pt0510
Joined: Dec 15, 2007
|
 Posted:
Nov 23, 2008 - 12:55 |
 
|
\m/
nice one Christer |
_________________
SELL YOUR SOUL TO NUFFLE AND BUY A CUPCAKE! |
|
CircularLogic
Joined: Aug 22, 2003
|
| Posted:
Nov 23, 2008 - 13:13 |
|
OK.. the are a few implications.. and I`m not sure if I`m too happy with some of them...
1) You can basically combine BR and BBR into one figure, as they both now directly offset each other (Though it`s good that they are separated.. I can hardly imagine how [R] would look, if casualties would raise your CR). Bashy winners will still get paired against non-bashy losers - just wth a higher TS advantage.
2) Dropping the minimum score cutoff means that you ensure coaches on either end of the BR/BBR spectrum get matches, though it will bring alot of 15TS matchups if there are not many coaches in the box - or more specifically an even BR spectrum. With the current formula it means that 17 points of BR give you the standard 15TS advantage. 10 points of BBR count as 3 points of BR. Matches between coaches with a higher difference become unlikely (which is a good thing).
3) Having a fixed TS number is problematic. For a TS100 team, 15 TS is huge, while for a TS250 team, those 15TS might be a niggle rolle failed or a bad handicap.. a minor inconvenience. Why not make it X% of the lower teams TS? I`d suggest 8%.
4) One of the main gripes of the formula: It evaluates the BR+TS match just as good as a perfect BR match.
Lets have an example for a BR165 coach with a TS150 team having 2 possible matches (BBR is assumed equal or calculated into BR and TS is weighted by racial factors):
1)BR165 with TS 150 vs BR 165 with TS 151 (993)
2)BR165 with TS 150 vs BR148 with TS 165 (997)
Even though most (I dare to say all) of us would agree that match 1 is nearly as good as it can get, the scheduler as shown at by the numbers in parenthesis will match 2), because in the formula 2) scores higher.
5) The problem of BR-metagaming is still present. If a person doesn`t care about BR (teambuilder for example), then it can use underpowered team or playing bad on purpose to artificially trash their BR. With the latest changes, it`s just advisable to avoid causing casualties (though that could be optimised once the exact way BBR is calculated get public). If you play that game enough, you will get 15 TS at every matchup - which is the time when you can create your 'serious' team that you want to build. Sure.. you still better have a durable team, but you could get your orcs/khemri/killerchaos just as high as by cherry picking in ranked. So should there every be tourneys for [B], as some people would like to, then this IS a serious problem.
Also the 'kill for the loss' group can just make a 'trash team' to artificially lower their BBR and not only get their 15TS advantage, but also reduce the probability to face serious coaches.
6) Assuming a spread equal to the CR distribution at the moment, this will still drive away the high BR coaches, because playing up 9 out of 10 matches at a considerable TS advantage will kill the fun to play. It`s more like work then. Plus if you are BR 180, then you give up 15TS to anything below 163. While you can beat (probably quite easily) BR 145 coaches at that TS dif, someone with solid coaching will qinn because of that advantage, even though you might be better. Skill doesn`t rise in a linear fashion. The skill difference between 133 and 150 will be huge, the skill difference between 173 and 190 tiny.
Dear Christer, I have now tried to give the best analysis my still sleepy brain could come up with in 45 mins. I still prefer the simple solution of:
distance=dTS(weighted)*dTS(weighted)+dBR (optional: +dBBR)
But I`m afraid that this might run against your mission statement of 50% win chance being the ideal. But looking at the recent changes, you are leaving this ideal more and more - so maybe it`s best just to look at the alternatives. If you don`t want to drop the calculation based on the CR formula, then I want to suggest the following tweaks:
1) Make the TS cap nomalized to the lower team. as I mentioned before, 8-10% feels fine.
2) Add dBR/200 and dBBR/500 to the distance. distance= abs(P+Pbash)+ abs(dBR/200)+abs(dBBR/500) |
Last edited by CircularLogic on %b %23, %2008 - %14:%Nov; edited 1 time in total |
|
Mr_Foulscumm
Joined: Mar 05, 2005
|
| Posted:
Nov 23, 2008 - 14:41 |
|
\o/
Sounds hot! Good job Christer! |
_________________
Everybody's favorite coach on FUMBBL |
|
Reisender
Joined: Sep 29, 2007
|
| Posted:
Nov 23, 2008 - 14:45 |
|
| vanGorn wrote: | | Wouldn't that prevent some teams from getting a fixture at all? |
minimum suitability prevented that as well
- happened to me twice in about 25 tries - which is no problem imo
probability raises ofcourse if you are on the high or low end of TS in da box |
|
|
Purplegoo
Joined: Mar 23, 2006
|
| Posted:
Nov 23, 2008 - 16:35 |
|
Big fan of the new text in the channel.
