32 coaches online • Server time: 10:13
Forum Chat
Log in
Recent Forum Topics goto Post DOTP Season 4goto Post Skittles' Centu...goto Post Secret League Americ...
SearchSearch 
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic
sk8bcn



Joined: Apr 13, 2004

Post   Posted: Aug 03, 2009 - 13:40 Reply with quote Back to top

CircularLogic wrote:
sk8bcn wrote:

5,1% under normal conditions, but 0.003% under the formula change.

Quote:
2. Subtract p by 0.5 (transposing the value to be centered around 0 rather than 0.5).


ok -4.997

Quote:
3. Define the distance as the absolute value of p


p distance= 4.997

Quote:
4. Apply a small random factor (adding 0 - 0.02)


skipped

Quote:
5. Normalise the distance to 0-1


That's were I'm blocked. I don't have the standart deviance (of what, actually? games in the round? possible TS-matchup of fictive teams? ...)

Quote:
6. Get the base suitability as 1-normalised distance

7. If the two teams are of the same race, multiply suitability by 0.97

8. If either team played the other in their last game, multiply suitability by 0.94

9. If there are any handicaps in the game, multiply suitability by (1-numHandicaps * 0.03)

10. Scale suitability to 0-1000 (ie, multiply by 1000 and round to an integer).


unfortunately, I can't go further.


The bolded part is the first error. if you subtract 0.5, you get -0.4997. The absolute is 0.4997. The absolute value can range from 0 to 0.5 so to normalize it, you just have to double the value, so that a spread between 0 and 1 is possible. So we get 0.994 - or a siutability of 6. Though I seriously doubt that a win% of 0.003 is right.


well, about the calculation, it's right. 0.4997. However, to me normalization is (X-m)/std dev.

I'm ok if it's |X-0.5| * 2 though I didn't understand this as normalisation.

About the 0.003%. The normal p value is 5,1% but Christer added a calculation line to deacrese the suitability of games over 5TS difference.

_________________
Join NL Raises from the Ashes
sk8bcn



Joined: Apr 13, 2004

Post   Posted: Aug 03, 2009 - 14:04 Reply with quote Back to top

Calculation to get 0.003%:

dT=100* (str1/min(str1,str2)-str2/min(str1,str2))
=100*(105/min(105,90)-90/(min(105,90))
=100*(105/90-90/90)
=100*15/90
=16.667

p=1/(1+10^(dT/70))
=1/(1+10^(16.67/70))
=1/(1+10^0.238)
=1/(1+1.73)
=0.366

P=p^(5^(1-2*r))
P=0.366^(5^(1-2*0.3426))
P=0.366^(5^(1-0.6852))
P=0.366^(5^(0.3148))
P=0.366^1.6597
P=0.18858

P=18.8% so under normal conditions (ok I was wrong here at the start)

mmmmmm, I start to get confused

NewdT=dT*3-10*dT
=16.667*3-10*16.667...
..

..

I m defo confused:

Quote:
- In the above ranking formula, if the normalised strength difference is above 5, use the following instead:
dT = (dT*3) - 10*dT (properly adjusted for who is higher or lower)


Is this right? 3dT-10dT=-7dT

It was dT^3-10dT no?

In order to exponentialy reduce the suitability. Am I right?

_________________
Join NL Raises from the Ashes
CircularLogic



Joined: Aug 22, 2003

Post   Posted: Aug 03, 2009 - 14:25 Reply with quote Back to top

I think there is a typo. It should be 3*dT-10. So every TS point over 5 counts thrice. In former posts, it has been explained that way. So the 15 TS difference between chaos and zons would be counted as 35TS difference. As I`m not sure, how racial factors are factored in when calculating p, I cannot give the win percentage for a give matchup.

From the lower siutability numbers Christer gave, the lowest p ever considered would be around 30% which is strange because it seems to be really high. Going for the TS85 fling vs TS97 CD the numbers would be:
dT=100* (str1/min(str1,str2)-str2/min(str1,str2))
=100*(97/85-85/85)
=100*12/85
=14.1

Which would yield a 32.3 TS difference when modified by dT= 3dT -10 because the difference being above 5. That would yield for p:
p=1/(1+10^(32.3/70))
=1/(1+10^0.462)
=1/(1+2.897)
=25.65%

As the calculation of P(p) is only for CR purposes, we don`t need it here.

