Best way to rank FPL records. (split from Billy's consistent managers thread.)

[Carlos Kickaball]

CK, I hope you appreciate the irony of you spending so much time nitpicking over which of two ever so slightly different systems (of measuring who the best managers are) is slightly more precise after spending most of the spring trying to tell us that the records of managers for whom either system would serve to differentiate between (say Ville and a regular forum 5k averager) are only really separated by luck. :p

Somewhat ironically I actually think if you look at the point scores and ranks of two players, and can't really tell which one is better easily, then there is probably not enough difference in their records to meaningfully conclude who is better anyway.

Ha ha. So you now ''can't really tell which one is better easily'' of a record averaging 1k and one averaging 5k despite spending the last page saying that raw scores are better than averages because they take into account how it's much harder to climb than it is to fall because the distance from 1k to 5k is actually very big, as big as 10k to 27k in this case.

Well someone with a 5k average ranking could easily have a comparable amount of points to someone with a higher average ranking, if they had a few very high ranked finishes coupled with some 9-10k finishes or even with just one 25k finish.

[Droughton]

I've always thought that people read too much, or too far into past results overall rank wise. The rules of the game have changed since 4,5,6,7 years ago. We now have vice captains, roll over subs, different price rise/fall levels that CTC struggled to pin down this year.

I often find that people had very impressive ranks 4,5 even 6+ years ago but have failed to hit those heights again. Maybe because the game rules changed and that benefited there competitors. Maybe because its more competitive now with various sites offering advice and critique....... or maybe because they are less interested/have less time for a fantasy football game.

It would be interesting to see how people rank up against each other pre/post vice-captain era or pre/post sub roll over era.

I always think you are only as good as your recent performances. My reasoning for this is that surely over time you acquire more knowledge, intelligence and skill in whatever it is you are doing, in this case FPL. I usually look at the last 2-3 years in somebodies history and make a judgement. I may look behind that and consider them to have 'potential' to improve on recent scores, but that wouldn't change my opinion of there current performance levels.


I have had one good finish in FPL. The amount of time I put into that season was huge. I had a RTM thread I posted in almost every day and definitely every week. I was doing write-ups in the beehive and trawling forums, watching games, highlights... reading endless threads, posts. Analysing form and upcoming fixtures 4-5 or more weeks in advance, immaculately planning transfers. I managed to finish 292 and was sitting in the top 200 until the final gameweek. So I know how much time/commitment it can take to break that top 1k barrier. It is very competitive to get to those heights and to manage it through a season and then to be there or there about (5k>) 2-3 years running. If you can do this it means you are one of the best. YES - You could get there without so much time input but it would be very difficult as a lot of things would have to be going your way, one thing wrong such as a injured player on the friday/saturday morning without you noticing or then you fail to login in time to make your ideal transfer.... your on a slippery slope down the overall rankings.

In summary

If somebody doesn't have a top 1k finish in the last 3 years then they aren't in the group of the best managers. This then has to be couple with another good finish say top 5k and then another respectable say top 15kish.

5,200
300
22,000

Is better than

5,900
4,400
6,000

In my opinion.

[Oatfedgoat]

.........
I have had one good finish in FPL. The amount of time I put into that season was huge. I had a RTM thread I posted in almost every day and definitely every week. I was doing write-ups in the beehive and trawling forums, watching games, highlights... reading endless threads, posts. Analysing form and upcoming fixtures 4-5 or more weeks in advance, immaculately planning transfers.
.......

I can relate to that!
The season I finished in the 500's I put so much time into that season.
The following season I slumped to 100k+ despite putting in even MORE time than before. Much of this time was beehive write ups.
What I found that season, and many since is that for all the good forums and research does, it is all too easy to fall into the trap of listening to the hive mind of FISO and assuming because everyone is saying something, it must be the right thing to do.

The last 2 years I have spent far less time on FISO and my results have massively improved with two top 3k finishes.

Things that spring to mind are, bringing in Aguero and Silva before they proved themselves after injury when the consensus was to wait until their fitness was proven.
Yes it was risky, but each time the potential reward of (c) them coupled with a great run of fixtures each time really paid dividends.