A fan of any tweaks to make TS gaps tighter. |
|
|
KenThis
Joined: Jun 28, 2007
|
| Posted:
Nov 23, 2008 - 16:40 |
|
I like the changes mentioned but am no good at maths.
do the changes mean I am now more likely to face coaches with BR closer to my own? |
|
|
JanMattys
Joined: Feb 29, 2004
|
| Posted:
Nov 23, 2008 - 16:50 |
|
| KenThis wrote: | I like the changes mentioned but am no good at maths.
do the changes mean I am now more likely to face coaches with BR closer to my own? |
Yeah.
Juicy thing for your dwarfs, eh?
 |
_________________
 |
|
CircularLogic
Joined: Aug 22, 2003
|
| Posted:
Nov 23, 2008 - 16:57 |
|
| KenThis wrote: | I like the changes mentioned but am no good at maths.
do the changes mean I am now more likely to face coaches with BR closer to my own? |
The probability to play coaches with a BR that is more than 17 points different than your sinks. For the rest it`s exactly the same (as in you have just as much chance to play a 17BR higher or lower coach than a coach of your BR). |
|
|
KenThis
Joined: Jun 28, 2007
|
| Posted:
Nov 23, 2008 - 17:10 |
|
thanks for the info.
any chance of reducing the possible tr/ts advantage and therefore increasing the likelihood of facing coaches with BR within +- 10 points?
Surely this would be what everyone wants all matches to have more even tr/ts and reduced probability of very good coaches playing very poor coaches.
or am I missing something?
Thanks Christer.
Looking forward to trying it out. |
|
|
pythrr
Joined: Mar 07, 2006
|
| Posted:
Nov 23, 2008 - 17:14 |
|
| CircularLogic wrote: | | KenThis wrote: | I like the changes mentioned but am no good at maths.
do the changes mean I am now more likely to face coaches with BR closer to my own? |
The probability to play coaches with a BR that is more than 17 points different than your sinks. |
Wha? |
|
|
nin
Joined: May 27, 2005
|
| Posted:
Nov 23, 2008 - 18:50 |
|
| CircularLogic wrote: | OK.. the are a few implications.. and I`m not sure if I`m too happy with some of them...
...
6) Assuming a spread equal to the CR distribution at the moment, this will still drive away the high BR coaches, because playing up 9 out of 10 matches at a considerable TS advantage will kill the fun to play. It`s more like work then. Plus if you are BR 180, then you give up 15TS to anything below 163. While you can beat (probably quite easily) BR 145 coaches at that TS dif, someone with solid coaching will qinn because of that advantage, even though you might be better. Skill doesn`t rise in a linear fashion. The skill difference between 133 and 150 will be huge, the skill difference between 173 and 190 tiny.
...
|
Few implications long text... and reading it, it'll be difficult for everibody to be very happy with the scheduler.
Still I think I like the new formula more han the old one.
Something extra concerning mostly this point 6:
1) If the ramdon modifier is keept in the formula, then those numbers are not extrictly accuate.
2) The hard point would be starting new teams and facing allways teams with over +10 TS.
3) A bit of alpha testing with linear functions will be good for Christer's sanity. I was starting to worry about the effects of another exponential formula on any human mind... may be in some days he can come up with a polinomial formula for the TS and BR gaps. |
|
|
CircularLogic
Joined: Aug 22, 2003
|
| Posted:
Nov 23, 2008 - 19:20 |
|
I prefer the newest formula over the previous incarnation. Yet, I think it`s not what it could be. I think it`s known, that I am rooted in the TS-only camp, so excuse me if my text makes people less likely to be very happy with the current system - I thought that was the point of making an arguement.
Yes.. there is a random modifier, but the influence isn`t huge so for making estimates and giving examples it can be neglected. Just as I neglect the racial modifiers, because else sensible examples become too complicated to show anything specific. |
|
|
SillySod
Joined: Oct 10, 2006
|
| Posted:
Nov 23, 2008 - 19:25 |
|
| CircularLogic wrote: | | I think it`s known, that I am rooted in the TS-only camp |
Really? Thats news to me |
_________________
Putting the "eh?" back into Sexeh.
"There are those to whom knowledge is a shield. There are those to whom it is a weapon. Neither view is balanced." |
|
westerner
Joined: Jul 02, 2008
|
| Posted:
Nov 23, 2008 - 19:33 |
|
| CircularLogic wrote: | | I prefer the newest formula over the previous incarnation. Yet, I think it`s not what it could be. I think it`s known, that I am rooted in the TS-only camp, so excuse me if my text makes people less likely to be very happy with the current system - I thought that was the point of making an arguement. |
Props for distinguishing between objective analysis and personal opinion. 
| nin wrote: | | 2) The hard point would be starting new teams and facing allways teams with over +10 TS. |
Indeed. It feels like there should a per-team weighting in addition to, or instead of, a per-coach weighting. Particularly for bashiness. |
_________________
\x/es |
|
odi
Joined: Aug 02, 2003
|
| Posted:
Nov 24, 2008 - 08:23 |
|
Yesterday I had my first game that looked on paper like I might be in trouble. But the game ended 1-1, I even had a slight chance of winning. So the scheduler must work  |
|
|
|
|
Conceding v Goblins/...
Search
