Discussion around good; very good or genius FPL players

[hancockjr]

I thought I'd posted before, but it doesn't seem to have worked, so briefly:

- Many many thanks to ChrisA, both for HoF and posting

- If you want a different ranking, you need to decide what it is trying to do and have a way of testing it.

Assuming the data exists, I'd suggest the ranking should be trying to predict who will do the best next year. Therefore you need 2 things:

- A proposed measure for success
- Proposed ranking methods/options

You can use the measure to assess the proposals.

As an example, I'd say for the measure you should take players who are accepted as good (a lot of them) and see which method best predicts their actual finsihing positions best. This can be done retrospectively, for instance using data up to 2 years ago to predict for last year's finsihing positions, and also using data up to 1 year ago to predict this year's results.

If the data is avaliable on Excel I'd give it a bash.

Options for ranking methods would include weighting of years, whether to use points, positions or something else (I think a log scale of finishing positions is probably optimal) and how to deal with people with differing years, and also maybe even discrding very poor years.

Those are my thoughts anyway, but again I'd like to say Chris A has done a great job!

[Stemania]

I agree entirely with that. The only two things i have particularly strong opinions on are

1) That the number of entrants each year should be taken into account.

2) That the data in any ranking table is wherever possible expressed in terms of finishing positions rather than points (regardless of how the rankings are actually calculated).

I do prefer the newer system ChisA proposed to the old one, but as Mav and hjr have said it's all a bit too subjective to be meaningful without a pre-specified goal.

[MoSe]

wow, I went more or less awol last week, and it looks like I have some serious ketchup to do now

also, I had promised I'd have expounded my thoughts on the rating methods started in Billy Whiz's "consistent players" topic, the one or two interested readers will have to forgive my slacking, of course my considerations would have connected with the above ones

I mostly find myself in agreement with hancockjr summary

great thanks to ChrisA for sharing and explaining what's behind the HoF
and... smartypants posting on FISO!

please allow me to briefly piggyback and drop my £.02 for now

It's an interesting debate. In fact, I'll have another go, this summer, at trying to work out what the optimum percentage for the downweight is. Should be pretty easy to do- look at optimising the correlation between what their HoF score would be without this season and their points this season.

Now, I was having a think about the downweight. [...] To me, this feels much fairer and I'll be looking to implement this in the future. What do you guys think?

I guess the thing I'm not sure on is how you can scientifically justify any of the factors that are applied to someone with a record that suffers as a result of weighting changes.

Could it even be an option for a viewer to choose their own weighting factor (100%, 80% etc) with a button to update the table for their personal view?

at first I thought too the factor you'd choose would just be arbitrary, and merely had to be chosen to best suit the purpose you had in mind, legitimately as long as you declared it, as you did by posting in this topic!
The idea to optimise the correlation with the latest season results, in order to turn it into a predictive tool, it's rather interesting, although at first guess a huge amount of noise would have to be filtered out...
Stem idea would allow each user to verifiy for himself the effects of modifying the weight
I had in mind myself to propose a "study" on the weighing factor to plot its effects on the "HoF space"...
in general theory, you should get something like a series of differently shaped lines, which should be enveloped by a surface, much like with a Quadric
http://en.wikipedia.org/wiki/Quadric
in simpler terms, you record for a notable set of players (say Top 10, or Top 20) what their HoF rating would be, by slowly changing the wheighing factor (you should limit to one factor if want to stick to our 3 dimensions ) by say 0.01
the purpose of this would be to highlight when the HoF positions would CHANGE
e.g. going from .85 to .86 3rd and 4th place would swap (totally making this up, for exemplification purposes)
and from there you start to reason about what the weight represents in real HoF terms and which is desirable in practice

IN SUMMARY that's indeed what hancockjr said:

- If you want a different ranking, you need to decide what it is trying to do and have a way of testing it.