Also, Austin(c) for their DGW was not the common opinion people were speaking about. Bucking that trend also helped me offset the many many visits to the triple millionaires club I had this season.

My results are so sporadic that I genuinely don't know where I fall as an FF player.
I'm no where near being amongst the best we have here on FISO, but amongst most of my ML I rarely fall out of the top 3.

[Stemania]

What I found that season, and many since is that for all the good forums and research does, it is all too easy to fall into the trap of listening to the hive mind of FISO and assuming because everyone is saying something, it must be the right thing to do.

The last 2 years I have spent far less time on FISO and my results have massively improved with two top 3k finishes.

I'm a bit surprised by your general outlook as the effect of sites like FFS and FISO on most people's rank is generally accepted to be extremely rank-boosting as far as I thought. There's definitely countless examples of managers whose rank has jumped immediately from the 50-100k average region to 10k-20k average or even higher as soon as they've found sites of this type.

What I would say is that 'consensus' on FISO' is strictly speaking very very rare in my experience - with captaincy polls for example it's not common for one player to get more than 50-60% of the vote. That leaves a huge proportion of the community that don't agree every week. There is almost never an actual consensus (90% or whatever), and I think the term is generally used around here to just signify the most popular of (possibly many viable) options if it's noticeably ahead of option 2.

There's always going to be a danger of the group mentality or slightly biased views surrounding the more divisive EPL players, sure, but the way to solve that and find the right answers is for those disagreeing with apparent consensus to speak more often and push the reasons they feel the consensus might not be right on a certain occasion. If the reasons then given don't hold water then maybe the consensus was right after all, if they are very good reasons they can be discussed. So if you are finding yourself disagreeing with the consensus often then I urge you to speak out more often rather than step away from the forum as posters who do disagree with consensus (like you seem to) are some of the most valuable to the forum.


You see a lot of people complaining in the captaincy polls every week (I'm not speaking about you now OFG) saying they followed the polls on the popular websites instead of going with their gut and because of it they've lost out and that listening to FISO has damaged their rank. Well, who's to blame there? I don't think you can just skim sites like FISO/FFS to find what the 'best choices' are (and then follow these moves because of their popularity) and then have any case to complain if it goes wrong. If you don't look fairly closely at the reasons people are actually giving for making or not making a decision and then deciding by yourself which of the reasons are stronger then you have absolutely no right to complain, you're just being a sheep who blindly follows the herd and obviously the herd is not always (or even possibly most of the time) going to be 'right'. If someone was just going with their gut before being convinced by a poll then they should just admit they were guessing in the first place, if they had good reasons to go against the most popular option but chose to ignore them over strength in numbers then they are again to blame. More importantly I'd urge them to air these reasons beforehand and challenge the consensus before the event.

The uncomfortably fact is though, every one of us is guessing to some extent all of the time, and that the 'right' decision is not the same thing as 'the decision which would with hindsight have scored the most points' - in fact they are pretty unrelated. I post a lot around here (probably too much), but to be honest it's not because I think I know a lot - it's because I find it a very good sounding board for ideas and a way of solidifying my own thoughts, and that's where I think FISO boosts my ranking (along with the fact that there are better players than me here who I can learn from, and the collected reservoirs of knowledge/team nws etc).

[Mav3rick]

I post a lot around here (probably too much), but to be honest it's not because I think I know a lot - it's because I find it a very good sounding board for ideas and a way of solidifying my own thoughts, and that's where I think FISO boosts my ranking (along with the fact that there are better players than me here who I can learn from, and the collected reservoirs of knowledge/team nws etc).

I agree with that. Sure, I could go round all the team news website collating injury news or I can use FISO to do that for me and then use the free time I have to work out rotation strategies, or to think about squad structure and share it with the community, while others looks at fixture lists, pre season lineups, potential DGWs or starting cheap defenders, etc.

Sure some lurkers will just take everything and contribute nothing, but the working together aspect lets me enjoy the game a bit more and has surely enabled me to boost my own rankings over the last three seasons.

That doesn't really have anything to do with the original subject matter, but I'm not overly worried about the best way to decide which of the regular top managers are better since I don't think it matters if x is marginally better than y.

