Which numbers to use, anaylsis

[Trinzia]

I ran some numbers and discovered that newer players & players with
low attendance were receiving too high priority over veterans with
100% attendance. Trust me, they noticed.

I attempted to fix this by:

1. Set the MIN_EP to be roughly equal to the number of EP that can be
earned in one average week. All this does is ensure that brand new
members can't bid on "contested" items for 1 week.

2. Set the BASE_GP to be roughly equal to the cost of 1 average item
(300). This is very important because it removes a priority spike
whereby newer members would always win contested items after 1 week,
no matter what. Many of us would not object to allowing a new member
unearned priority once. However, without BASE_GP, the new member might
win 2 extra items, by having say 600:1 priority, while a veteran has
1200:600. Think about how often you get a new member, & pretty soon
your veterans may not be able to win anything, since each new guy wins
2 items first. (And before you say to use subtraction instead of
division - I tried that & it's worse.)

3. Set decay to 10% per week. I do not understand why people say that
decay changes priority. The ratio of 100:50 is exactly the same as
90:45. All this does is change the EP and GP graphs to parabolas
instead of straight lines, so that both approach a constant. Not sure
though, what will happen when someone decays past the BASE_GP. Seems
to me, at that point, EP should keep dropping, while the BASE_GP
cannot, so they would eventually, with very long absences, be returned
to the status of a brand new member, 0:300 priority.

4. Then, it also appears to be necessary to add in a weekly bonus for
players who attend every night, to further correct the system. This
bonus is around 10% of the points earned per week, that must be
returned to those with 100% attendance. I am unsure exactly what this
number should be, but it's a clearly necessary correction.

I will link my Excel file in case you wish to look at it. The numbers
are the same regardless of whether I multiply our EP amounts by 100 or
not (which surprised me at first). But I suppose that, since the
parabola approaches a constant, the ratio to item value does also, and
rounding becomes more important, the longer you use the system.

So, why do I need this correction? Something in the math... that I'm
not getting. I tried doing the EP and GP decay separately in the
Excel, for example, 10% decay on EP and 20% decay on GP, but it didn't
work out the same at all. I suspect the rate of decay should be
related to the number of hours a person attended.

** Do you plan to add in an attendance tracker & offer this correction
as standard?

Here's the file: http://bluemoo.guildhaunt.com/epgpchartX100.xls

Please note, while looking at the chart, it is the pattern of 1 and 0
we should be most concerned with, not the totals. Player A, attending
twice as often as Player B, should win exactly twice as many items.
Note that if you remove the decay or bonus, the pattern breaks.

[Trinzia]

disht replied (moving topic from curse):

That's some interesting analysis lorelye. Instead of the comments here
on curse do you want to take the discussion on the epgp forum?

MIN_EP: If you use BASE_GP then MIN_EP is not necessary imo. With a
BASE_GP set to the GP price of the average item, no new member will be
able to bid on a contested item. If your decay is somewhere around 10%
your 100% raider will have 10x the EP given per week as his total EP
and maybe 2-3x the GP of the average item. Hence the PR of a new
member will be 3-4x lower than the PR of your 100% attendance raider.

BASE_GP: What you say is true, and base GP should be changed overtime.
When we were in SSC/TK BASE_GP was set to 333 (head piece GP value of
T4) and when we hit BT/Hyjal BASE_GP was changed to 410 (head piece GP
value of T5).

Decay: You can do per week or you can do per raid. I personally find
the per raid easier to administer, but the per week is a bit more
fair. Per week is more fair because you might clear Hyjal/BT in 3 days
or 2 days depending on how fast you cleared. Why should the 2 day
clear be decayed twice and the 3 day clear thrice?

Weekly bonus: I do not see how this is necessary. In the excel sheet
the EP bonus is set to 0 for both player A and B.

As for plans there are plans to add a transaction log and an online
system to upload all the information and sync between it and the addon
but available time is the bottleneck. If I had to guess I would say
this is going to be something for WotLK.

[Trinzia]

In the Excel sheet, I left the weekly bonus at zero and marked a few
numbers in red. The pattern should be, with Player A attending twice
as often:

1 0
1 0
0 1
1 0
1 0
0 1

Notice that in a sequence of 60 wins, the pattern breaks twice. This
may seem small, but is multiplied by every member with lower
attendance. So you might start to see the veteran competing with 4-5
members with lower attendance, and then it would become every sixth
turn or so, that he's "losing a roll".