___

Re: the idea of weighing itself, in the "consistent players" topic I referenced above, CK had proposed to use a "Weighted rolling average" or WMA, which is a simple concept most of us would know, using a linear factor
in the same wiki page you'll find EWMA too, that is a moving average using an Exponential factor
http://en.wikipedia.org/wiki/Moving_ave ... ng_average
that's exactly the concept you're applying, wiki doesn't actually bring any groundbreaking info but could be a useful reference for all yours considerations
___

(I think a log scale of finishing positions is probably optimal)

in my ranking charts (posted mostly in the Overall Rank topic) I used indeed a base 10 log scale, merely because that was the excel tool/option at hand
can't recallt if I had already hinted it in the "consistent players" topic, anyway another option worth considering is the Harmonic Mean
http://en.wikipedia.org/wiki/Harmonic_mean

Typically, it is appropriate for situations when the average of rates is desired.

the wiki page should provide some food for thought, and in short it's the less-punishing for bad seasons (or conversely, it's the most rewarding for top results)
___

finally

1) That the number of entrants each year should be taken into account.

to that end, I think I had already quoted elsehwere smartypants FPL History analysis https://fpldiscovery.wordpress.com/2014 ... l-history/
where I took this graph from

I had started considering the idea of using GW1 entrants (instead of end-of-season ones) but then I hadn't time to properly reflect on it (i.e. to analyse the ranking distribution of those who started GW2, GW3, etc.. and thus their impact on one's ranking)

something similar should be also done if/when using points instead of ranks, albeit in this case is much more difficult to choose a fair "normalising" factor: Overall DT? Season Winner's score? something factoring Top10/100/1k scores? SUM of each GW Highest? SUM of GW DTs? A combination of the above?
S_P's articles provide insights on such proposals too
______

forgive me if I had to be brief but real work is calling now

============

PS:
lest I forget it in the near future:
let me borrow this graph

and imagine it's instead a Points(Force)/Rank(Extension) representation for a given season
given you have a points score somewhere in the middle, even if the Season winner score is THE SAME, your same points score would have a completely different meaning and value if the bulk of FPL players had a "standard" performance, with proportionally more teams at the lower scores (the red line), as opposed to a season where a lot of teams got higher than usual scores, "bunching at the top" (blue line)
and this can only be represented by your RANK

[Stemania]

Good God MoSe (I enjoyed it)!

I am nominally on the side of Ranks are a better measure than Points for the record. No-one goes into a season aiming for 2200 points or judges their own performance at the end of a season on how many points they scored. It's meaningless. Everyone cares only about rank.

Nobody judges how successful an EPL team has been in the league in the last decade on the number of points they've scored - it's always expressed in terms of the number of titles or the number of top 4 finishes etc. Any measure is pretty arbitrary ans subjective, but at least a ranking based one carries some sort of intuitive meaning of success (although if the process is complicated enough, this will probably be lost anyway making the method used pretty unimportant).

[Carlos Kickaball]

Points gets over the problem of the amount of entrants for a start.

I think there seems to be a lot of bias towards ranks here, specifically because you do indeed play for a rank each year, but the aim of the game is to be the best at getting points (and your actions really have very little effect on other managers).

Rank is only a measure of comparative points over a season, why if you are deciding who the best manager is over a period of seasons you'd decide to chop up the history into pieces and take ranks, I'm not sure. You wouldn't chop the season in to quarters and then assess the quarterly ranks to decide who the best manager was.

given you have a points score somewhere in the middle, even if the Season winner score is THE SAME, your same points score would have a completely different meaning and value if the bulk of FPL players had a "standard" performance, with proportionally more teams at the lower scores (the red line), as opposed to a season where a lot of teams got higher than usual scores, "bunching at the top" (blue line)
and this can only be represented by your RANK

If we are comparing records over the same time however, this criticism does not hold. If you win with a low amount of points, you still make ground everyone else who played that season in the hall of fame, and if you have a poor rank with a high amount of points you still lose ground on everyone who was ranked above you.

Using points and comparing totals (or averages) over a long period is still a comparative measure.

Nobody judges how successful an EPL team has been in the league in the last decade on the number of points they've scored - it's always expressed in terms of the number of titles or the number of top 4 finishes etc. Any measure is pretty arbitrary ans subjective, but at least a ranking based one carries some sort of intuitive meaning of success (although if the process is complicated enough, this will probably be lost anyway making the method used pretty unimportant).

And similarly you wouldn't say Ville is rubbish because he has never won the FPL title. Due to the amount of entrants, and the way the game works, it's completely different.

[davidoff]

Is there anyway I could view the the overall rankings or my own previous rankings before the 06/07 season?

I'm sure I finished around 55th overall in one of the previous seasons (awful memory) I would love to know if I did.

Thanks in advance.

