|
On March 11 2009 04:59 semioldguy wrote:Show nested quote +On March 11 2009 04:01 malongo wrote:On March 11 2009 00:56 Abydos1 wrote:On March 11 2009 00:43 malongo wrote:On March 11 2009 00:22 Abydos1 wrote:On March 10 2009 23:50 malongo wrote:
I came to think that the anti team cost is too high. Considering that the actual "real" team is about 25 points value (once you take about 5 points to choose your team) 13 points to spare in the anti team is like 50% of your team. Anyways if the rules are the same for everyone its still fair. 30 pts / 7 (6 players + 1 team) = 4.3 pts per player/team 4.3 * 3 (3 players anti team) = 12.9 or when rounded 13 You're having to pay the same amount on average per player on your anti team as a player/team on your main team. -_-' thats the same i was saying, since you have 3 players in your anti team its like 50% cost of your team (6 players without the team). Why is that too high, its the same ratio you're allocated for your main team? (and 3 is 50% of 6) I think its too high because the point system values then the ability to predict players that will perform well and players that will underperform in a ratio 2:1. However, this adds too much randomness about players in the middle segment (because i doubt someone will choose a ss class in his Anti team and succeed). Well just my thought, as i said as long as the rules are the same for everyone its ok. Please more important id like to know how is calculated the trade value of each player. Thanks. The cost is the same for putting on either team. An S-class player may underperform and earn less points than his cost would suggest he earn. If you don't think so then it is your choice not to put him on your negative team. Figuring out whether or not it could be a good/bad idea is part of the fun and no one will really know whether it is truly a good/bad decision until they can look on it in hindsight. Because none of us can predict the future exactly. As far as how trading value works.... Based upon costs players have a certain expected return value for the upcoming season which we split evenly in a points per match sort of formula. The remaining games are always these fixed amounts and gmaes already played get replaced by how many points they have been earning. All these totals are added up (and divided again to keep from having ridiculously large numbers) and that becomes a player's "Trade Value." So if a player is performing above expectations, their value starts going up, if they are slumping their value starts to go down. Players with a small enough trade value, probably less than one (which are all the players who aren't expected to get wins or even get appearances can all be traded among each other since their value is so low that it doesn't break the balance. This has all been tested out and works pretty well.
Ok as i said its just an opinion about the Anti-Team value. Im interested in the trading value, i sort of understand, but id like to see a numerical example if possible, taking any player and making a simulation for week 1 and 2. This requires to know the expected return value and i guess its a lot of work, but since the ability to trade is going to be increased it may be much more important.
|
United States7488 Posts
Abydos1 uses a much more clear (and exact) formula and explanation at the bottom of this page.
+ Show Spoiler +Lets use Leta as an example.
His cost is 9 points. His expected point total for the upcoming season would be approxiamtely 45 points (for any player just multiply their cost by five for expected return value based upon previous performance, no personal biases included in costing)
At 45 points he would be expected to earn either and average of 6.43 points per week or 4.09 points per match.
Let's use points per match for example... his trade value at the beginning of week one would be 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 = 45.00
If Leta earned zero points in week one and played two matches his trade value at the beginning of week two would be 0.00 + 0.00 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 = 36.82
We then divide this figure by a variable based on the number of matches they have remaining to ensure that players with more played games don't simply just get dumped off for alternatives with less games played if both players are at equal strength such a trade would not necessarily be fair.
This isn't how it works verbatim, but this is the general concept behind calculating trade values.
|
Is there a reason why a player who hasn't played since 2000 is pickable? ( SoulKey, costs 1 point)
Is that player supposed to be  Neo.G_Soulkey?
|
Team:
Zero 5
free 8
PianO 4
Guemchi 3
Woongjin Stars 4
Anti-Team:
Bisu
GOGOGOOGOGOGO
|
uh miseiler, the anti-team is suppose to be min.13 points and your team suppose to have 6 players max 30 points
|
are they gonna change the rosters before part 4 begins? wemade lacks protosses, skt only has 2 zergs, etc.
|
United States7488 Posts
I believe the current rosters will not be changing again until May.
|
Ok, so there will be a web app for this? We don't have to PM our teams to semioldguy?
|
On March 11 2009 08:01 Qatol wrote:Is there a reason why a player who hasn't played since 2000 is pickable? ( SoulKey, costs 1 point) Is that player supposed to be Neo.G_Soulkey?
Oops, I was going off the roster post which had SoulKey listed, updated the op, thanks.
|
On March 11 2009 16:57 rushz0rz wrote:
Ok, so there will be a web app for this? We don't have to PM our teams to semioldguy?
Correct, please don't PM semioldguy 
|
United States7488 Posts
On March 11 2009 17:11 Abydos1 wrote:Show nested quote +On March 11 2009 16:57 rushz0rz wrote:
Ok, so there will be a web app for this? We don't have to PM our teams to semioldguy? Correct, please don't PM semioldguy
seconded
|
On March 11 2009 07:07 semioldguy wrote:
Lets use Leta as an example.
