Friday, February 10, 2012

Gameweek 25 Captain Data

P90 Points scored per 90 minutes at home/away based on this week's opponents QG% Percentage of games played where the player scored 6 or more points xPct The percentage of his team's goals a player has accounted for, multiplied by the expected goals for this week Rating Based on the player's rankings in the other three categories.

I've finally taken the plunge and tried to provide some hard rankings rather than just objective data. The rating column is based on a player's performance in the other three categories, currently weighted towards the xPct ranking, which I believe to be the best indicator of future success. Over the next few weeks I am going to be posting a few reviews to try and determine the best weighting for each of these factors but until then I simply going with 50% xPct, 25% P90 and 25% QG%. I feel this gives the best chance to recognise underrated players, without giving too much attention to players who have played sporadically (Drenthe kept popping up in the old rankings) and also avoiding overvaluing players who assist too many goals (who obviously pickup less points than their goalscoring colleagues).

I must say, I am surprised to see Ba come in second as he has a much harder fixture than Bale, Dempsey and even Rooney on paper, but when you're accounting for 59% of your team's goals (second only to Van Persie's 65%) you're always going to be riding high.

After last week's captain-gate I'd be remiss if I didn't at least mention that Van Persie is still due for a rest at some point and Arsenal have to travel to Milan in midweek. You wonder though if he will try and power through until Gervinho returns from the ACN, although Chamakh is said to be fit again. I guess the lesson last week, if there was one, is probably to not be overly concerned with downside in the face of such extreme upside. Blackburn at home is about as good as it gets on paper and the risk of a 1 point effort was probably eclipsed by the potential spoils Van Persie could (and did) bring to your team. This week is less clear cut and the decision to opt for a Rooney or Bale is probably more logical. At the moment though, I am sticking with Van Persie (caveat emptor for all).

I will try and add to this tomorrow night with any late fitness or team news, but if nothing emerges, the rankings for the week will stay as they are above.


mike said...

Chris (or anyone) - I need to bench one - right now I have Fletcher on the bench, but I'm going back and forth between Valencia and Fletcher. I'm a bit worried that Valencia may be spending more time at right back. The point is moot if RVP rests, but any suggestions who to bench this week?

Dempsey - Bale - Sess - Silva - Valencia
Ba (VC) - RVP (C) - Fletcher

Kalix said...

I think weighting QG and P90 at all is somewhat flawed. The main reason being that they don't account for the fixture in question. If i had to captain one player for a whole season, then maybe it would be useful.

A (simplified) example:
A player always gets 2 pointers when his team plays one of the top6. But he gets 6+ points in every other match. This could easily give him a QG% score of 75%, and a high average p90.

Under the current weighting, his overall captain score will be inflated vs higher teams (where he always scores 2 points, obviously not a good captain).

Typically when we choose our captains its based entirely around the fixture in question. so giving a 50% weighting to data which ignores the current fixture seems counter-intuitive.

I think xPct alone has served us very well over the last couple of weeks (yes yes, small sample size, but i think it will be a success in the long run too!).

Kalix said...

I do think somehow accounting for assists vs goals in the xPct is a great idea though :)

makes perfect sense.

Kace said...


Firstly, Do you think that with Adebayor's current streak of low scoring games, that he should be as far up the table as he is? Is there a way to account for recent form in these rankings? Secondly, I believe that if it were possible to account for ups and downs in form, to put into a line graph format would be beneficial so that people could make out form vs. points scored against a timeline of gameweeks. Might be hard to do but the result could be very interesting I think.

Fozz said...

Where would Sigurdsson fit into all this? Good fixture and he has 2 good and 2 poor returns out of his lowly 4 games thus far.
I guess he hasn't played enough games to figure but (and I agree with Kalix) the fixture this week places him (in my view) well above the likes of Kuyt etc.

Chris Glover said...

Mike - that's a great looking side and it's really hard to bench anyone from that group. I would play Fletcher given that Wolves appear more likely to score several where as United have struggled against Liverpool in past years, though they did score 3 last season. It's really close but with Nani and Young now supposedly fit, Valencia has an increased risk of rotation or at least early substitution which might be the tie breaker.

Chris Glover said...

Kalix - interesting inputs and I'm interested to hear what others think as this is very much a work in progress. A couple of points though. The xPct rating is not based on fixture alone as it accounts for a player's scoring/assist record to date. Fixture alone would lead us to all captaining Bolton players this week, which would obviously be a foolish move.