[Stemania]

I think there seems to be a lot of bias towards ranks here, specifically because you do indeed play for a rank each year, but the aim of the game is to be the best at getting points (and your actions really have very little effect on other managers).

Does there?

I don't think anyone has stated any strong opinions on the matter in this thread. Many posters have expressed arguments for preferred weighting in terms of ranks, for example "which is why in imo, finishes of say 3k,2k,3k is more respectable than a 300, 4k, 25k nowadays". But that's because it's the only language that it makes sense in. The sentence "which is why in imo, finishes of say 2311,2287,2210 is more respectable than a 2454, 2293, 2122 nowadays" is pretty meaningless.

Rank is only a measure of comparative points over a season, why if you are deciding who the best manager is over a period of seasons you'd decide to chop up the history into pieces and take ranks. You wouldn't chop the season in to quarters and then assess the quarterly ranks to decide who the best manager was.

No you wouldn't, but that's because nobody cares about quarterly ranks. You wouldn't measure any real life sports team in terms of how many points they've amassed over a decade, because nobody cares about that either.

Look, I don't really want to see the thread break out into an argument over points vs rank - the best should depend on what you are trying to show or find a measure for as hjr says, not on some pre-existing preference.

[Carlos Kickaball]

I think there seems to be a lot of bias towards ranks here, specifically because you do indeed play for a rank each year, but the aim of the game is to be the best at getting points (and your actions really have very little effect on other managers).

Does there?

I don't think anyone has stated any strong opinions on the matter in this thread. Many posters have expressed arguments for preferred weighting in terms of ranks, for example "which is why in imo, finishes of say 3k,2k,3k is more respectable than a 300, 4k, 25k nowadays". But that's because it's the only language that it makes sense in. The sentence "which is why in imo, finishes of say 2311,2287,2210 is more respectable than a 2454, 2293, 2122 nowadays" is pretty meaningless.

Yes there is, and I explained why in the rest of the quote, it's the only reason you'd frame the argument the way you just have. Of course finishing points on their own are meaningless without the context of other scores, but comparing and then ranking the total scores themselves over that period of time is meaningful.

You may as well say season ranks are meaningless because they only compare points.

[MoSe]

... I am nominally on the side of Ranks are a better measure than Points for the record.

yesyes, I'm on the side of the ranks too, but I didn't want to rekindle that discussion here (*damn TOO LATE, I should've known better )...


...I'm wholly satisfied with hjr approach

Options for ranking methods would include weighting of years, whether to use points, positions or something else (I think a log scale of finishing positions is probably optimal) and how to deal with people with differing years, and also maybe even discrding very poor years.

leaving/keeping mind open also to options we disagree with

to be fair, once CK explained it, using plain pts has its rationale I can acknowledge even in disagreement, but I wanted to address in detail my criticism to it in another thread (* alas I see CK has already jumped on it as I was replying to Stem)

I think the subject it's too much a sidetrack in this topic, forgive me if I added as a Post Scriptum the Elastic_Hysteresis picture here as I had it at hand
I'll only very briefly comment:
viewtopic.php?p=2712493#p2712493
on reviewing it, I decided to refrain form commenting it here, as I 'd have defeated my very own good purpose,
and moved it to its dedicated topic instead

now I really have to give woerk the priority, as I'll probably have this whole week...

[Stemania]

Is there anyway I could view the the overall rankings or my own previous rankings before the 06/07 season?

I'm sure I finished around 55th overall in one of the previous seasons (awful memory) I would love to know if I did.

Thanks in advance.

I don't think so I'm afraid.

[MoSe]

what you two made me forget to post, it's that it wasn't my intention to turn the HoF into such a hassle

definitely not a hassle-HoF !

[Stemania]

to that end, I think I had already quoted elsehwere smartypants FPL History analysis https://fpldiscovery.wordpress.com/2014 ... l-history/
where I took this graph from

I had started considering the idea of using GW1 entrants (instead of end-of-season ones) but then I hadn't time to properly reflect on it (i.e. to analyse the ranking distribution of those who started GW2, GW3, etc.. and thus their impact on one's ranking)

It's very useful info that (thanks s_p and MoSe). It's interesting to notice how the dip in 12/13 doesn't seem to have effected the steady rise in popularity. If I remember rightly, that dip was due to a humongous screw up by fpl in which their site was down for a couple of days before the GW1 deadline. I think the GW1 scores were essentially wiped (or at least there was an option of starting your leagues from GW2) and everyone got a GW2 wildcard.