His cost is 9 points. His expected point total for the upcoming season would be approxiamtely 45 points (for any player just multiply their cost by five for expected return value based upon previous performance, no personal biases included in costing)
At 45 points he would be expected to earn either and average of 6.43 points per week or 4.09 points per match.
Let's use points per match for example... his trade value at the beginning of week one would be 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 = 45.00
If Leta earned zero points in week one and played two matches his trade value at the beginning of week two would be 0.00 + 0.00 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 + 4.09 = 36.82
We then divide this figure by a variable based on the number of matches they have remaining to ensure that players with more played games don't simply just get dumped off for alternatives with less games played if both players are at equal strength such a trade would not necessarily be fair.
This isn't how it works verbatim, but this is the general concept behind calculating trade values.
Could you explain how this variable is calculated exactly.
I mean for the example you gave Leta - new cost value not taking into account the games that are to be played will be 7.36 meaning total value earned 36.82 divided by the 5 factor.
Is this 5 changing based on the games remaining?
|
United States7488 Posts
Abydos1 uses a much more clear (and exact) formula and explanation at the bottom of this page.
+ Show Spoiler +The five is only used for player costing at the beginning of the season for team creation purposes, not in the calculation of trade values.
I chose five in the very first Fantasy season because it seemed like a nice number while still leaving enough variation in player costs. That and I don't think people trying to calculate out making a 150 point team would be a very nice thing to do :p
(it was also not just an arbitrary decision, everything down to how players are costed and points allowed per team went through various stages of testing and tweaking to get it to be as functional as possible. Maybe one day I'll write a detailed account of my almost year-long process of creating and testing Fantasy Proleague with myself and figuring it out how best to work it... and then how it all got screwed up because of a change in Proleague format. Today is not the day I am going to write about that)
As far as changing trade values based on number of games left, it will change based on the week, since the variable when progamers are split between having 7 or 8 games remaining is different than them being split between having 2 or 3 games remaining. The latter will have larger variance between the two groups, while the variable in the former will not alter the values nearly as much percentage-wise as seven-eighths is a lot closer to 1.00 than two-thirds.
So basically progamers that have 8 games remaining will use a modifier of 1.00 while players with 7 games remaining would use a modifier of 0.875, making their value slightly less since they will have less future opportunities to play.
|
well pls can we have a practical example? like the one with Leta above. Sorry to insist but I want to know this detail because from the new system I think trading, keeping a close eye on the league and next week line-ups each time will make a big impact on the final rankings. Thus I'am thinking that my line-up should be a "powerfull" one taking into account first week and see from there...
I am not saying its good or bad - just want to adapt better to the new system.
To continue with Leta and assuming in week 1 he is sent out on both games and ends up with 0 point - is it ok to say that his new "cost value" will be 6.44 dropping from 9?
|
United States7488 Posts
Abydos1 uses a much more clear (and exact) formula and explanation at the bottom of this page.
+ Show Spoiler +In the example above his new "cost value" would be 7.36, dropped down from nine. This is a more extreme example of early value drop though, as Leta earning absolutely zero would be unexpected and most value fluctuation will happen much more slowly as the round progresses. In the example of Leta, if that were to happen, here is an idea of how other players could rise (or retain) a trade value to be traded for Leta after week one. Lets try to see what scores in week one other players would need to get to obtain that cost value. In all the below instances the example of performances would make it so that Leta could not be traded for the players in the example, but those players could now be traded for Leta. If you want to see what Leta would able to be traded FOR, then just assume that the below example did poorer than their expected outcomes provided. 8 points: ?.?? + 9 * 3.64 = 36.82 ?.?? = 4.09 This player would need to earn at least five points in two games to have a greater value than Leta at the end of week one (this player would be expected to earn about 7) (9) Leta - 0 >>>> (8)  free - 4- (9) Leta - 0 <<<< (8) free - 5+ 7 points: ?.?? + 9 * 3.18 = 36.82 ?.?? = 8.18 This player would need at least nine points in his first two matches to match Leta's value for a trade (this player would be expected to earn about 6, so he would have to perform above expectations). (9) Leta - 0 >>>> (7) Mind - 8- (9) Leta - 0 <<<< (7) Mind - 9+ 8 points (but only 1 game played) (?.?? + 10 * 3.64) * 1.1 = 36.82 ?.?? + 10 * 3.64 = 33.47 ?.?? = 0.92 This player would need to only earn one point in his first game to exceed Leta's value (which is slightly easier to do than other 8-pointers, since this 8-pointer has one more game left to play in the season when compared to all the other players) (9) Leta - 0 >>>> (8) HiyA - 0 (9) Leta - 0 <<<< (8) HiyA - 1+ Of course every single player will have these calculated values every week (thank goodness for automation!) and you will be able to unload a player for ANY player with less value on your team. Or in the case of your negative team you can only trade for a player with more value (the opposite of your regular team)
|
i think this antigroup thing is a bad idea too... gamers are suposed to be good on gaming, not on some gypsy future vision voodoo random thing or whatever, because it is coaches job already!