If pushed, I actually think Billy Whiz's original algorithm does a good job as when you look at the names that are up there I'd find it hard to argue that any of them are anything less than exceptional managers who also make valued contributions to the forum. I like putting them under the same banner (the "elite") rather than trying to create an explicit ranking since I think that's better from a community aspect too (promoting information sharing and inclusivity over competition and secret moves).

[MoSe]

I'm sorry, I wanted to address it properly and orderly, but I let a tiny gap open addressing a wider subject, and CK jumped in on the detail forcing me the hand....

viewtopic.php?p=2712483#p2712483

• 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,

  you say yourself you play for a rank..... this means the "rank-bias" not here but in the WORLD it's because if you play for a rank, the aim of the game is to be the best, period, i.e. to get the best rank by the mean of getting the most points, or whichever other metric the rules state it has to be USED TO DETERMINE THE RANK
  

  why if you are deciding who the best manager is over a period of seasons you'd decide to chop up the history into piecesand take ranks, I'm not sure.

  what made you think I "decided to chop" ????
  I did NOT decide and did NOT chop anything:
  seasons are BORN CHOPPED without my decision.
  History is made by putting together atomic seasons. Whereas a Season is not made by putting together months, it's the opposite, the race is born as a season, and to add something to it you focus on intermediate steps.
  PL or FPL do NOT declare a competition over 5 years, and then look at single season as intermediate steps of their quinquennial plans, awarding season titles as mere byproducts...

please consider this is just a draft, and that when I'll have the time I'll address the whole discussion in a tidier and systematical way, starting it over

[Stemania]

If there is a rank vs points argument craved, this is defo the best place for it imo. Thanks, MoSe.

[Carlos Kickaball]

And no thanks for originally creating this thread.

To my mind the only reason to use ranks is if those yearly ranks are specifically what you're interested in as acheivements. This is perfectly logical preference.

In terms of measuring performance points is the best measure, and seemlingly most of the attacks on using points and ranking them over a longer period (using averages or totals) are mostly to do with a misinderstanding of the consequences. By taking ranks at the end of each year, you're actually reducing the information in the data (in order to give context to single entries).

I find it somewhat amusing that using a log score for ranks, is only really an attempt to get them to mirror points scores more closely.

[Stemania]

And no thanks for originally creating this thread.

Going back to this, I refer you to my post quoted in the OP.



Whether you sort a HoF or ranking table by some calculation based on ranks or based on points, the difference in the order will broadly speaking be pretty small. The top managers in the current HoF will still be around the top of the other method's. The people around 1000th will be not too far from 1000th - they certainly won't be suddenly near the top.

It seems to me, CK, that for you to be able to argue with any credibility at all statements like:

In terms of measuring performance points is the best measure, and seemlingly most of the attacks on using points and ranking them over a longer period (using averages or totals) are mostly to do with a misinderstanding of the consequences. By taking ranks at the end of each year, you're actually reducing the information in the data (in order to give context to single entries).

....then you would first have to do one of the following:

1) Concede that your repeated statements all spring that it's essentially luck that separates managers like Ville from the bulk of 100 or 1000 or 3000 (or however many) very good managers that exist were wrong; or,

2) Concede that the differences in the two methods (one of which you keep pushing is much stronger than the other) is pretty meaningless in the context of your denial of point 1).

So which is it?

[Carlos Kickaball]

Neither. Your logic and possibly appreciation of my stance incorrect.

[Stemania]

So what would we learn from a manager ability table calculated by total points over an ability table based on rank in view of 1)? Both will give a fairly subjective list of managers, the first however many (100, 1000, 3000) of which are only separated by luck? Great. What have we learnt?

[Carlos Kickaball]

Doesn't matter though, we're discussing the accuracy of the measurement here, in fact further we're discussing what is the ideal system for making the measurements. You're also misrepresting my views on another topic by taking them in their nascent form and simplistic terms (but that's for another thread. ).

PS you can probably PM me about this kind of thing, because I don't think we're adding to the debate anymore.