(PS: If anyone knows, can someone possible PM me with how you can resize images linked to a URL in a given post - sorry, the one above is biiiiiig and I'm not so good at php!)

[MoSe]

...
Thanks for the clarification triggerlips, also it would be great to have you posting more frequently here as I'm one of the ones who was introduced to your blog by liquidfootball2 this season and I enjoyed (and agreed with) pretty much everything you posted there.

spurred by the recent flurry of celebrity posters, and by the several interesting subjects touched, I remember I had read on Triggerlips blog one statement which raised my curiosity
https://whichtransfer.wordpress.com/201 ... in-review/

If captains didn’t receive double points my rank would have ben much lower

without any polemic intention in the least, I sincerely wondered whether that was just his gut feeling or if he had based it on some actual data analysis

profiting now of SmartyPants having posted here too, from his Season 2014/15 Summary – Ownerships section, chapter 4. Captaincy

How did the most popular captain choices of the top 10K perform?
Frankly, not so good. Picking most popular captain each week would have brought you only 240 points (480 points if doubled), or 6.3 (12.6) points on average.

• from the FPL Statistico site TL used too in his blog for his captains breakdown, we see indeed TL earned 292pts with Captains (582 if doubled)
  That's 52p more than the Top10k "most popular" choices each GW!
  with 2257p instead of 2309 he'd have ranked 2k places lower (2599 instead of 592)
  
  although "most popular" is not exactly the same as the average in all Top10k teams
  here SP comes to help again in his last sections Season 2014/15 Summary – Squads 3. Average Points for Captains, Bench, Autosubs
  the weekly average is even lower, 12.41p instead of 12.6 for the "poll winner"
  and you also have to consider that's not the average captain points of the teams who ended the season in top10k, but of the Top10k teams entering each GW

the above data supports TL statement that he benefited from Captain picks significantly better than the Top10k average and/or "mode"
he'd have lost ~ 2k places had he picked like the other Top10k did

and still he has whetted my curiosity about his exact statement, that is "If captains didn’t receive double points" at all
I wonder if that's something someone with the programming skill & tools (which I lack) used for the above sites could easily sketch and obtain...
I'd have to individually insert and check the "Total Base Points for the above Gameweeks (no captain doubling)" line from FPL Statistico.... (TL is 2018)
I might indeed be trying it for top 20, or top 50, surely not for top 10k!!!

(BTW, personally I got +233p from C, which is almost in line with Top10k avg [not mode] of 236p, despite I'm much lower ranked at 175k)

[triggerlips]

It seems more sensible to give a higher weighting to recent years, as they more fairly reflect current ability. Most people who have used the same account since the beginning (like myself) have a few shocking stats in the early years as they learnt to play. Why should a season from 2008 carry the same weighting as this season for example, when in effect it was nothing more than a dead team.
It gives an unfair advantage to those who discarded their original accounts, starting new ones to make the stats look better

[Wyld]

Any sign of the Ville yet?

[Stemania]

Any sign of the Ville yet?

Here you go:

http://fantasy.premierleague.com/entry/ ... history/1/

His progress/moves (along with a few other top end players) are being followed and analysed in the FPL Stalker thread (in the side games section):

viewtopic.php?f=82&t=116788

So, that thread is doing one of the jobs this thread did last year. Everyone is still (as far as I know) free to pick a top player to report on that hasn't already been chosen if they want - though I guess there needs to be a reasonable limit on the number to keep the thread readable, whatever Mo Bot decides. If it becomes popular enough I might even move it to the main forum.

[Le Red]

Wow I got 11 players in common with him
Guess my thought process is in the right direction then

[Impossible_United]

I was better than him GW1 And i have 7 players in common with him not that bad too. But i prefer my forwards than his

[WINNING!]

A few interesting blogs of note:

https://www.triggerlips.com/
https://fpldiva.wordpress.com/
https://innsie.wordpress.com/
https://matthewmartyniak.wordpress.com/about/

[Carlos Kickaball]

A popular thread last year, and I'm interested to see how the top 10 in the HoF fare this season.

