pizzamogul
Joined: Jun 13, 2005
|
  Posted:
Nov 18, 2008 - 15:25 |
|
So, if I win and get my BR up I won't get to face new coaches not on IRC with connection issues taking 4+ minutes per turn while grumbling about rough play and/or ball control? I will instead be matched against coaches of similar experience, play style and ability???
Bummer. |
_________________ "Don't expect mercy."
-Woodstock |
|
CircularLogic
Joined: Aug 22, 2003
|
  Posted:
Nov 18, 2008 - 15:36 |
|
OK.. to clarify how the scheduling works I will try to work out the perfect match for a set of coaches. As I think BR and BBR are similary calculated as the CR, I will put their numbers into the same order of magnitude.
Lets say Peter, Jim, Pauly and Olli are all supposed to have an even game and for ease of understanding they are all playing the same race and the stronger coach is having a team with TS150 and noone is getting a handicap.
Peter: BR = 170 - BBR = 140. He is good at winning, but because he doesn`t foul too much and plays only softer races, so he suffers more cas then he deals.
Jim: BR = 165 - BBR = 160. He bashes and fouls to win and he is good at it.
Pauly: BR = 145 - BBR = 180. He likes the splat sounds players make when they die and gets sometimes abit carried away so he doesn`t win that often.
Olli: BR = 135 - BBR = 135. He recently joined FUMBBL and because he had ruled his local league he decided to join [B] right away
So Peter brings a TS150 team, then the scheduler thinks, that Jim should have a TS157 (909) team for that, Pauly should have a TS175 (857) team and Olli should have a TS 183 (897) team. The numbers in parenthesis are the highest scores possible and that is before last match and racial penalty as well as without the randomization. Note that without the latest tweak in the scheduler, these numbers would be TS 161/215/239 for Jim/Pauly/Olli.
Lets see what the scheduler thinks best for Jim... Pauly should bring a TS170 team (877) and Olli should have a TS 179 (979 - the only really good matchup).
This leaves us with the matchup Pauly vs Olli. If Pauly has a TS150 team, then Olli should have a TS 163 (886) team according to the scheduler.
General thoughts about the result:
For me surprising was the fact, that high BBR doesn`t offset low BR - I didn`t read that out of the formula. That means that people with high BR and low BBR are extremely unlikely to be matches against people with low BR but high BBR as the highest possible score is getting lower and lower the greater the difference. I think that is a plus for the scheduler. You meet more people that play like you. On the other hand the system will continue to throw harder and harder opponents at you as long as you increase in BR or BBR. That means that if you try for a high BR, you will still win only 50% only that each game is really really hard. I think a problem here is that the difficulty to overcome a TS difference is not increasing in a linear fashion, but rather exponential. That in turn means that while the system will give a fitting handicap in a BR-wise close match bringing it to about 50%, it will give too high of a handicap in matches with a larger BR difference, bringing it to significantly under 50% chance for the higher BR coach.
All in all after looking at the suggested TS differences for different BR-spreads I`m thinking about rejoining [B] because I liked the ready availablility of even matchups when everyone was around the same BR. Now that the TS influence was decreased, most realistic matchups seem to be playable.
In the end I would like to suggest a few tweaks to the formula, that would improve match making in my opinion:
1) remove the modifier for mirrormatches
2) add dBR and dBBR to the distance calculation. Example: distance = sqrt(P*P + bashP*bashP + dBR/200*dBR/200 + dBBR/200*dBBR/200) - this will make it more likely that coaches with similar abilities are paired.
3) weight dTS with an exponential factor. At the moment dTSweighted= dTS [+ (dTS-5)*2 if dTS>5]. I`d suggest: dTSweighted= dTS^1.5 [+ (dTS-5)*2 if dTS>5] |
Last edited by CircularLogic on %b %18, %2008 - %15:%Nov; edited 1 time in total |
|
PhrollikK
Joined: Nov 04, 2006
|
  Posted:
Nov 18, 2008 - 15:37 |
|
Pirog wrote: | To make a rather exaggerated comparison, winning a sprinting contest over Usain Bolt if he could only use one of his legs wouldn't feel very cool or fun, even if I'm sure it would be challenging to him. |
Well I think you are missing the point in that it's supposed to be a challenge for the both of you. Sure you won over Usain Bolt with him using one leg, but at the same time you run as fast as Usain using only one leg.