[Stemania]

Doesn't matter though, we're discussing the accuracy of the measurement here, in fact further we're discussing what is the ideal system for making the measurements.

What do you mean ideal, or accurate? It just doesn't matter - there's too much noise. The current FFS HoF puts me 191st. Another system might put me 240th. The 191st best fpl manager doesn't exist - it's a made up concept. Another system might put me 160th. It's not an exact science, every vaguely sensible system is going to give a fairly similar result. No rating system is going to put me up with the likes of Ville/trigger etc, and if it does it's extremely contrived. You can't say this guy's better than this guy because he's 191st and the other guy is 192nd on my arbitrary list. Which is better? A record of 15000, 50000, 1300, 4000, 10000, 5400, or a record of 490, 22000, 12000, 600, 3000, 42000? There is no ''accurate'' or ''ideal'' answer. It doesn't exist, it's completely subjective. (And 'skill' level whatever that is, is certainly not a fixed quantity year on year anyway but dealing with that is a whole other debating issue. )

Ville for example will likely be in the top 5 or so (at absolute worst) whichever system you use. The difference between Ville's results and my results is not arbitrary - i.e, the difference between the top few managers in the FFS HoF and those around 200th odd is noticeable. Ville's record is 277, 90, 4668, 1631, 362, 569. It's clearly much better than mine. Whichever sensible system you use (points/ranks possibly with a bit of weighting, whatever) has done it's job by putting me in a different category.

The advantage of a ranks based exposition is that it has meaning. If you say to someone Ville has an average finishing position of 1k they know what you mean (this stays true even if you bend the truth and normalise the previous ranks by the number of competitors) - they know immediately how impressive it is. If you say 'mean rank' and use some complicated average or choose to include some logarythmic scaling then people will still know what is meant. If you say Ville has had four top 1k finishes in the last six years they will know what at that means. If you say Ville averages 2340 points, or Ville averages finishing about 200 points behind the winner they will in most cases look at you blankly.


Further, you say you want to rank people based on points as it explicitly gets round this logarythmic rank-points relationship. Well, how are you going to list your final table, because guess what? That ordered list is going to have the very same logarythmic property you were trying to avoid - the difference between 10th and 20th is much bigger than the difference between 50th and 60th in the HoF table. What then would it mean to be 191st in your table?


(Btw, I assume the 'Doesn't matter' and 'PM the conversation' is a sidestep as to not see your previous posts discredited publicly. In my eyes you can't at the same time argue that luck is too much of a factor to read anything meaningful from someone being 1st and someone being 191st in the HoF - which you cannot deny you were saying as you mentioned my name explicitly in this context a number of times - and then demand pin-point accuracy on the way such ranks are determined in the first place.)

[Carlos Kickaball]

Further, you say you want to rank people based on points as it explicitly gets round this logarythmic rank-poits relationship. Well, how are you going to rank your final table, because guess what? That's going to have the very same logarythmic property you were trying to avoid - the difference between 10th and 20th is much bigger than the difference between 50th and 60th in the HoF table. What then would it mean to be 191st in that table?

Yeah the final table having this relationship isn't the problem, it's components [season ranks] being summed or averaged together that makes the problem. The fact you have still not appreciated this shows as distinct lack of understanding on your part, and I have little interest in explaining the same things over and over again.

(Btw, I assume the 'Doesn't matter' and 'PM the conversation' is a sidestep as to not see your previous posts discredited publicly.

They're not, my suggestions have been taken on board by even those that disagree with me. You are trying to suggest that my posts on the topic are not valid at all, rather than refute the points they make.

[Stemania]

'Distinct lack of understanding'? Classy.

I'm not trying to suggest that your posts on the topic are not valid at all. I'm saying that they either refute points that you have previously repeatedly made or are by your own definition irrelevant. If it makes you feel better I think that the fact they are irrelevant is independent of your previous comments on luck, for the reasons I tried to outline above.

[Carlos Kickaball]

Yeah, and the reasons you think they conflict is that you have given simplistic accounts of my positions, and used poor logic to infer a contradiction (where there is in fact only a slight irony that I would have interest at all ).

Which is certainly not classy...