Also, it is worse, the more you lower Player B's attendance. In the
spreadsheet, try setting Player B's attendance at 1/4 of Player A's.
Then Player A should win 4 times as many items as Player B. The
disparity increases.

Now it seems odd that my patchwork solution is to first decay everyone
10% at the beginning of the week, and then to return ~10% EP for
perfect attendance at the end of the week. I did this only because it
works; I cannot think of any math to explain this, or a more proper
solution.

PS - I don't know the meaning of "WotLK".

:) Lorelye

[Andrew]

Why not simply increase the min EP And the base GP? At some point new
raiders will have priority over veterans, this is one of the
underlying principles behind EPGP. Min EP and Base GP exist so you
can delay the point at which a new raider will pass a veteran. When
the new raider does get loot, their PR will drop dramatically and it
will be longer before they get another piece of loot, relative to the
veteran raiders.

> PS - I don't know the meaning of "WotLK".

Wrath of the Lich King, the next expansion. :)

[Trinzia]

Because, this is not the issue I am trying to address.

Look for the difference in these patterns:

1 0 1 0
1 0 1 0
0 1 0 1
1 0 1 0
1 0 0 1
0 1 1 0
1 0 1 0
1 0 0 1
0 1 1 0

Do you see the skipped turn in 2nd column? It is marked in red in
spreadsheet.

[Alkis Evlogimenos ('Αλκης Ευλογημένος)]

I think there is some error in your spreadsheet somewhere. Since you don't use $D$3 for static references it is a bit hard to debug and edit. I created another one:

http://spreadsheets.google.com/pub?key=pKq15kNAMau13pzV9O3zwDQ

I believe you will have the option to copy this to your account (if you have a google one) and edit it at will.

What you will find is that by increasing BaseGP you can turn the table in favor of Player A. That is not true for Player B though. At no value of Base GP is Player B receiving more than hours played B / hours played A * items received A items.

I added rows up to 60 weeks. You can add more if you want. Now the thing to find is that the system is fair in the sense that the ratio of EPs given result in the same ratio of items looted. +/- 1 errors should not be counted simply because the system deals with single items. There is no such thing as giving 1/3 of an item to player A and 2/3 of an item to player B on every raid. If you find a configuration where you see 1 item more on player A or B just add 1 or 2 more weeks. In either 1 more week or 2 more weeks things will be exactly as you predict. Why is that? It is simple: Player A should receive twice as many items as Player B. If you have 6 items this is simple: 4 to A and 2 to B. But what if you have 5 items? Depending on who looted and when, it might be 3 to A and 2 to B or 4 to A and 1 to B. In the former case, the next week A will receive an item and make it 4:2 and in the latter B will receive an item making it 4:2 as well.

--

Alkis

[Trinzia]

On Mar 26, 2:50 pm, "Alkis Evlogimenos ('Αλκης Ευλογημένος)"

<al...@evlogimenos.com> wrote:
> I believe you will have the option to copy this to your account (if you have
> a google one) and edit it at will.

Since I cannot edit your file or see your formulas, this doesn't help
me to isolate the problem.
Thanks.

[Trinzia]

PS - I didn't know that trick for Excel, and have overwritten my file
with it included:
http://bluemoo.guildhaunt.com/epgpchartX100.xls

I also changed it to be more obvious by making Player A attend 4 times
as much as Player B. I marked each sequence in red, where Player A
skips a turn, ie, when the pattern is not 4:1 but 3:1 in wins. This
is 8 times in a series of 60.

[Alkis Evlogimenos ('Αλκης Ευλογημένος)]

You will have the option to edit it if you have a google account. Anyway, I have exported this to an excel sheet for you to play around with.

--

Alkis

[Trinzia]

Again, I cannot view this file correctly, since it's referencing
column B, which somehow I do not have in the spreadsheet... but... it
doesn't matter that much, it will still suffice to show you what I
mean, even though I can't see all your formulas. The first disparity
is at line 34 in your spreadsheet.

Now, change Player B's average EP to 500, and the pattern changes
more. Should be 4:1 ratio of wins, but it is not even close. The
disparity happens more often, as you have a greater difference between
the attendance of two players.

~Lorelye

[Alkis Evlogimenos ('Αλκης Ευλογημένος)]

If you change Player B's average weekly EP to 500 and leave the starting the same you have a disparity for the following reason:

Player B with weekly EP of 500 starts with 5000 EP and decay 10%. This means this player has been getting 500 EP for a while and reached the plateau.

