|
|
|
Value the players based on when they are playing for instance some teams dont play until weeks 3, 4, 5 and 6... their players waste roster spots and acquire no points until they finally play. Maybe a breakdown of the most efficient ELO vs week by week trading to have the most possible players playing
|
Wow i didn't notice you wern't natively speaking english, definitely no need for an apology
man its times like these i wish id put much more thought into it
|
Fellow physicist here.
Your assumptions are a bit off, IMO.
1.) the only reasonable value for unranked players in terms of Elo is 2000. 1700 is really arbitrary and it's pretty much impossible to be that low anyway (do you know how Elo works?)
If you adjust the unranked players to 2000 (because their performance is relatively unknown in Korea) you'll fare better.
Also, it might be useful to use a rule of thumb to correct Thorzain's and Naniwa's Elo using their foreign Elo, since there are stats for that (I believe)
2.) Your main team needs to weigh Elo vs points (as does your anti). So it probably would be good to have an Elo/cost rating, and then match it up with total cost and Elo. This allows you to decide whether you want a bunch of mediocre-strong players or one/two very very good players and a bunch of lower ranked players.
3.) you need to add in the team's value somehow as well. For example, a low ranked player will most likely be pulling in points solely on the performance of his/her team. Since I think SlayerS performs better than oGs in TL, I'd rather have Cella over Cezanne, for example.
4.) your anti needs players with high cost but low relative Elo. You want the exact opposite of your anti. thus, MMA might kill your team because he's likely to rack up kills and streak breaks. IMO it may be better to get 2 mediocres and a weak or 3 mediocres or something like that.
|
At the end of the day this kind of analysis proves somewhat useless (edit: maybe not useless, but extremely difficult to do in any way that takes all the important stuff into consideration) for several reasons:
1. Having a schedule, along with the ability to trade, has a great deal of implications for player value. Having bomber for 5 games may not be nearly as good as, say, some combination of MVP and MKP over 7 or 8 games - even though their ELO is lower. I spent entirely too long procrastinating work, instead figuring out which team combinations will allow for the most player games, with the fewest number of trades, while maximizing the easiest/hardest matchups. (Hard matchups good for strong players since they are more likely to have to play, easy matchups good for weak players) At the end of it all I had to make several potentially fatal assumptions just to ensure my entire week wasn't consumed.
2. It does not consider how often a player is likely to be played. This would, presumably, be a function of the difficulty of the match-up, the quality of the rest of his own team roster, etc. Some teams have more difficult schedules than others, so you would expect the top players on those teams to play more. A team like IM, for instance, plays a LOT of easy matchups and so it would be unlikely for a player like Nestea or MVP to even get to play in most games since they will be 4-1 and 4-2 a lot.
3. The corrections suggested by wherebugsgo are important too and after reading his I had to delete a few of mine, haha.
|
Relevant to my interests and I am working on a similar guide for Fantasy GSTL. Here's a note regarding the anti-team.
Because of the scheduling and trade-rules, there's an unique, optimal solution for the antiteam.
Team F.United does not play week 1-5, which means their players (Moon-5, Lyn-5, Thorzain-4, Naniwa-4, Soccer-1) will not incur a single point for those weeks. Therefore, the optimal solution for a -13 point antiteam are Lyn, Thorzain, Naniwa.
Since the trade-rule stipulate that you can only trade one antiteam a week, simply swap the players out for an underperforming player or FXO player (whom won't player week 6-10) as you approach week 6.
|
On June 22 2011 06:01 Primadog wrote:Relevant to my interests and I am working on a similar guide for Fantasy GSTL. Here's a note regarding the anti-team. Because of the scheduling and trade-rules, there's an unique, optimal solution for the antiteam. Team F.United does not play week 1-5, which means their players (Moon-5, Lyn-5, Thorzain-4, Naniwa-4, Soccer-1) will not incur a single point for those weeks. Therefore, the optimal solution for a -13 point antiteam are Lyn, Thorzain, Naniwa. Since the trade-rule stipulate that you can only trade one antiteam a week, simply swap the players out for an underperforming player or FXO player (whom won't player week 6-10) as you approach week 6.
