan competitions, crescent upgrades will be less common in smaller clans than in larger ones.
• Because upgraded crescents give more XP, upgrades will award more XP to members of larger clans than to members of smaller clans, creating a systemic disadvantage.
• Therefore, members of large clans have a built-in XP advantage over members of small clans.

Or, as one CON put it:

My biggest fear with this change being implemented is this: Member A in my unit is sitting there, and they're watching themselves fall farther and farther behind member B in the big clan. And they don't know why, all they know is that they are (because most folks don't dig into the rules). They feel like the problem is them, not the system. That they aren't good enough. They keep having negative interactions with the club, not knowing it's the system that's making them move forward slower. And so they just drift off and leave. Systemic disadvantage is really good at making people feel like they're shit.

Whether or not that’s actually true is a complex question to answer, because there are many factors to XP and we’re trying to predict the future. So let's break that down into separate questions that are easier to answer and we will work our way up to the crescent issue.

Question 1: Does having more participants in a competition generally give you more XP?

Question 2: Based on what we know about activity, should we expect larger clans to get more crescent upgrades?

Question 3: What is the impact if they do?

This is going to be a long report for two reasons. First, predicting how XP will work across thousands of submissions to hundreds of comps by hundreds of members is a complex question and has a complex answer. Second, we already gave the short answer and people were not comforted by it. The people demand rigor, and so rigor they shall have. I have put the key points in bold and code blocks that turn the text pink and there is a conclusions section at the end, so if you don’t want the full treatment, hopefully it’s structured in a way you can effectively skim.

Because it’s so long and dry, and because it gets a little heavy at the end, I’ve taken the liberty of including “precious photos of baby birds” that the Audubon Society assures me are “just what you need” through the report body.

I'll also publish my spreadsheet here, with the caveat that I don't expect anybody to actually look at it, so I haven't spent any time trying to make it look pretty or easy to read. You shouldn't need to look at the spreadsheet to follow this report. But if you want the same set of numbers I used, have at them.

For our first question, we're looking at how competition XP behaves in general. I'm going to call this part the Theory, because it should accurately predict what real world data will look like. In the spreadsheet, that's on the first sheet, “Abstract Data”. If you look at the spreadsheet at all, this is the part to look at.

For the other two questions, we're going to look at a set of all participation data for a six month period in the first half of 2025, because that's the data I had available and already cleaned up and ready to use. I'll talk more about this in the next report, but the gist is that I calculated the actual XP awarded under the old system and the XP that would have been awarded for that same activity under the new system. We're going to call this part the Model, because I'm modeling how XP will change under different policies. The stuff you care about is on sheet 2, “Clan Data, H12025”, with the raw data in the two other sheets.

1. Does having more participants in a competition generally give you more XP?

The short answer is No, generally more participants will get you less XP.

The key word is generally. Again, the concern is a systemic disadvantage. That means accounting for your overall pattern of activity across many competitions over time, including everything from your best work to token entries barely eligible for participation credit. And it means doing that for everybody.

Your actual probability of getting a crescent will vary too much to be predictable, because it depends not only on how good your entry was, but who else entered and how good they are and how much work they put in. As long as there were more participants than crescents available, the precise odds will be different and unknowable every single time. So we're going to work with averages and abstracts for this question.

If there are 1-3 participants, your odds of getting a crescent are 100%. That drops to 43% if there are 7 entries, or 3 in 7, because there are three crescents for seven entries. After this point, your odds never get better than 43%. Your odds of placing are worse if there are 8 or 9 or 10 or 15 or 25 or 50 entries. Again, those odds will probably not be spot on for any specific competition, but it's a generally true statement about how placement in competitions works.

Another way of thinking about this is by the average XP awarded across everyone in the comp; i.e. (Participation XP * # of entries) + (sum of all crescent XP) / # of entries. But this follows the same trends. On average, people do better in comps with low participation because if there were only 5 entries, most people got a crescent, and if there were more than 6 entries, most people did not, and crescent upgrades will not change that. On the whole, entering low-participation competitions to farm reliable crescents is still the best XP farm. It's not as good as it was because those are Cr-Es now and not Cr-Rs (for clubwide) or Cr-As (for clan comps), but it still beats a hard fight.