On the downside, I am disappointed with all the basher coaches out there. I really don't have anything against bashing teams. But the current TS calculation favours slow high AV teams and faster low AV teams are imo penalized unduly in having to play teams with similar TS but imho the real Team Strength is very different. When the BBR only weighs in at less than 1/3 I can't see it alleviating this. Which is why I retired my skaven and created Orcs |
_________________ Visit my MiniCity
Mega Punk!
Malicious Rex, Daemon Adoptable |
|
Optihut
Joined: Dec 16, 2004
|
  Posted:
Nov 18, 2008 - 15:38 |
|
Reisender wrote: | since everybody ignored it so far, i would like to come back to my first post myself (last one on page one).
the question i ask there is: are you sure that the current scheduling as christer decribes it really tends to give TS-advantadge to weaker coaches? actually i do not think that is true, but i am not sure... (please see first post for details) |
I'm not so sure either, as I faced a stronger team (131/138 vs 162/158) with a higher ranked coach (CR 161.94 vs 177.3) in my last game. It was still fun, though. |
|
|
nin
Joined: May 27, 2005
|
  Posted:
Nov 18, 2008 - 15:47 |
|
Reisender wrote: | since everybody ignored it so far, i would like to come back to my first post myself (last one on page one).
the question i ask there is: are you sure that the current scheduling as christer decribes it really tends to give TS-advantadge to weaker coaches? actually i do not think that is true, but i am not sure... (please see first post for details) |
I'm afraid you are wrong and it does, but not too much imho.
The scheduler aplies the BR and the BBR to TS (and racial factors) to get 2 numbers that are related to the probability of winning (in touchdowns and in a casualty contest (don't know if there are racial factors for casualties in the formula))
There Christer explains it.
(I may have made my own mistakes, but I think I understood most) |
|
|
CircularLogic
Joined: Aug 22, 2003
|
  Posted:
Nov 18, 2008 - 15:50 |
|
Optihut wrote: | Reisender wrote: | since everybody ignored it so far, i would like to come back to my first post myself (last one on page one).
the question i ask there is: are you sure that the current scheduling as christer decribes it really tends to give TS-advantadge to weaker coaches? actually i do not think that is true, but i am not sure... (please see first post for details) |
I'm not so sure either, as I faced a stronger team (131/138 vs 162/158) with a higher ranked coach (CR 161.94 vs 177.3) in my last game. It was still fun, though. |
If you read the first part of my wall of text further up this page, you will find the answer. The scheduler - at least the way I understood him - scores BR180 coaches against BR180 coaches as high as BR 180 vs BR130 with a 31.3% TS advantage - in both cases the score is above 998. What is given a penalty is if BR-BBR is different. So a BR180 BBR170 coach can be matched perfectly with BR170 BBR160 or BR 140 BBR130 (each can be given matches with scores over 998). But a the best matchup between BR170 BBR 160 and BR150 BBR 150 scores at 959.
Hope that answers your question.
Edit:
I have written an Excel sheet that gives you a distribution of scores for different combinations of BR/BBR/TS. You insert 2 sets of BR/BBR and the TS of the stronger coach at 150. The sheet then plots the score of the matchup against the TS of the weaker coach. I think that gives a good visualization how the scheduler works. It incorporates everything but the racial chart and - of course - the randomizer. If you are interested, drop me a PM. |
Last edited by CircularLogic on %b %18, %2008 - %15:%Nov; edited 2 times in total |
|
BillBrasky
Joined: Feb 15, 2005
|
  Posted:
Nov 18, 2008 - 15:53 |
|
BlackBox is the best thing in A LONG TIME here on Fumbbl. Christer you Rock. Keep up the good work. |
|
|
westerner
Joined: Jul 02, 2008
|
  Posted:
Nov 18, 2008 - 15:59 |
|
CircularLogic wrote: | So Peter brings a TS150 team, then the scheduler thinks, that Jim should have a TS157 (909) team for that, Pauly should have a TS175 (857) team and Olli should have a TS 183 (897) team. The numbers in parenthesis are the highest scores possible and that is before last match and racial penalty as well as without the randomization. Note that without the latest tweak in the scheduler, these numbers would be TS 161/215/239 for Jim/Pauly/Olli. |
You seem to understand how the scheduler works. I'm not getting it, though. Would you mind explaining how you calculate these numbers based on the info Christer posted?
