treborius
Joined: Apr 05, 2008
|
  Posted:
Jun 16, 2009 - 00:12 |
|
Over the last couple of days i've been quite happy to see the number of Box-games increase...
...i think that it's probably partly due to the fact that there has been talk about Christer wanting to introduce tourneys to the Box
...however, i'm also witnessing some really strange match-ups, lately - for example consider these two (at <700):
[Monday, 23:30] Processing round.
Found 10 coaches.
- Sandsack vs spathii (Halfling 135/110 vs 149/123 Ogre) @823
- DeMoorck vs zhraia (Undead 148/147 vs 141/160 Necromantic) @806
- def909 vs Crimsonwolf (Orc 146/146 vs 134/126 Chaos) @645
- Chingis vs cealisson (Amazon 153/153 vs 136/133 Chaos) @631
- Snappy_Dresser vs anisdrin (Chaos 111/99 vs 116/94 Lizardmen) @867
Scheduled 5 matches
...i haven't seen such bad match-ups, before (TS-diff of 15 even *after* HC-modifier) - in fact, i always thought it was max_TS_difference = 14 ??? (cutoff=15)
...i think with more and more coaches joining the Box we should think about improving the match-ups, because this looks really awful to me
i've proposed it before, but here i go once again: i'm for a relative component in max_TS_difference, i.e. if you fix the max_TS_difference for a team at say 8% of its TS, i think that would be an improvement, already!
it would also be possible to have each coach choose his own max-difference (within certain limits, say 5%..15% or 5TS..20TS), such that (s)he could find his/her own preferred balance between getting games and having close games (in TS).
match-ups would only be considered by the algorithm if they satisfy both coach's settings. |
|
|
CircularLogic
Joined: Aug 22, 2003
|
  Posted:
Jun 16, 2009 - 00:34 |
|
The really strange thing here is, that Orc vs Amazons and Chaos on Chaos would be both possible and probably better matches. |
|
|
Chingis
Joined: Jul 09, 2007
|
  Posted:
Jun 16, 2009 - 00:47 |
|
Just for info, the Amazons weren't my preferred team (the Dwarfs were at 106/105).
It should be noted that both def909 and ceallisson had one box team, so that may be causing some discrepancy where it's harder to match a single team, although I see the point about the Chaos vs. Chaos and Orc vs. Amazon. Is it an over-eager "non-mirror" function?
(If I can read the racial match-up table correctly Orc-Amazon should be almost 50-50 at those TSs too).
Edit: having a look, it looks like five of the ten coaches in the round entered with one team. Of course people can enter as many teams as they want, but just in looking at the matchup values that will go some way to explaining the "strained" nature of picking this particular Blackbox round. Still doesn't explain why Chaos-Chaos and Orc-Amazon weren't better matches though... |
|
|
treborius
Joined: Apr 05, 2008
|
  Posted:
Jun 16, 2009 - 00:55 |
|
Chingis wrote: | Is it an over-eager "non-mirror" function? |
'only explanation i see, so far - trading an additional TS-difference of 8 (15 instead of 7) for producing non-mirror-matches does seem "over-eager", indeed
EDIT: as far as i see, neither of those 4 coaches (2xchaos, orc, amazon) would've had a repeated-match vs. same race either
EDIT2: maybe this should get moved to "Bug-Reports" or something similar? |
|
|
Kryten
Joined: Sep 02, 2003
|
The scheduler doesn't check all the possible matchups. If I recall correctly, it makes three passes through and then chooses the best set (best being the highest sum of the pairing suitability numbers).
For mirror matches, I believe their is a 3% reduction in suitability. |
|
|
lawman
Joined: Oct 07, 2008
|
  Posted:
Jun 16, 2009 - 03:32 |
|
I'd prefer to see less matchups like the one i had to face yesterday :
Norse TR/Str 175/180 Norse
TR/Str 249/205 Chaos
Can you guess what made the chaos team so bloated?
Thinking especially with high TS teams, that difference between TR makes a huge difference. If you get three sad rolls on the pre-matchup table (like I did) you can pretty much being looking at a boring game because mostly you will be outgunned. I think the TS factors for some chaos players (especially with claw/mb/RSC combos maybe slightly underweight compared to baller players. |
|
|
treborius
Joined: Apr 05, 2008
|
  Posted:
Jun 16, 2009 - 08:57 |
|
I'd prefer to see less matchups like the one i had to face yesterday :
Norse TR/Str 175/180 Norse
TR/Str 249/205 Chaos
i looked it up...
lawman wrote: | Can you guess what made the chaos team so bloated? |
seems like it was loads of spp?
lawman wrote: | Thinking especially with high TS teams, that difference between TR makes a huge difference. |
i agree, i'd rather see HCs removed from the Box than have them based on TR, but i think at that TS-range it rather favors the less-TR-team (on average)
lawman wrote: | If you get three sad rolls on the pre-matchup table (like I did) you can pretty much being looking at a boring game because mostly you will be outgunned. |
yes, but if you get 3 average..good ones, they're easily worth their TS (and more)
lawman wrote: | I think the TS factors for some chaos players (especially with claw/mb/RSC combos maybe slightly underweight compared to baller players. |
possibly true (i honestly don't know), but certainly not in the game you played (0 claw-rsc chaos-players, but 3 +AG ones ) |
|
|
sk8bcn
Joined: Apr 13, 2004
|
  Posted:
Jun 16, 2009 - 13:40 |
|
well, some of this games were a bit unfair, which, as always, doesn't make the box unfair.