You're point on a player having success against weaker sides is a very good one, and it's something I have tried to measure to date. Indeed, I think I have sufficient data for this and last season to make this work so I will try and get to this over the next couple of weeks. I do think though that you need to know a players' overall production as rightly or wrongly, there is no way in hell anyone is captaining Danny Graham this week despite him having a xPct higher than Dempsey and Van Persie.

Quality games is definately the weakest of the factors and I may well remove this. However, I do feel that you need to account for players who are cosistently good rather than throwing out the odd 19 pointer and a bunch of 2s. Of course, it depends on how you want to pick your captain, but I try and pick a player who has a high percentage of adding 8 points rather than taking risks on a guy who might get 15 or might get nothing. As with Van Persie last week, perhaps that makes me too risk averse, but hopefully the other data provided will allow you to make your own decision there (hence my hestitancy to rank the cpt picks in the first place).

Thanks for the input though, and please respond below if there's something you would recommend including in the rankings.

Chris Glover said...

Kace, interesting ideas. I used to factor 'form' into my captain rankings but I did some analysis earlier in the year which, for goals at least, showed little predictive value in a player scoring this week on next week's result. The analysis was overly simplistic and needs to be developed but I'm not sure how much I really believe in 'form' in the way we understand it.

Van Persie isn't in good form, he's just really good. Indeed, he's actually scored at a similar rate in the past, just never over such a sustained period, so the key driver to his success is really fitness. Some players do appear to be streaky (I always felt Rooney was) but I didn't really find much evidence that this was the case. If you take toss a coin 16 times, you have a 31% chance that 3 of those tosses come up heads but we wouldn't suggest that the coin is in form. Indeed you have a 6% chance that it's heads every time. Players, of course, aren't robots (or coins) and do have feelings, emotions etc but from the stats I've seen, player talent and the week's fixture have a much bigger influence on the score for the week, hence my neglect of form of late.

Chris Glover said...

Fozz - Sigurdsson isn't in the listing as he doesn't meet the minimum minutes. I have to draw the line somewhere, but yes he's done well of late. That said, he isn't that close to captain territory for me.

Kalix said...

Hey chris,

Thanks for the reply.

xPct is a teams predicted goals for the week, modified to estimate how many of those goals a single player is involved in. So at its core it's still based on the next fixture. This is what i mean when i say the other two are fixture independant, but xPct is not.

Good points about not wanting to captain players with two 19's and a load of 2's. This may be where the limits of computer intelligence comes in...

We look at those players' history screens on the FPL site and think: "no way i'm captaining him", takes us about 2 seconds.
But getting some code/stats alone to quantify those kind of judgement calls might be so complex its not worth saving ourselves the 5-10 seconds it takes to make the judgement ourselves :P

Keep up the great work!
Always pushing the limits of what stats & trends can do for us

vanilla said...

Thanks again for the blog, Chris. think there's a post coming up where you evaluate the Strength Of Schedule for the remainder of the season?

Kace said...


It was funny that I focussed on Adebayor in my comment because he's pulled out a 20pt gameweek here after weeks of poor results - good practical reinforcement of the point you had made earlier :)

amtosh said...

Chris (or anyone) - I have a question.

This week is the first round of an FA cup style comp in my minileague, so I need to beat the other guy's points tally to progress.

My team at the moment is:

Vorm (Mignolet)
Simpson Kompany Evans (Enrique, S. Kelly)
Sessegnon Dempsey Bale Silva (Moses)
Dzeko Ba RVP

2.8 in bank

I'm afraid that the return of Balotelli and (possibly) Tevez means that Dzeko's on the outer, and Aguero is obviously the best choice.

I can swap Dzeko for Aguero as long as I swap Moses for Surman, but I will lose 4 points.


Is it worth losing 4 points?

amtosh said...

Is McLean's form sustainable?

Since MON took over, he's averaging 7.29 points a game with 4 CS, 3 goals and 2 assists in 7 games.

What do the numbers say Chris? Have his underlying stats been good?

I'm thinking of dropping Sess for McLean to free up some dollars.

vanilla said...

Maybe I missed mention of a vacation, but no post since 2/11 on Twitter either... hope you're well, Chris!

amtosh said...

Nah, he's retweeted a couple since then including something yesterday.

Must just be taking a well-deserved break from EPL while he can.