Player A with weekly EP of 2000 starts with 10000 EP and decay 10%. This player hasn't been on the plateau. Given his starting EP of 10000 it means on average so far he has been raiding half as much for a weekly of 1000 EP.

In the end Player B gets 14 items and Player A gets 46. This is not 1:4 as you would expect because of exactly the reasons above. You are comparing apples to oranges.

P.S: Column B is a hidden column (as is column G). You can unhide it somehow in excel I don't remember how off the top of my head.

--

Alkis

[Trinzia]

Hi,
Still don't understand. Start them both at zero EP and it's the
same. What is apples and oranges?

On Apr 1, 3:31 pm, "Alkis Evlogimenos ('Αλκης Ευλογημένος)"

[Alkis Evlogimenos ('Αλκης Ευλογημένος)]

When you change the starting EP to a value that is not normalized for both you cannot compare properly.

But lets take starting EP 0 for example. With weekly EP of 2000 and 500, item GP 400 and base GP 300, A gets 47 items and B gets 13. This is not exactly 4:1 and the reason for it is BaseGP. If you change Base GP to 600 this will get you exactly 48 items for A and 12 items for B. You just need to adjust the variables a bit to get exactly what you want. For more pragmatic ratios like 2:1 attendance pretty much any value is gives you exactly the same ratio of loot as for attendance. For extreme cases of 4:1 you need to choose the numbers a bit better that's all. The reason behind this is decay. Decay while it saves you from hoarding, it reduces the effort of the most attentive members more than the effort of the least attentive ones. Base GP does exactly the opposite. The higher Base GP you have the more you decay the loot penalty for people that get more loot. This way you balance things out and have a happy guild. I would be nice if someone comes up with a patch to estimate average weekly EP and GP for the guild and suggest a proper BaseGP on the fly given the current Decay. That would be ideal.

--

Alkis

[Trinzia]

I agree. More dynamic numbers here would be incredibly nice!

I'm reluctant to change the base GP because it will affect mostly:
1. Players who have the best gear, who therefore had very little to
roll on during our 2 months in SSC/TK (haven't killed Vashj dangit),
will decay to the base GP too quickly, and start losing priority. Why
penalize your best geared players?
2. New members would take twice as long to get priority. I mean,
until their EP crosses a certain level, it's being decayed, while
their GP is not, so they're losing priority for a while until they
spend some points I guess. Increasing min_GP to 600 would increase
that delay in priority from 1 week to about 2.5 weeks, I think.

I use EPGP because my guild does not require attendance. I actually
do have players who attend only 1 night per week -- mostly standby
healers etc. I also have players who attend 6 nights/week. And I
have some in the middle, for example, our rogues all attend about 60%
in rotations. So this small issue when magnified by 50-60 raiders in
the guild, and another 50-60 friends/family who are either gearing up
to raid or waitlisted, becomes very important to me. I'm very
concerned with being absolutely mathematically fair. As I see it,
that's my job as guild leader.

I can understand why this hasn't come up that often for other guilds
though. I think most hardcore guilds have required attendance, and
most casual guilds are too worried about filling the raid seats to
think about much else. My guild might be one of the only "hybrid"
guilds out there that actually raids 5-6 nights/week (have to, to make
enough raid seats, people's friends keep joining, lol).

Thanks for listening. :)

[JebJoya]

Hi, sorry for rather barging into this topic, but I'm interested by
your points in this last post.

Recall in 1) that the point of EP/GP is to reward effort by members.
So, if a veteran in great gear does not get a chance on any items for
ages, he can still keep his EP up as his GP falls by going on raids
etc. Once the member has decayed to his base GP, assuming his EP is
similar to the EP of other people he's raiding with, he will have
priority over them (i.e. the well geared who only need minimal drops
will take priority on those items over the less well geared who have a
big pile of items they need to get and thus have higher GP).

Regarding 2), this is simply a decision that your guild need to make,
base GP is useful essentially so new players have to build up some EP
before they can start winning items (unless uncontested of course).

Finally, going back to your original points, I don't think there's an
issue with people with, say, 4:1 ratio of times played getting
different to 4:1 ratio of items, I think the problem you've had is
that you have a player who jumps from a large number of EP per week to
less (the higher attendance guy), which causes some minor
inconsistencies. I'm not 100% sure on this, but I'll spend some time
on the maths to try to make a sensible reasoning...

Jeb