|
On March 12 2009 01:00 PIJAMA wrote:
i think this antigroup thing is a bad idea too... gamers are suposed to be good on gaming, not on some gypsy future vision voodoo random thing or whatever, because it is coaches job already!
i didnt know that there are gypsies in Brazil too :O
nevertheless quite funny post 
|
Hey, every team has opponents... at least we get to pick our own with the anti-team.
|
The gist of Semioldguy's calculations are essentialy how the calculation is done, however for clarity this is the official formula that is used:
Trade Value = (P / (4.56 / 11)) / M * M/11 + C * (11 - M)/11
P = Player's points.
M = number of matches played.
C = Player's original cost
4.56 is the number of points a 1 point cost player is expected to score over the entire round
11 is the total number of matches each team plays in a round.
(4.56 / 11 ~= .41)
Basically, this is figuring out what a player's point cost would be given their current performance (ie someone scoring 4.56 points in one game would have a cost of 11) this is then weighted with their original cost; the weights are percentage of games played for the "new" cost and percentage of games left for the original cost. I hope this makes sense, if not feel free to ask questions.
As an example let's say Leta (9 cost) plays in two matches the first week going 1-0 with a Team win in the first match and 2-0 in the second with an ace win and team win (4 pts 1st game, 9 pts 2nd game [3 wins * 2, 2 lineups * 1, 1 ace * 2, 2 team wins * 1, 1 Three Win Streak * 1]) = 13 pts for week 1 with two matches played out of 11 for the round:
T = (13 / .41) / 2 * 2/11 + 9 * (11-2/11)
T = 31.4 / 2 * 2/11 + 9 * 9/11
T = 15.7 * 2/11 + 9 * 9/11 (At this point we can see he has an estimated cost of 15.7 if he continues on this scoring trend whereas his original cost is 9--these are then weighted based on games played and games left)
T = 2.85 + 7.4
Trade Value = 10.21 (Leta's trade value at the end of week one with him scoring 13 pts over 2 games)
Trade Value is directly comparable with a players original cost thus someone with a Trade Value of 9 is rated the same as a player that started at Cost 9.
Adjusted Trade Value is what is used in the actual trade calculations and is simply Trade Value * Number of Games Left.
Thus, Leta at 10.21 with 9 games left would have an Adjusted Trade Value of 91.89; say free plays one game in week 1 and his new trade value is 9.5 (just picked this randomly fyi), free with 10 games left would have an Adjusted Trade Value of 95. Notice how free's Adjusted Trade Value is higher than Leta's because he has the extra game left to score additional points in.
|
For reference, a player scoring 0 points over two games in the first week will have the following trade value (note this is the absolute minimum trade value someone could attain after week 1):
+ Show Spoiler [0 pts over 2 games] +11 Cost: 9
10 Cost: 8.18
9 Cost: 7.36
8 Cost: 6.55
7 Cost: 5.73
6 Cost: 4.90
5 Cost: 4.09
4 Cost: 3.27
3 Cost: 2.45
2 Cost: 1.64
1 Cost: 0.82
0 Cost: 0
What if a player scores half their expected pts (so an 11 point cost player would score 4.56 pts over 2 games)
+ Show Spoiler [1/2 expected pts over 2 games] +11 Cost: 10
10 Cost: 9.09
9 Cost: 8.18
8 Cost: 7.28
7 Cost: 6.37
6 Cost: 5.45
5 Cost: 4.55
4 Cost: 3.64
3 Cost: 2.73
2 Cost: 1.82
1 Cost: 0.91
0 Cost: 0
Now what if a player scores 13 points over 2 games in the first week? (similar to the previous post example 3-0 (one ace) with 2 team wins):
+ Show Spoiler [13 pts over 2 games] +11 Cost: 11.85
10 Cost: 11.04
9 Cost: 10.22
8 Cost: 9.4
7 Cost: 8.58
6 Cost: 7.76
5 Cost: 6.95
4 Cost: 6.13
3 Cost: 5.31
2 Cost: 4.49
1 Cost: 3.67
0 Cost: 2.85
Now, assuming you pick a player that gets 13 points in week 1 over 2 games, what slumping players (0 score over 2 games in week 1) could you then turn around and trade him for?
+ Show Spoiler [Optimal Trades?] +0 Cost >>> 3 Cost
1 Cost >>> 4
2 Cost >>> 4
3 Cost >>> 5
4 Cost >>> 6
5 Cost >>> 7
6 Cost >>> 8
7 Cost >>> 9
8 Cost >>> 10
9+ Cost >>> 11
As you can see from these examples, the highest trade under more than optimal circumstances would be a player for someone that originally cost 3 points more than them. Remember though that the chances a 0 cost player is going to score 13 points week 1 are very slim. Thus, at most you're probably looking at a 2 point difference.
|
|
|
|
|
|
EPT