So flings have a 1 in 4 shot BEFORE racial modifiers. Seeing that the racial factor for this matchup is somewhere between 32% and 45% (depending how the racial table is applied) that means that p is lower than 25%, which will defenitly result in a siutability score of less than 500. Yet the matchup registers as 590. Where did I go wrong?
sk8bcn



Joined: Apr 13, 2004

Post   Posted: Aug 03, 2009 - 17:50 Reply with quote Back to top

CircularLogic wrote:
I think there is a typo. It should be 3*dT-10. So every TS point over 5 counts thrice. In former posts, it has been explained that way. So the 15 TS difference between chaos and zons would be counted as 35TS difference. As I`m not sure, how racial factors are factored in when calculating p, I cannot give the win percentage for a give matchup.

From the lower siutability numbers Christer gave, the lowest p ever considered would be around 30% which is strange because it seems to be really high. Going for the TS85 fling vs TS97 CD the numbers would be:
dT=100* (str1/min(str1,str2)-str2/min(str1,str2))
=100*(97/85-85/85)
=100*12/85
=14.1

Which would yield a 32.3 TS difference when modified by dT= 3dT -10 because the difference being above 5. That would yield for p:
p=1/(1+10^(32.3/70))
=1/(1+10^0.462)
=1/(1+2.897)
=25.65%

As the calculation of P(p) is only for CR purposes, we don`t need it here.

So flings have a 1 in 4 shot BEFORE racial modifiers. Seeing that the racial factor for this matchup is somewhere between 32% and 45% (depending how the racial table is applied) that means that p is lower than 25%, which will defenitly result in a siutability score of less than 500. Yet the matchup registers as 590. Where did I go wrong?


As I understand it, the P(p) formula is exactely the racial factor:

from Help:

Code:
r = A racial modifier by TS


Code:
There has been a CR tweak, meaning that p in the first formula is replace by P. This new P is calculated using the old p, as follows:

P = p ^ ( 5 ^ ( 1 - 2*r ) )



link to help page

_________________
Join NL Raises from the Ashes
CircularLogic



Joined: Aug 22, 2003

Post   Posted: Aug 03, 2009 - 18:13 Reply with quote Back to top

Ah... should have read better. Just stopped at 'CR tweak' and dismissed it`s relevance for [B]. So.. for the fling-CD matchup it would be:
P=p^(5^(1-2*45.21%))
P=0.2565^(5^(0.0958))
P=0.2565^1.167
P=0.204
Timlagor



Joined: Feb 13, 2009

Post   Posted: Dec 10, 2009 - 04:52 Reply with quote Back to top

Not sure how open to consideration this still is but the thread's open Smile

I would generally prefer to play (though I agree that a right to cancel for the underdog in extreme games would be reasonable) because I see the box going in cycles quite a lot with one activation haveing 3-8 matches and the following one only a couple of coaches only to come round again at the 2nd or third activation. Thus it's often not a case of waiting just 30 minutes for a better game.

I would also like to see a way to weight your matchup towards a specific team while also submitting other teams: you want to play Team A but would rather play Team B(/C/D/..) than play an unreasonable match or not play at all or are willing to play an alternate if it will *significantly* improve the matchups for others but would still prefer your Team A if the matches are "good" anyway.
EXAMPLE: I might have several teams active but want quite badly to get a new Ogre or Vampire (/halfling/..) team off the ground. These are hard to find opponents for since they have TS well below 100 initially. I'd definitely prefer a game angainst any TR100 team to playing with my other teams but this is my time to play BB and I'd hate to miss the activation altogether because there happen to be no matches that the bot likes for me (almost as much as I'd hate to miss a great match with thtat team because i activated too many other teams)
EXAMPLE2: I much prefer playing at high TR/TS but my really developed team has trouble finding matches. I don't want ot miss out on a good match just because there's a 'better' one with a team I care less about at the moment (and it may only be a marginally better scedule by the numbers which will inevitably not be perfect in any case).
EXAMPLE3: I just feel like playing khemri today but playing is the more important consideration.

I see two ways this might be implemented:
- EITHER each player's 'Team A' gets a +X or *X boost to its matchup ratings. This should be pretty simple to implement I imagine. Finding a number that is high enough to matter without being high enough to distort badly could be tricky but I think would be well worth the effort and I'm sure it's not beyond you Smile

- OR the bot gets a minimum criterion for matches (all matches over X* are 'good'; or all schedules over 'X*Coaches entered' are good) and then tries to maximise the number of 'A' teams playing. [there are other variations on this kind of idea but I think there'd be a lot more programming in doing it this way]

Interface-wise you could just give the activation page two columns of tick-boxes that the coach can select.