That may not actually be optimal, since you are guaranteed to lose 3 points from the trades, and you still end up with at least 3 player matches being played. Also, there is no guarantee that there will be 3 eligible people to trade for if everyone on the other team is ranked too low.
edit. I better not say too much though because I may give away my advantage 
|
On June 22 2011 06:03 Gnial wrote:Show nested quote +On June 22 2011 06:01 Primadog wrote:Relevant to my interests and I am working on a similar guide for Fantasy GSTL. Here's a note regarding the anti-team. Because of the scheduling and trade-rules, there's an unique, optimal solution for the antiteam. Team F.United does not play week 1-5, which means their players (Moon-5, Lyn-5, Thorzain-4, Naniwa-4, Soccer-1) will not incur a single point for those weeks. Therefore, the optimal solution for a -13 point antiteam are Lyn, Thorzain, Naniwa. Since the trade-rule stipulate that you can only trade one antiteam a week, simply swap the players out for an underperforming player or FXO player (whom won't player week 6-10) as you approach week 6. That may not actually be optimal, since you are guaranteed to lose 3 points from the trades, and you still end up with at least 3 player matches being played. Also, there is no guarantee that there will be 3 eligible people to trade for if everyone on the other team is ranked too low. edit. I better not say too much though because I may give away my advantage
You lose resource points (which I prefer as ![[image loading]](http://wiki.teamliquid.net/starcraft/images2/a/a4/Minerals.gif) ), not scoring points. It's the optimal solution in that you cannot find a better solution without relying on contingencies.
|
On June 22 2011 06:16 Primadog wrote:Show nested quote +On June 22 2011 06:03 Gnial wrote:On June 22 2011 06:01 Primadog wrote:Relevant to my interests and I am working on a similar guide for Fantasy GSTL. Here's a note regarding the anti-team. Because of the scheduling and trade-rules, there's an unique, optimal solution for the antiteam. Team F.United does not play week 1-5, which means their players (Moon-5, Lyn-5, Thorzain-4, Naniwa-4, Soccer-1) will not incur a single point for those weeks. Therefore, the optimal solution for a -13 point antiteam are Lyn, Thorzain, Naniwa. Since the trade-rule stipulate that you can only trade one antiteam a week, simply swap the players out for an underperforming player or FXO player (whom won't player week 6-10) as you approach week 6. That may not actually be optimal, since you are guaranteed to lose 3 points from the trades, and you still end up with at least 3 player matches being played. Also, there is no guarantee that there will be 3 eligible people to trade for if everyone on the other team is ranked too low. edit. I better not say too much though because I may give away my advantage You lose resource points (which I prefer as ), not scoring points. It's the optimal solution in that you cannot find a better solution without relying on contingencies.
I'll definitely have to look more into this distinction between resource points and scoring points...
edit. Upon looking into it, I believe you made an error. There is a score tax of 1 against your score for each trade you make. Read the bottom line of the "Scoring" explanation.
You do have to rely on contingencies in that you still rely on 3 player games, and you have to rely on FXO players appreciating in value, and F.U players depreciating in value, in such a way that allows you to actually make the trades, since you can't trade a player for a player of lesser value for the anti-team.
You missed the first step of optimization - looking into the mechanics
|
On June 22 2011 02:11 wherebugsgo wrote:
Fellow physicist here.
Your assumptions are a bit off, IMO.
1.) the only reasonable value for unranked players in terms of Elo is 2000. 1700 is really arbitrary and it's pretty much impossible to be that low anyway (do you know how Elo works?)
If you adjust the unranked players to 2000 (because their performance is relatively unknown in Korea) you'll fare better.
Also, it might be useful to use a rule of thumb to correct Thorzain's and Naniwa's Elo using their foreign Elo, since there are stats for that (I believe)
2.) Your main team needs to weigh Elo vs points (as does your anti). So it probably would be good to have an Elo/cost rating, and then match it up with total cost and Elo. This allows you to decide whether you want a bunch of mediocre-strong players or one/two very very good players and a bunch of lower ranked players.
3.) you need to add in the team's value somehow as well. For example, a low ranked player will most likely be pulling in points solely on the performance of his/her team. Since I think SlayerS performs better than oGs in TL, I'd rather have Cella over Cezanne, for example.
4.) your anti needs players with high cost but low relative Elo. You want the exact opposite of your anti. thus, MMA might kill your team because he's likely to rack up kills and streak breaks. IMO it may be better to get 2 mediocres and a weak or 3 mediocres or something like that.