Now there are two exceptions to the "more people entering always makes you less likely to place" rule. The first applies to everybody, and that's that if the comp gets a crescent upgrade, your odds get a little better from the extra medal before they go back down again. And I do mean a little better. Your odds at 9 entries are 33%, at 10 entries they are 40% because a fourth-place medal is on the table, and then by 12 entries they are back down to 33% and keep falling. By the time the full set of 7 crescents is up for grabs, your odds of getting any of them are down to 28%, or 7 in 25. So all that means is that it’s slightly better to be in the clan where the average comp gets 10 entries than a clan where the average comp gets 9 or 11 entries, and that’s not something roster size will do for you.

The exact numbers are a little more complicated for average XP than the straight probability, and it’s true the average XP per participant is 2-4 points higher at 25 entries than at 10 or 15 entries. But that’s not because XP values are trending up, it’s because we keep throwing more XP out to try to flatten the decay. The first two upgrades make the average XP the same at 6, 10, and 15 entries, so maybe the Cr-D and Cr-R values are marginally too high, but for perspective the average XP at 20 or 25 entries is still lower than the average XP at 5 entries. We’re talking about at most 4 XP here for a T1 comp. Simply getting one more or one fewer person to throw a token entry in will have a bigger impact regardless of how many total participants that puts you at.

The second exception is that you are, and I say this fondly, a freak. You don’t play by averages. You are somebody who will consistently beat your clanmates, regardless of how many there are. And we have had those people around before. For cluster races, it was Warp. For Jedi Academy brackets, it was Blade. For ACC ladders, it used to be that Mark would just do twice as many matches as anybody else and win by default. If you are the only dedicated artist or writer in your clan, your clanmates might as well be there to spoon feed you placement XP. Now being a freak is relative; you might be the best artist in Vizsla by a mile but fourth- or fifth-best in Odan-Urr. For that person, being the big fish in the small clan is better XP because placement will be much more reliable. Because if you always place because you’re the only good artist in your clan, adding another 15+ people who want to do art comps to your clan is probably going to change that.

But if you are the genuine best in the club, or even in the top 5, yes, you want as many people as possible to enter competitions because even as the average person's XP goes down, your 1st place XP goes up. But frankly I'm not going to cry for those people if their clan comps don't get upgraded, because they can and should be more than making up the difference from clubwide competitions.

Nevertheless, even freaks still regress towards the mean. Even if you are the person who will always wins your small clan’s fiction comps, if you want more XP, you are going to enter other competition types besides fiction. The more XP you want, the more of them you’ll do. The more of them you do, the more your overall competition placement record will average out, and therefore the more you will be impacted by the fact that the average participant gets more XP if there are fewer participants.

So the long answer is that relying on crescent upgrades to get better XP is an unpredictable edge case unless you're so good that you know you will place, in which case you're already raking in so much XP that I do not want to hear your (probably) EL2 ass complain about not having enough clanmates to steamroll.

Anyway, that’s the Theory. Let’s test it against reality to see how it holds up.

2. Based on what we know about activity, should we expect larger clans to get more crescent upgrades?

The short answer is Kinda, but it's messy.

It's self-evidently true that if the CONs of CSP (25 members) and COU (52 members) both run a clan competition, CSP does not have the numbers to reach the maximum crescent upgrades without Rayne participating in her own competition, while COU would only need half the clan to show up. So it's at least true that this is possible, and that does sound like a bad deal for Scholae.

Now I was going to talk about the due diligence we did in Report #3, and I still will in (hopefully) Report #4, but part of that was that I pulled all of the actual participation data from the two most recent quarters at the time, which turned out to be the first half of 2025. That's 10162 entries across 876 competitions, 512 RP participations, 38 ACC participations, and just over 50k clusters worth of information about how DBers behave in real circumstances. I calculated the XP that was actually awarded, and then I calculated the XP that would have been awarded for the same activity under the new system. And I will talk more about that later, but the short story is that I had that data handy and ready to work with. Nobody needs to, or should, speculate about what hypothetical competition activity looks like when I can tell you what it looks like for real.