I say 15TS is ok. If you start to find 15 beeing too much, and reduce it to 12, then 12 will be too much...
How many games are that loopsided so far?
I certainly have less than 1/10. And I rather play that 10th game than beeing unscheduled. |
_________________ Join NL Raises from the Ashes |
|
odi
Joined: Aug 02, 2003
|
  Posted:
Jun 16, 2009 - 13:53 |
|
The scheduler is doing a good job in my opinion... Sure sometimes you get matches that feel a bit off, but remember, sometimes you get the bear, sometimes the bear gets you. And even being slaughtered can be fun |
|
|
CircularLogic
Joined: Aug 22, 2003
|
  Posted:
Jun 16, 2009 - 14:02 |
|
I`d say that one in every 8 is abit questionable. And I rather be unscheduled instead of playing that 8th game. But I accept it as the box in total is a good one.
I still would be extremely interested, how those miss-schedules happen. Maybe if it would be possible to implement a bowlbot command that returns the three generated lists + scores?
I also think, that the racial mirror factor could do with some tweaking. Either only apply it, when the TS difference is below 4 or just tweak random factor. Instead of multiplying the final score with a random factor, modify the TS of a team with a two decimal value ranging from -1 to +1. This way you would also avoid a mirrormatch having a 'perfect match' more likely. |
|
|
def909
Joined: Oct 25, 2005
|
  Posted:
Jun 16, 2009 - 14:10 |
|
- def909 vs Crimsonwolf (Orc 146/146 vs 134/126 Chaos) @645
I was playing the Orcs and felt my team was _a lot_ stronger (4 guards, dp and BG vs. no guard, almost no block, no dp and no BG - just two meager claws) - my oppo never had a chance. I´ve had another game where my oppo was up 15TS too though. However, generally the matchups seem more or less fair and in the end it all equals out if you play enough games. Matching will never be perfect and I agree with sk8bcn that I rather play that "unfair" 10th game if that means more games are scheduled. I see it as either a real challenge or a good opportunity for spp farming every now and then (depending which side of the matchup you are). |
_________________ Don´t ask Nuffle for better dice, ask him to make you a better coach. |
|
ibambe
Joined: Jun 27, 2005
|
  Posted:
Jun 16, 2009 - 14:55 |
|
|
arw
Joined: Jan 07, 2007
|
  Posted:
Jun 16, 2009 - 15:30 |
|
treborius wrote: | it would also be possible to have each coach choose his own max-difference (within certain limits, say 5%..15% or 5TS..20TS), such that (s)he could find his/her own preferred balance between getting games and having close games (in TS).
match-ups would only be considered by the algorithm if they satisfy both coach's settings. |
I think that is not too bad.
In fact I already mentioned it in the Bettering Blackbox (Summary) thread.
To make it more attractive however to accept a broad TS range I emphasized to show this in the team bio as "difficulty-setting" you play. Playing a tough race but accepting more challenging games would be more accepted that way I guess.
If you like it just second it in the thread please.
It is referred to as "Difficulty options for playing up/down TS for awards"(unseconded). |
|
|
CircularLogic
Joined: Aug 22, 2003
|
  Posted:
Jun 16, 2009 - 17:00 |
|
Wait.. wasn`t blackbox created, so that all coaches could play matches on an even footing from the matchmaking angle? So lets bring in this option, so that 'picking low-TS-range only' becomes the new cherrypicking excuse.
Oh.. and accepting a larger TS range doesn`t make it more difficult. It just moves your win% towards 50% as there are more matches with a fixed outcome. So no.. playing bashers is still as lame as before.. no wait.. it`s even lamer, because you have to do even less thinking. Either you have advantage, then you win and main, or you have the disadvantage, then you just main. |
|
|
treborius
Joined: Apr 05, 2008
|
  Posted:
Jun 16, 2009 - 20:31 |
|
sk8bcn wrote: | well, some of this games were a bit unfair, which, as always, doesn't make the box unfair.
I say 15TS is ok. If you start to find 15 beeing too much, and reduce it to 12, then 12 will be too much... |
i agree, that the box is ok in general, but i think the 2 match-ups above are not acceptable (i'd rather not play than having a match-up like that).
the real question still remains: why the 2 chaos-teams weren't scheduled vs. one another as they had only a TS-difference of 7 (instead of 15)?
personally, i haven't had such bad match-ups myself, so far (as far as i can remember) and i'm just wondering whether or not something has changed as i can't remember ever seeing a matching-score below 650
also, i think that the acceptable TS-difference should (at least in part) be dependent on the TS-band (i.e. TS-difference=15 or even 20 is really not that much around 250, but it's a whole lot around 100) |
|
|
|
| |