* I haven't taken the time to really understand the mathematics of this system but eyeballing the examples I reckon any match rating over 850 is pretty good (900 to be safer) and +20 or *1.05 would be a good place to start for the matches of preferred teams.
..not even sure if those figures are in any sense linear for the quality of matches and obviously everyone would have their own opinion of what constitutes a 'good match' and how important it is to play your preferred team(s).

_________________
Time for a new .sig
Timlagor



Joined: Feb 13, 2009

Post   Posted: Dec 10, 2009 - 04:56 Reply with quote Back to top

P.S.: There's no real need to limit yourself to two preference columns but I doubt that there's much demand for more than 3 (+20, =, and -50 spring to mind) and 2 is probably plenty to make everyone happy on this (apart of course from anyoen who thinks the whole idea is terrible Wink).

_________________
Time for a new .sig
CircularLogic



Joined: Aug 22, 2003

Post   Posted: Dec 10, 2009 - 12:40 Reply with quote Back to top

It`s always a tradeoff between coding work, confusion it might cause and gain. So I`m not sure it will be implemented.
Timlagor



Joined: Feb 13, 2009

Post   Posted: Dec 10, 2009 - 15:07 Reply with quote Back to top

Nor am I but more likely if I suggest it Very Happy

I can't see an extra column next to "Activate" saying "preferred" being terribly confusing for anyone but I have no idea how hard it would be to code (though if adding a static modifier is hard I'd be surprised) ...whether Christer thinks there's significan gain is another question altogether however Smile


...any word on why BWR and CR are seperate? (aside from CR being clearly inferior and not wanting to contaminate ourselves with it that is)

_________________
Time for a new .sig
clarkin



Joined: Oct 15, 2007

Post   Posted: Dec 10, 2009 - 15:19 Reply with quote Back to top

One thing I'd like to see is coach's BWR/BBR listed in match reports between B teams. It can go below the CR rather than replace it.
maysrill



Joined: Dec 29, 2008

Post   Posted: Dec 10, 2009 - 15:42 Reply with quote Back to top

clarkin wrote:
One thing I'd like to see is coach's BWR/BBR listed in match reports between B teams. It can go below the CR rather than replace it.


Heh, it would be a nice feature. There are some coaches who play so little Ranked that their CR is really unrepresentative of their coaching level.

It would also be nice to see the various discrepancies between CR and BWR. Mine had been getting to be close to even, but a rash of awful luck has split them up a bit again (was once as low as 3 points, now over 10). Mopping up in Ranked and getting kicked around in the box isn't as impressive.
odi



Joined: Aug 02, 2003

Post   Posted: Dec 10, 2009 - 15:48 Reply with quote Back to top

If we go into other things than the scheduler, the coach ranking by race thingy would be fun to see implemented for B Div.
Calthor



Joined: Jan 24, 2006

Post   Posted: Dec 10, 2009 - 16:29 Reply with quote Back to top

clarkin wrote:
One thing I'd like to see is coach's BWR/BBR listed in match reports between B teams. It can go below the CR rather than replace it.


Seconded.
treborius



Joined: Apr 05, 2008

Post   Posted: Dec 10, 2009 - 21:59 Reply with quote Back to top

clarkin wrote:
One thing I'd like to see is coach's BWR/BBR listed in match reports between B teams. It can go below the CR rather than replace it.


Thirded.
treborius



Joined: Apr 05, 2008

Post   Posted: Dec 10, 2009 - 22:03 Reply with quote Back to top

i have another request for the box:

this team is scheduled in the current b-minor against an oppo who's not decided yet.
after last game, there are some new skills to give to the players.
i don't want to do so just yet, since i want to adapt to my next oppo.

meanwhile i cannot apply with any of my other 11 b-teams since it says "not eligible" instead of "activate" until i have chosen skills for the team in the tourney.

is this a bug? Sad

Edit: i suppose many coaches prefer giving skills no sooner than when they know their next oppo, all those coaches will apparently not be able to play box-games until they have done so...
Display posts from previous:     
 Jump to:   
All times are GMT + 1 Hour
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic