Friday, 25 January 2013

GW24 Buying Guide [Explained]

I've been working on a new way of representing transfer targets for this Buying Guide.

This post will serve as the explanation.  The results are found on the Buying Guide page and I  link to it at the end of this this post.


First though, the explanation. 

One thing that has intrigued my through setting up this site and analysing player data and their points projections is how a player's points projection compares relative to other players in the same position and relative to players in other positions.  The requirement in your FPL squad of 5 Defenders, 5 Midfielders, and 3 Forwards places constraints on your selection. 

The available selection of just 3 forwards is crucial, but the thing with forwards is they invariably score all the goals and every, or most, teams have a "striker" who you can reasonably expect and predict a set number of goals (e.g. Norwich's Striker might score 10-12 goals, Liverpool's might score 15-20, Man United's 25-30). 

To try and bring this context to my projection data I have defined a few terms that are based on relative points rather than absolute points.  The theory being that you could theoretically fill your entire squad with cheap , but fairly decent players, and expect a weekly gameweek score of, say. 25 points.  The assessment therefore is how many more points would you get by signing Striker X compared to bringing Defender Y??



 
PAL - Points Above Lowest:
For each position there is a player in my projection data who has the lowest projected points forecast for his position.  Generally, defenders have the lowest projection and forwards the highest.  The PAL value is how many more points the Player in question has over this lowest forecast player.
Example:  Agbonglahor is the lowest ranked forward in the data with a  forecast over 10 weeks of 33 points.  Robin Van Persie is forecast 75 pts over the next 10 GW so is +42 PAL.

PAN - Points Above Next
Similar to PAL but this is simply how many more points a player is to forecast compared to the next best ranked player in his position.  This gives a way to see the relative ranking in pecking order of a player at a glance.

£price; %own
Fairly self-explanatory I hope

Value Add
This is a calculation of a player PAL divided by their price.  

Deviation
This is an important parameter I think - but I don't know how yet, or whether a positive or negative value is good or bad.  Essentially it's how different a player's statistical performance (F.SCORE) differs from their actual points scored.   

A positive value means the players has more points than you would expect from their shot data.  A negative value indicates they have not scored the actual points from their actual shot data according to the model.  A couple of extreme players in this respect are Walters and Walcott.  The former has a DEV -7 and has not scored as many points as the model would expect.  Walcott on the other hand with  DEV +12 has scored many more points than one would expect, based on league average conversion rates.  

The decision as to whether a player performance above or below the model is sustainable or due to regress is up to you.  In my example, I'd suggest Walters and Walcott's deviations are accurate due to the contrasting play styles and quality of the two teams in question. 

Hope this makes some sense, please let me know in the comments if not, or anything can be explained any better.  

Here's the Buying Guide's "Transfer Targets" for the next 10 GW (GW24- GW33).  In future, once this is bedded in a bit, I will add some of my thoughts and narrative on the players.

3 comments:

  1. Sir, I enjoy your blog immensely. It is thoughtful & well measured. Thank you for all your work.

    Further, I recognize am woefully under qualified to engage you in a meaningful dialogue regarding your statistical analysis of players.

    With that being said, I would like to discuss "PAL - Points Above Lowest"...

    There is a 'sabermetric' baseball statistic called Wins Above Replacement (WAR) that is used to show how many more wins a player would give a team as opposed to a 'replacement level' or bench player at that position.

    Now, I feel that being able to compare a 'top player' to someone who is 'average' or 'middle of the road' may be more valuable than comparing them to the 'worst' at their position.

    That is to say, I would see more value in comparing Van Persie to an 'average' striker or someone I may actually consider having in my side. (eg. Crouch or Holt) As opposed to the worst player who wouldn't warrant a spot in my side.

    Thank you in advance & I welcome hearing your opinion regarding "PAL" vs. a comparison to an 'average' player.

    ReplyDelete
  2. Kyle, this is perhaps the most splendid comment I have ever had! WAR and sabremetrics are something I am trying to learn from and I have thought about doing Points Above Average, or even Points Above Median, and will do exactly this for the GW25 update.

    ReplyDelete
  3. Any more suggestions drop me a line at shots_on_target@hotmail.co.uk or @shots_on_target on Twitter.

    Thanks Kyle!!

    ReplyDelete