League points based rating system for roto leagues. Besides using mean and standard deviation, there is another potentially more accurate way of player rating. This system tries to estimate how many league points a player will contribute assuming every other starter you have is average. According to my calculations, KG last year would have contributed 12 points in a 9 category league. Of course, this assumes that you have not otherwise maxed out in any categories already. The downside of this system is that you need to collect a lot more information. You need to find out how many assists per game more the #1 guy in your league had than the #2 guy and then #2 over #3 and so forth. What you are getting here is how many assists per game separates one position in the assist category. You will then need to multiply that number by X (X = # of starters in your team). That number is the denominator for each of your categories where you divide (assist - average assists) by the denominator to get the player rating for that category. You need to do this for each category. Now, if that wasn’t complicated enough, there is a question of robustness of the data and outlier teams that you have to deal with. For better accuracy, I would recommend taking the full year information for at least 10-15 leagues and then also subjectively eliminating the outliers like the top and bottom category teams.

I have talked about using mean and standard deviation before when creating a player rater. Just to review for each category.

Mean: (rebounds-average player rebounds)/average player rebounds
SD: (rebounds-average player rebounds)/standard deviation of rebounds

I personally prefer just using the mean rather than adjusting for volatility. Here, I think simplicity makes more sense. The advantages of mean is that over a cumulative season, 10-12 starters, and active management to balance categories, the probability distribution is very much centered around the mean anyway. The full season stats and active management effectively dampens much of the volatility anyway.

So now that we have the rating info so we have to interpret the data. If you have a total rating sum of 0 across thee categories, you have an average player. If you have 9 categories and the player has a rating of 9, then he is as good as 2 average players. That will not happen but it gives you a reference point when comparing players. If you do not want to set the average player with a rating of 9, you can also do that. Just add nine to everyone else's rating. Ratings are useful but do not lose sight of category needs. In that case, the player rating may not matter.

The % categories are the same calculation for a player rater analysis so I will just go over the FG% calculation and you can do the same to come up with the FT%. First, you have to calculate the average FG% of all players (top 120 if you are in a 12 team x 10 starters). Do the same for FGM and FGMsd. The FGM and FGMsd are for the correct weighting of the category. For example, Dwight Howard and Brad Miller had a higher FG% than Garnett last year but because Garnett shot more last year, he actually has a higher contribution in the FG% than either of those players.

According a recent PR calculation that I did, the averages for FG% stats were 45.7%, 5.9, and 7.2. Separately, the averages for FT% stats were 77.0%, 3.4, and 1.1. Now for some math. We need to get FGmsd% which is (1-FG%) = 54.3%. We have all the component parts for the calculation now. The calculation is (FGMsd% x FGM) - (FG% x FGMsd). The average player shooting 45.7% will have a rating of 0 and anyone shooing better or worse than that will have a positive or negative rating.

This deserves another segue. This is a great tool to use when you are drafting and how to balance your strengths when it come to the % cats. We all know that players like Iverson (although less so last year) and Shaq hurt the FG% and FT% a lot. I have heard some people say that all you need to do is pair the two together and you should be alright in FT%. While that is true in FG%, that is simply untrue for FT%. Iverson and Shaq have a FG% rating of -0.8 and 2.1 so you are still positive. However for FT% rating, they are 0.7 and -3.2, respectively. Yes, that is a negative 3.2 and the sum of the two is –2.5. You are still going to lose the FT% cat by a mile. To put it in perspective, if you have Shaq, you will need Dirk, Peja, Ray Allen, Nash and Billips just to get back to 77.0% and be average in that category. If you have Shaq, move on.

Here's are the steps to create your own player rater for a roto style or H2H league:

1) Download all the relevant stats (depending on your categories) onto a spreadsheet. Make sure the stats are on a per game basis.

2) Most of the cats are similar and straight-forward except for any % cats like FG% and FT%. For FG% and FT%, you will need to download FGM and FTM for each player (you can also download FGA and FTA if you want to). The bottom line is that we need to tabulate FGM and FGMsd and FTM and FTMsd to correctly weight the % categories.

3) Find the average of each of the cats for the top x (if your league starts 10 and there are 12 GMs, then x=120) number of players. This will serve as your reference point for an average player. A player rater is just a way of figuring how good a player is in reference to an average player.

4) Now that you have all the raw info in your hand, this is where it gets tricky. I will exclude the % cats here for now since they are different and deal with just the cumulative cats (pts, 3s, ast, reb, etc…). Philosophically, in calculating how a player's rating, some people like to use the mean and some people like to use some sort of volatility adjusted mean.

5) Using the mean is straight-forward. If the average player (in a 2 cat league) gets say 15 pts and 1 blk and player x gets 30 pts and 1.5 blks then player x will get a rating of 1.5 (1 pt for pts and 0.5 pt for blk). The average player is always rated 0. You just do the same for all the cumulative cats.

6) The reason some people like to use volatility adjusted mean is that they say that cats like 3s and blks are very different between players. What they are indirectly saying is that you can catch up in 3s and blks a lot easier than pts and asts. Hence they use the excess over am average player divided by the standard deviation of the cat to derive the rating for a player.

I will leave the % cats for a later post.

Lets get some basics down about a player rater. First, I am assuming that people are playing in a scoring system where all categories are weighted the same in Roto and H2H as opposed to some scoring systems where points are worth 1 point, assists 3 pts, etc... Now the basics:

1) Player raters should be based on average stats per game and not total unless your league allows you to go above 82 games per season. Some may ask, "what about injury prone people? Wouldn't it be better to use total?". No! Average is still better and you can make adjustments for average if a player is injury prone.

2) All categories are worth the same. Winning the points category is just like winning the Blocks category. I thought that I said this already but it doesn't hurt to repeat it.

3) PRs are based on averages. That means how much better is a player over the average player.

4) Previous performance is still the best indication of future performance. For exceptional situtations, you can make projections which should be highly tied into how many minutes a baller is expected to play this season.

5) Number of categories, number of starters, and number of players in your league matters. There is no one size fits all. The average player in a 10 (starters) x15 (players) league is much different that an average player in a 10x10 league.

6) There is no additional premiums for positional eligibility (centers are usually the most valuable), although there are ways to adjust for that. I'll get to that one of these days.

Next post will be on what you need to actually get down with it.

To be a successful fantasy b-ball player, there is nothing more important than a player rater. There are many people that think that they know the players well enough and do not need a player rater. Those are probably the same people that pick Shaq instead of Marion or Jamison way ahead of Kurt Thomas. Needless to say, based on last year, Marion and Thomas were the better fantasy players. If you take a look at any player rater, the numbers do not lie. That's how you find the next Doug Christie, Eddie Jones and AK-47 of yester-year. Incidentally, there's someone that looks like the old Christie and Jones this year. That person is Iguodala. He's not going to score 20 points but he is going to be above average in every other category.

You can probably find a player rater in your fantasy league pages. However, those may not be your best bet. For example, Yahoo's player rater sucks period. One PR that is free and useful is the Basketball Monster PR. It is one of the few PRs that allows you to customize your categories and roster size (which is usually one of the overlooked things). Espn.com has a pretty useful and widely used player rater but it is no longer for free. I prefer the Basketball Monster PR anyway. For the avid baller who wants to squeeze out any advantage that he can, the only way is to construct your own PR, which can be easily done using Excel. Regardless of what you decide to do, consult your player rater before you draft. You will not be disappointed.