Good Points.
1) If you put them to 2000 you the bruteforce will most likely take a cheap 1 point guy. 1700 is much more arbitary then 2000 - yeah. But im not sure if 2000 is good?
2) If you put in just the weighten ELO it will totally fail. A 2000 ELO, 1Point guy would have a weightend ELO of 2000. Bomber and so would have much less. In the end you may maximize both or so?
3) Yeah i pointed out the team value in my first point. I'll think about it.
4) The Antiteam is not good. I recognized it at the start. You are prefectly right.
In the end its all about a (or 2?) value function. There you can put in like a few aspects:
ELO
Weightend ELO
Teams ELO
Probability of playing
Its sad that there is no Sc2 Powerrank - which would be quite good.
It could be something like
V = E * T * C * A
With
V = Value
E = Players ELO
T = Team ELO
C = Teamsynergy, 1,X if it is on your chosen team 1 if it isnt
A = Chance of appereance (however you would calculate that one..)
|
This is interesting but the team factor matters a lot when determining a player's value, because of the point that a player gets for showing up.
Each team will send out 4 players, so usually the top 3 players will always be sent out, the 4th player will be sent out most of the time, the 5th player some, and the rest get scraps.
So there's a big dropoff from 3rd to 4th, from 4th to 5th, which varies widely depending on what team you are on. But this is still something that you can estimate using the ELO ratings on each team.
|
On June 22 2011 07:53 yoshi_yoshi wrote:
This is interesting but the team factor matters a lot when determining a player's value, because of the point that a player gets for showing up.
Each team will send out 4 players, so usually the top 3 players will always be sent out, the 4th player will be sent out most of the time, the 5th player some, and the rest get scraps.
So there's a big dropoff from 3rd to 4th, from 4th to 5th, which varies widely depending on what team you are on. But this is still something that you can estimate using the ELO ratings on each team.
To further elaborate on that, there's 170 players in the Fantasy draft list, some teams more players than others:
![[image loading]](http://wiki.teamliquid.net/starcraft/images2/5/52/SlayerSlogo_std.png) SlayerS - 22 players, 51
+ Show Spoiler +![[image loading]](http://wiki.teamliquid.net/starcraft/images2/8/83/IMlogo_std.png) Incredible Miracle - 13 players, 41
+ Show Spoiler +![[image loading]](http://wiki.teamliquid.net/starcraft/images2/8/8e/Startalelogo_std.png) StarTale - 13 players, 37
+ Show Spoiler +![[image loading]](http://wiki.teamliquid.net/starcraft/images2/e/ee/TSLlogo_std.png) Team SCV Life - 9 players, 33
+ Show Spoiler +![[image loading]](http://wiki.teamliquid.net/starcraft/images2/4/41/OGslogo_std.png) Old Generations - 19 players, 45
+ Show Spoiler +![[image loading]](http://wiki.teamliquid.net/starcraft/images2/8/87/FOulogo_std.png) For Our Utopia - 16 players, 34
+ Show Spoiler +![[image loading]](http://wiki.teamliquid.net/starcraft/images2/0/05/ZeNEXlogo_std.png) ZeNEX - 16 players, 26
+ Show Spoiler +- MVP - 19 players, 39
+ Show Spoiler + - Prime - 17 players, 42
+ Show Spoiler + ![[image loading]](http://wiki.teamliquid.net/starcraft/images2/9/90/HoSeologo_std.png) New Star HoSeo - 11 players, 20
+ Show Spoiler +![[image loading]](http://wiki.teamliquid.net/starcraft/images2/a/a2/FXOlogo_std.png) FXOpen - 10 players, 16
+ Show Spoiler +![[image loading]](http://wiki.teamliquid.net/starcraft/images2/a/a2/FUnitedlogo_std.png) F.United - 5 players, 19
+ Show Spoiler +
|
|
|
|
|
|
EPT
Bomber 9 2264
Alicia 5 2183
viOlet 2 2075
GuineaPig
![[image loading]](http://wiki.teamliquid.net/starcraft/images2/e/ec/MVPlogo_std.png)
![[image loading]](http://wiki.teamliquid.net/starcraft/images2/8/84/Primelogo_std.png)