Thanks |
_________________ \x/es |
|
shadow46x2
Joined: Nov 22, 2003
|
  Posted:
Nov 18, 2008 - 16:03 |
|
you know circular....
for as much as you talk about B, positive and negative, and flaunt about how you know so much about the league...
you sure do avoid it pretty well...
--j |
_________________
origami wrote: | There is no god but Nuffle, and Shadow is his prophet. |
|
|
CircularLogic
Joined: Aug 22, 2003
|
  Posted:
Nov 18, 2008 - 16:06 |
|
westerner wrote: | CircularLogic wrote: | So Peter brings a TS150 team, then the scheduler thinks, that Jim should have a TS157 (909) team for that, Pauly should have a TS175 (857) team and Olli should have a TS 183 (897) team. The numbers in parenthesis are the highest scores possible and that is before last match and racial penalty as well as without the randomization. Note that without the latest tweak in the scheduler, these numbers would be TS 161/215/239 for Jim/Pauly/Olli. |
You seem to understand how the scheduler works. I'm not getting it, though. Would you mind explaining how you calculate these numbers based on the info Christer posted?
Thanks |
Simple - I have a spreadsheet that calculates the score for every TS from 100-220 for a given BR/BBR combination of coaches under the assumption that coach 1 brings a TS150 team. I plot the score versus the TS and get a maximum of the curve. The coordinates of that maximum are the best TS and the highest possible score between those 2 opponents. |
|
|
Mr_Foulscumm
Joined: Mar 05, 2005
|
  Posted:
Nov 18, 2008 - 16:07 |
|
BillBrasky wrote: | BlackBox is the best thing in A LONG TIME here on Fumbbl. Christer you Rock. Keep up the good work. |
Hehe, Bill... if it wasn't for the fact that [B]ox arranged those games I would call you a dirty picker! |
_________________ Everybody's favorite coach on FUMBBL |
|
Unstoffe
Joined: Aug 22, 2004
|
  Posted:
Nov 18, 2008 - 16:09 |
|
Seems like the main bone of contention here is how much, if any, a TS advantage should balance out differences in coaching skill.
But never mind that... I've spotted something else that worries me a bit. Bear with me here, there will be diagrams
Well - one diagram anyway.
This illustrates my understanding of how P and bashP for a given match are used to calculate the 'distance' used in the suitability calculation. Quite simply, for any P and bashP we can plot a point on this graph, and the distance is then the distance from that point to the origin.
I've plotted five points A-E as examples. A is the 'ideal' match with P and BashP at zero. But consider the other four :
B is a match where one team is likely to both outscore and outbash the other - not a fair match then.
C is the reverse of B, the other team will both outscore and outbash.
D and E though - in these games, one team will probably outscore, the other probably outbash. How unfair are these matches? I should say that your typical 'fair' game between elves and dwarves, say, would be in this sort of position (though I've probably chosen exaggerated values of P and bashP for the examples).
My point is that the 'distance' for B, C, D & E are the same, all four are therefore as likely to be assigned by the box. But speaking personally, I'd greatly prefer to be given D or E than C. I could probably live with B
I think there's even an argument here for saying, any matchups on a line drawn through A, D & E are essentially fair - although maybe matchups nearer to 'A' should be made more frequently.
Well - did that make any sense? Or if it's wrong, somebody let me know why. |
_________________ British or thereabouts? Check out the White Isle League |
|
CircularLogic
Joined: Aug 22, 2003
|
  Posted:
Nov 18, 2008 - 16:16 |
|
Well.. in this case the scheduler would match either C and B or D and E first, because the distance between those 2 is max. Which one is decided by the randomization process of distance. Lets say C and B get selected, then either of them (lets say the randomization process picks B) is matched with A (the best match for B). C in this cas is matched with E or D (best matches for C after A). Again, randomization kicks in pairing C and E, leaving D unpaired. |
|
|
Unstoffe
Joined: Aug 22, 2004
|
  Posted:
Nov 18, 2008 - 16:24 |
|
To clarify... A, B, C, D & E don't represent teams needed to be matched to each other. They represent possible matchups. The P and bashP on the graph are the values from Christer's calculation : distance = sqrt(P*P + bashP*bashP)
Unless I've completely misunderstood that... |
_________________ British or thereabouts? Check out the White Isle League |
|
CircularLogic
Joined: Aug 22, 2003
|
  Posted:
Nov 18, 2008 - 16:30 |
|
Hmm... true.. You diagramm represents the 4 possible matchups of A. |
|
|
|
| |