This is about if size matters and if your girlfriend is lying to you, so I took the average sizes of the clans for Q1 and Q2 2025 and from biggest to smallest they were:

Clan	Size
Odan-Urr	62
Arcona	44
Vizsla	41
Taldryan	35
Naga Sadow	29
Plagueis	28
Scholae Palatinae	28

Just eyeballing that, you might expect COU to have twice as many comps get upgraded as CNS/PLA/CSP. I’d expect to see a big difference between COU and everyone else, ARC+VIZ+TAL in the same ballpark, and CNS+PLA+CSP bringing up the rear. But seeing what you expect to see is boring, so thank God we don't.

The first takeaway is that the average competition is not getting upgrades in any clan. Across 7 clans and 5 activity tiers, the only things with an average of 10 or more entries were:

• T5 comps in Naga Sadow (22)
• T5 comps in Odan-Urr (13)
• T4 comps in Scholae Palatinae (10)

Clan	Tier 1	Tier 2	Tier 3	Tier 4	Tier 5
Odan-Urr		8	3		13
Arcona		7	4	2	4
Vizsla		7	5		9
Taldryan		2	5		6
Naga Sadow		6	5		22
Plagueis		9	3	7	5
Scholae Palatinae		6	2	10	5
DJB-Wide	36	18	7	39	18

No Arcona. No Vizsla. COU, ok, obviously, but CNS and CSP? This is unexpected and we will need to dig in to understand what is happening.

(For club-wide competitions, the average for T3 comps was 7, for T2 and T4 comps was 18, and for T1 and T5 comps it was in the 30s. Club-wide comps frequently get upgrades, clan comps do only rarely even in big clans.)

CSP’s T4 participation average is very easy to explain. That was a single container comp where the crescents ended up valued at T4 under the new algorithm.

COU had 24 T5 comps. 10 of them were either the Scimitar of Lord Hoth, a clan-wide competition where anyone who was active that month is given participation credit, or the Lord Protector, a very similar competition for House Hoth that isn't run anymore. Those comps averaged 25 participants. The other 14 T5 comps averaged 4 participants. So this is limited to a specific monthly competition format that’s stuck at the least valuable Tier and can’t really scale up. Jack could change the rules to fit the requirements for a higher tier, but then it wouldn’t get so many participants.

CNS is in a similar situation because they had a similar monthly competition, the Order of the Blackguard, and it averaged 25 participants. The only other T5 comp the clan ran during this period got 3 submissions.

So unless activity patterns change in a big way, crescent upgrades for clan comps aren't going to be a normal thing unless it's a competition where the summit manually enrolls every member who qualifies.

But there doesn't need to be a discrepancy in the average competition for there to be a systemic disadvantage. Averages can be deceptive. The average of 1, 1, 1, 1, 1, 1, and 99 is 15. They're good for a gut check and in this case my gut, at least, did not make an accurate prediction. Let's look at all the comps that did get crescent upgrades and where they were:

Clan	Tier 1	Tier 2	Tier 3	Tier 4	Tier 5
Odan-Urr		11	2		10
Arcona		1	2	2	1
Vizsla		1	4		10
Taldryan		0	1		0
Naga Sadow		0	1		6
Plagueis		0	2	1	5
Scholae Palatinae		0	0	1	3
DJB-Wide	23	25	20	39	235

So… it seems that size doesn't not impact crescent upgrades. COU does in fact have the most at 23 compared to VIZ's 15, but if we weed out the Scimitar/Lord Protector, it's 13 to 15. ARC is bigger than Viz but only had 4 comps with upgrades. The three small clans, CNS, PLA, and CSP, had 7, 8, and 4, respectively, though again Sadow's were all the Blackguard.

It's also worth remembering that the clans don't all run competitions at the same rate. COU ran 127 clan/house/BT comps during this period. TAL ran 11. So there's a spectrum. Let's look at what proportion of each clan's competitions would've gotten crescent upgrades, going from largest clan to smallest:

Clan	Size
Odan-Urr	18.11% (11.43% excl. Scimitar/Protector)
Arcona	11.43%
Vizsla	26.32%
Taldryan	9.09%
Naga Sadow	43.75% (6.25% excl. Blackguard)
Plagueis	13.56%
Scholae Palatinae	6.25%

So, excluding the COU and CNS leaderboard comps, the two clans with the best record of getting crescent upgrades are two of the biggest, COU and VIZ. Except Viz was significantly better at it than COU despite being only 2/3rds the size. Plagueis, which was tied for smallest clan in our sample, was the third-best. If we don’t throw out any comps, then it’s CNS, VIZ, COU.

So I think we can be pretty confident that roster size does not translate directly to crescent upgrades. This is not really feeling like a systemic disadvantage to me, but it is nevertheless still true that the 3 biggest clans got more crescent upgrades than the 4 smallest clans. Let's proceed to the next question...

3. What is the impact if they do?

Short answer: Small, uneven, and not what some people have been predicting.

Ok, so it's at least plausible that there is some discrepancy in outcomes here, and we can dig into that to see how much the difference in crescent upgrades actually accounts for it later. First, I want to know how big a discrepancy it is.

But before we even get into that, I just want to put this in perspective. Modeling the first half of 2025, we would've given out 510,641 XP for comps/RP/ACC/clusters, and an unspecified other amount for administrative tasks, SA, society advancement, scrolls, and non-site XP. 120,078 XP of that would've been from crescents. 30,907 XP of that would've been for clan, house, or BT crescents. So clan/subunit placement only accounts for 6% of the XP we know about. And only 17% of clan/subunit comps qualified for a crescent upgrade. So when we talk about the XP from upgraded crescents, we're talking about 1% of the pie, divided among about 265 clanned members. Somebody said this was an argument about pennies on the dollar. No. No, this argument is about one penny, divided two hundred and sixty-two ways. And yet Electorate members have threatened to quit the club over this. I have been threatened with the Chamber. At least one person has been reported to the Chamber.

So, and I know this sounds silly coming from the person with the 12k row spreadsheet of internet points, maybe some of you should rethink how you approach your leisure time.

ANYWAY! We know how much placement XP we gave out per participation per tier per clan. Is there a discrepancy between Palatinaeans and Odanites?

YES!

HOW BIG IS IT?!

10, 6, or 0 XP PER ENTRY FOR TIERS 2, 3, and 5 RESPECTIVELY!

A 10 XP GAP PER ENTRY? HOLY SHITBALLS!

RIGHT? WON'T SOMEBODY THINK OF THE POOR ODANITES?!

...wait, what?

Yeah, COU got 15 XP per Tier 2 competition entry and CSP got 25. It was 23 versus 29 for Tier 3. They got the same for Tier 5.

For Tier 2 comps, Taldryanites got an average of 101 between the two of them, because they only had two T2 entries so both got a crescent. CSP was the runner up with 25, followed by CNS. COU was at the bottom.

For Tier 3 comps, CSP was on top again at 29, followed by COU at 23 and CNS rounding out the bottom at 10.

For Tier 5 comps, Taldryanites got 20 XP per entry, Arconans 11, and Vizjizzles the least at 6.

I didn't count any clan comps as T1 and only PLA and CSP had anything that came out to T4.

Clan	Tier 1	Tier 2	Tier 3	Tier 4	Tier 5
Odan-Urr		15	23		8
Arcona		19	18		11
Vizsla		21	16		6
Taldryan		101	18		20
Naga Sadow		24	10		9
Plagueis		15	15	7	9
Scholae Palatinae		25	29	12	8

In other words, the Model confirms the Theory. For most members, in most competitions, having 62 members to throw at a competition gives less XP than having 28 members.

Well, for the same number of competitions. Bigger clans tend to have bigger summits and subunits and thus tend to run more competitions, which means you can do more competitions, so if you’re highly active you might still come out ahead even if it means you’re getting less XP per entry. Also, bigger clans have subunits, so if you’re doing house or battleteam comps, you no longer have the disadvantage of the huge roster to compete with.

Factors That Would Change Our Analysis

It's best practice to think about what new information would change your conclusions, so let's do that.

For starters, because the new system doesn't discourage clan and subunit competitions like the old one did, we should expect to see more clans doing more of them. That's a big shift in competition behavior from the data we used in the Model. Based on the Theory, I don't think it will upend our conclusions, but it's possible and that's something I'm going to have to check once we have fresh data displaying the new behaviors.

Second, clan sizes may change. One of the red flags for people was that our smallest units are about the same size as the threshold for Cr-Ds, 25 members. Units could shrink even more, and a systemic disadvantage that doesn't exist for a unit of 25 may appear for a unit of 20. Personally, if the clan is that small, I think you have bigger problems, but again, I have to concede the possibility. Conversely, maybe there’s a sweet spot where a unit’s active population, whether we’re talking a clan, house, or BT, tends to produce better outcomes, and I just haven’t looked at a unit that size yet.

Third, maybe our Model turns out to be kinda useless. After all, we did change our incentive structures with the intent to change member behavior. So ideally member behavior will change, and the more it changes, the less predictive a Model based on old behavior will be. The fact that the Model and the Theory ended up reaching the same conclusion means I don’t worry too much about it, but that said, if our new reality matches our Model exactly, this whole project was probably a waste of time.

Fourth, it's possible that clan summits will figure out how to intentionally manipulate this system to leverage crescent upgrades in an unfair way. Now, I'm at loss for how they'd actually do that. You could run fewer comps to try to concentrate participation, but that gives significantly less XP to everybody than just running as many comps as people will do and spreading the activity out. I guess there could be some sort of weird collusion to ensure any comp that's close to a crescent upgrade gets an extra bullshit entry or two to trigger the upgrade? But this is a lot of work for very minor gains for a slim minority of members, and anything that's actually effective will either be a) cheating or b) the kind of good leadership they should be doing anyway

But in all of these cases, we're not talking about information I can just go pull and look at now. It's information that doesn't exist yet. We're just going to have to let this shake out in practice and I can, and will, follow up in six months or a year or whenever the new patterns are established. I already do quarterly analysis for the GM and DGM so that's not asking a lot of me.

Analytical Conclusions

I’m not bolding this whole section, but this is where you should stop skimming.

1. For most people, more participants in a competition results in less XP over time, not more.

2. The exception to that is people who are consistently able to place. But if you're consistently able to beat 24+ people in a given competition, you don't need to roll your clanmates because you're going to get better XP from clan-wide competitions, especially T1, regardless of what your clan is like.

3. If you can consistently place in clan comps but not club-wide comps, not getting Cr-Ds because you don't have 24 clanmates to beat isn't unfair to you. You're not complaining that you don't have enough competition. That competition is there in the club-wide comps. You're complaining that you don't have enough lightweights to farm. If you're 1st out of 10, we're not going to change the rules to give you the same reward as someone who's 1st out of 25 just because you can imagine having another 15 people who all just happen to be worse than you are.

4. Even for Odan-Urr in H1 2025, which was a very large clan with a good participation ratio that ran way, way more clan/subunit competitions than anybody else, crescent upgrades were just not that common. Even if we scale up how many clan comps they run, they weren't popping off at significantly greater rate than other clans. Crescent upgrades are something we'll likely see in club comps, which have a much much bigger pool of competitors than even the biggest clans.

5. When we do look at how placement XP plays out in real clans, the small clans have more of an advantage than a disadvantage. Fewer competitors means more crescents for the people who do compete, so for the average DBer, a small unit will get them more XP than a large one.

6. If I were the COU CON and trying to maximize XP for my giant roster, I wouldn't be trying to run 30-entry clan comps. I'd be pushing comp organization down to the QUAs and BTLs to get more crescents to more Odanites. And that's great! If there's an imbalance, we want it to hurt the bigger clans because they can get around it by running things at the House level. Just like clan comps aren’t worth less to run and participate in than club comps anymore, house and BT comps aren’t worth less to do than clan comps. For years it’s been an open question what the point of having subunits is, and this is one answer.

Darth Strix