0 |
Subject: ESPN player rater
Posted by: Edgar
- Leader [00458944] Thu, Sep 09, 2004, 04:51
Last year was my first playing Yahoo and it was great fun. I noticed there are several ways to rank the players available. One of the most useful I found is the ESPN player rater. I have been trying to figure out the basic formulas for each category to see how the formula compares players. It is also interesting how the various categories are combined to reach the overall total.
The player rater uses linear formulas for most cats. For example blocks and rebounds. The linear formula is based on an average. By simply plotting the ratings for example I found out the formula for three is: -1,47+(number of 3s of the player/ 47,33). This means a player gets zero points when he has made approximately 70 3s in the year. So this should be considered the average the rater uses.
On the other hand some categories work differently: the FG% and FT% points. This because the have to be the results of 2 factors: the actual % and the number of FG/FTs made. I have not cracked the formulas for these categories yet.
Can anybody help me to: - find out the fomulas used; - explain how the average is calculated - how categories compare - Give his/ her opinion on whether the formula used is fair and whether the way the rater totals points is fair.
Maybe this issue has been addressed earlier as well. If anyone knows any useful links I would be very interested. |
1 | Mike D Sustainer
ID: 41831612 Thu, Sep 09, 2004, 13:38
|
Baseball Player Rater formula discussion
ESPN actually provided some info. It looks like their link has changed, but the thread still has some helpful stuff.
|
2 | Edgar Leader
ID: 00458944 Thu, Sep 16, 2004, 04:14
|
Ok, here are all the ESPN formulas:
FG% = (FG%-0,4451)*FGA*0,02926 FT% = (FT%-0,7625)*FTA*0,03254 Points = (Points*0,002696) - 3,33 Rebounds = (Rebounds*0,0059778)- 2,92 Assists = (Assists*0,0091404) - 2,49 Steals = (Steals*0,0294483) - 2,66 Blocks = (Blocks*0,0244408) - 1,51 3-Pointers = (3pointers*0,0210917) - 1,47
The formulas are based around an overall average per category. If a player exactly matches this average he gets 0 points in the player rater.
While doing some research on the averages I found out that the averages the player rater uses is very close to the league average of the top 100 players in the rater.
Here are the averages of the top 100 players in the rater and the difference with the average used in the ESPN rater to calculate the points:
FG% = 44,81% - (0,66% higher then ESPN average) FT% = 77,11% - (1,13% higher then ESPN average) Points = 1197 - (3,04% lower then ESPN average) Rebounds = 479 (1,85 lower then ESPN average) Assists = 268 (1,85% lower then ESPN average) Steals = 89 (0,89% lower then ESPN average) Blocks = 62 (0,25% higher then ESPN average) 3Pointers = 70 (0,35% higher then ESPN average)
If you would like to make a good analysis for your own league I think it would be most fair not to use the top 100 players but the number of players used in the leagues starting spots. This means the top 100 would do if you are in a league with 10 players and 10 starting spots (10*10=100). Is the assumption correct?
Also next step is to see how the rater's different categories sum up into a total. I have not figured out how this works yet. Maybe someone can provide some help here. I will look at this myself as well and provide info when I find something.
My basic argumentation on how it sums up is that the total spread in a catogy should be related to the average the same for each cat. For example the total spread for rebounds is 992 (Garnett is the best and Jason Williemas the worst). While the average is 479. So the factor is 992/479. I would expect this factor to be related to the ESPN score for each cat but this is not the case it seems.
|
3 | Guru
ID: 330592710 Thu, Sep 16, 2004, 10:36
|
I use that type of approach for my own rating system. For the averages, I use the number of active players for all league rosters. So, if we have a 12 team league with 12 active players on each roster, then I use the top 144. To scale the ratings for each category, I use the standard deviation of the results in that category (for the top 144 players, in this example).
|
4 | Aris
ID: 166571316 Thu, Sep 16, 2004, 20:10
|
Edgar: use the standard deviation (as Guru suggests). What you call "spread" is too unstable. For instance, Ratliff's block numbers with the Blazers would make everyone else seem almost indifferent and useless.
I use: (REB - ave(REB))/std(REB) to get a score for each category and then add those, discount for injuries, more PT, bad karma, etc. Then, I finish in the middle of the pack because you cannot project stats for all players correctly and you cannot afford to take only players whose stats you can project (team stability, same PT, 4-8 years in the league).
|
5 | Edgar Leader
ID: 00458944 Fri, Sep 17, 2004, 04:22
|
Aris, the thing you state is also the way that ESPN seems to do it. The scores are 'normalised' which is (AVGof the player - AVG of the population)/ standard deviation of the population. This is the method to compare categories. However when you use this method to calculate scores they do not resemble the ESPN scores. I will try to post a comarison between both scores later. Maybe ESPN does use some method to factor in category scarcity, for example we know that blocks and steals are always hard to come by.
Also I have not figured out how to tackle the FG% en FT% correctly. This is more difficult since you need to take into account both the % and the number of attempts. Is it better to 'normalise' the population based on the formula (FG%- FGAVG)*FGA or is it better to normalise FG% and FGA separately and then multiply? The outcomes seem to differ from the ESPN rater as well.
|
6 | Aris
ID: 166571316 Fri, Sep 17, 2004, 20:58
|
I wasn't trying to reverse engineer the ESPN formula. I think you can add scores the way I describe. Why should a difference of 10 steals over a whole a season be considered significant? It's not.
Guru suggested a way for FG% and FT% a few years ago. I don't have time to explain it right now. Search the forum or wait and I'll post later.
|
| Rate this thread: | If you wish, you may rate this thread on scale of 1-5. Ratings should indicate how valuable or interesting you believe this thread would be to other users of this forum. A '5' means that this thread is a 'must read'. A '1' means that this is a complete waste of time. If you have previously rated this thread, rating it again will delete your previous rating. If you do not want to rate this thread, but want to see how others have rated it, then click the button without entering a rating, or else click here. |
|
|
Post a reply to this message: (But first, how about checking out this sponsor?)
|
|