I could think all managers were of exactly equal ability (I don't), and still have a preference of how I would rank the best records.

[ChrisA]

I hope nobody minds me bumping this thread but I've just come across it and found it a very interesting read.

I thought I'd shed some light on how the FFS HoF operates (I'm the one who wrote and maintains it). Despite the fact the FFS HoF displays ranks for past seasons, it uses points and not rank when deciding how highly to rate someone's season.

Reading this thread was really interesting. Carlos K comes across as a bit of a statistical perfectionist (so I think we are quite like-minded! ) and I find myself agreeing with his rationale of why points are a better indication of performance than rank when compiling a HoF. Saying this, the FFS HoF is designed to be a measure of how good an FPL manager currently is and not an indication of 'consistency'. It is certainly interesting to look at consistency of FPL performers in terms of average rank, but it shouldn't be used to say, "this manager is better than this one". If I wanted to use a measure to do this, I would rather use points, and not rank, as the metric.

By the way, the live HoF at FFS has just been updated! Ville stills leads the way by some margin!

[Ruth_NZ]

Well, here's my two-pennorth.

Sites like FFS make it easier for managers to do reasonably well. By which I mean that top 100k should be quite easily achievable for someone willing to put the time in and play sensibly.

It is when we get inside the top 10k that things become more complex. There are some (many?) in there who play for a top 10k finish more than they play to win it. In other words they would rather make percentage plays and secure a safe finish around 5k than playing with more freedom and aiming for the top whilst risking slipping to 20k or 40k maybe.

These "safety" players tend to use % ownership as a guide (whether consciously or simply by following other managers like Ville or Triggerlips). Some of them pay great attention to things like captaincy polls and like to captain the most popular choice most weeks (or always).

Grading these two approaches against each other isn't easy because they aren't actually aiming for the same thing and there is no entitlement for anyone to arbitrarily decide which approach is more valid. You can have your personal preference (as I do) but that is all it is. So any attempt to comparatively grade managers is doomed to failure (or dispute) as far as I can see, whatever measure you use.

At least until someone wins FPL twice, that is. Then we'll all be able to agree who is the best manager.

[Le Red]

I hope nobody minds me bumping this thread but I've just come across it and found it a very interesting read.

I thought I'd shed some light on how the FFS HoF operates (I'm the one who wrote and maintains it). Despite the fact the FFS HoF displays ranks for past seasons, it uses points and not rank when deciding how highly to rate someone's season.

Reading this thread was really interesting. Carlos K comes across as a bit of a statistical perfectionist (so I think we are quite like-minded! ) and I find myself agreeing with his rationale of why points are a better indication of performance than rank when compiling a HoF. Saying this, the FFS HoF is designed to be a measure of how good an FPL manager currently is and not an indication of 'consistency'. It is certainly interesting to look at consistency of FPL performers in terms of average rank, but it shouldn't be used to say, "this manager is better than this one". If I wanted to use a measure to do this, I would rather use points, and not rank, as the metric.

By the way, the live HoF at FFS has just been updated! Ville stills leads the way by some margin!

Hello Chris! Coincidentally I was checking out the HoF parameters a few minutes ago and got to your post.
If I have anything to say, it's that I'd like to see two different ranks: a "historical" and a "current", for the reasons explained by Droughton in the second post.
I think the "historical" rank, which the HoF represents, is not a good indicator of who are the best FPL players at the moment. It would be good to have a rank considering only the last 4 seasons or so to measure that in my opinion.
Also, mind, the historical rank is a tremendous let down to players such as myself, who gave the game a break, albeit still participated in every season since we started playing. If you check my history out, you'll see I'm a rather decent player, but I had a couple of seasons where I barely played at all, leaving the team inactive for a good while too. This way, I have two ranks of 354334 and 553390 as perpetual stains in my history. Even with HoF's 70% system I'll have to play for another decade before these scores become *nearly* irrelevant.
If we had a rank that disconsidered long past seasons and gave more focus to the recent ones, reflecting more accurately who the best players are right now, I think it would be a great plus in the whole "who's best" debate. Of course, no such thing as a perfect system, but it would still be a welcome addition to the HoF. Do you have anything to say on that?