Top ten HoF direct links

1. Ville Rönkä
2. Triggerlips
3. Jay Egersdorff
4. Martin Button
5. Derek Simpson
6. Craig Johnson
7. Alianto Henry
8. Simon Walsh
9. Martin Gupta
10. Jon Sumner

Notable others

15. Matthew Martyniak - Previous winner and FISO regular.
17. Julian Hancock - FISO regular with excellent record.
22. David Meechan - Dropped out of top 10 despite top 10k finish.
24. Rasmus Sundman - Dropped out of top 10 despite top 10k finish.
32. Ulrik Nylund - Dropped out of top 10 despite top 10k finish.
41. Simon March - Last season's champion.

Hopefully this will kickstart it.

Here is a recap of last seasons top 10.

FPL HOF ranking, and finish this season [14/15].

1. Ville Rönkä - 277
2. Rasmus Sundman - 7640
3. Evs - 37429
4. Kenneth Tang - 10095
5. Mike Varcoe - 11138
6. Nick [Triggerlips] - 592
7. Ulrik [Nylund] - 5724
8. David Meechan - 3179
9. Paul Richardson - 45860
10. Chris Denbeigh - 30189

[matmutte]

Great stuff.
Just having a look at the top 5 of HOF + spidermatt, they all but one played their Wc in gw2/3 and are currently over the 1million overall rank.....

[hancockjr]

Super - I've managed to get in the millions without using my wildcard!

[RidleyMTB]

Looking at all of their current points, I'm feeling a little less down on my season so far...

[Stemania]

Anyone interested in the progress of the above list of names may also be interested in the FPL Stalker thread, where a few of the moves of a few of them are being regularly analysed.

[Sid22022]

Guys any of the top 10 HOF players who are on Twitter or run their own personal blog barring Matt? Thanks a million

[Blaze]

The man, the myth, the legend - Mr Ville Rönkä, the best person to ever play FPL.

Overall rank: 161,904
Team value: £104.0m
Wildcard: Gameweek 3
Number of hits: 0

Meanwhile most of us 'good players' have only just, or are yet to break the 1 million rank.

[tsubasa_sama]

He's made it back to the top of his local ML :p

[Finisher1]

I'm glad to see that Ville is also a big fan of maximising team value.

[Monsignor]

The man, the myth, the legend - Mr Ville Rönkä, the best person to ever play FPL.

Overall rank: 161,904
Team value: £104.0m
Wildcard: Gameweek 3
Number of hits: 0

Meanwhile most of us 'good players' have only just, or are yet to break the 1 million rank.

You forgot to mention

Captain: Aguero

[tsubasa_sama]

A popular thread last year, and I'm interested to see how the top 10 in the HoF fare this season.

Top ten HoF direct links

1. Ville Rönkä
2. Triggerlips
3. Jay Egersdorff
4. Martin Button
5. Derek Simpson
6. Craig Johnson
7. Alianto Henry
8. Simon Walsh
9. Martin Gupta
10. Jon Sumner

Notable others

15. Matthew Martyniak - Previous winner and FISO regular.
17. Julian Hancock - FISO regular with excellent record.
22. David Meechan - Dropped out of top 10 despite top 10k finish.
24. Rasmus Sundman - Dropped out of top 10 despite top 10k finish.
32. Ulrik Nylund - Dropped out of top 10 despite top 10k finish.
41. Simon March - Last season's champion.

Hopefully this will kickstart it.

Here is a recap of last seasons top 10.

FPL HOF ranking, and finish this season [14/15].

1. Ville Rönkä - 277
2. Rasmus Sundman - 7640
3. Evs - 37429
4. Kenneth Tang - 10095
5. Mike Varcoe - 11138
6. Nick [Triggerlips] - 592
7. Ulrik [Nylund] - 5724
8. David Meechan - 3179
9. Paul Richardson - 45860
10. Chris Denbeigh - 30189

Thought I'd give this thread a little bump.

All in this list bar David Meechan and, you guessed it Ville Rönkä have failed to crack the top 100k as of this stage of the season (Jay fell out of the top 100k for the first time since week 8 last week). David lies at ~22k in the OR with a season peak of 8k last week. Ville peaked in week 8 with a ranking of ~33k, he now lies in position ~57k.

Both have, as expected yet to use any of the big chips. Ville used the AOA in GW12, with both having played their initial wildcard.

It seems even the "experts" are having a tough old season this year!