Goalkeeper rotation - a case study.

A Fantasy Football forum for news on fantasy football games run by the Premierleague (FPL).
Post Reply
User avatar
jabbamunter
Wideboy
Posts: 60
Joined: 02 Oct 2006, 10:34

Goalkeeper rotation - a case study.

Post by jabbamunter »

I turn up here sometimes, don’t post a lot. Thought it was about time I offered something vaguely useful/interesting…

I’ve always felt rotating goalkeepers was an intuitively obvious thing to do. I dipped into a thread on here at the beginning of the season and it seemed the majority on FISO had a contrary view. I’m well versed in stats but I don’t tend to bother at a detailed level for FPL and rely on gut feel. Someone challenged me at the beginning of the season to test it and I’ve looked back at my performance over the year to try to assess whether there is any substance in my view.

Simplistically, at the start of the season I have 100m to spend and I want 2400 points, the additional 0.5m I spend on a rotating pair therefore needs to yield me at least 12 points over the season to be worth it. I pick the two 4.5m keepers who I think will score the most (but in reality probably some consideration to those that rotate best over a shorter period). My expected return comes down vs picking the first choice alone (assuming I’m right with that pick) but I take advantage of my skill to exploit fixture strength to bring it back up and beyond. I am also protected against injury, being dropped, blank weeks and have a bench boost keeper. I view any non-chip goal keeper transfer as a -4 in terms of opportunity cost and I would expect 1 over the season. Really this reduces the points that I need to make up through skill to 8 (N.B. I lost both Heaton and Leno to injury this season; N.B.2: the top scoring keeper, Pope, averaged ~4.5 points per game over the season). And, I have two shots at identifying the best keeper correctly.

I had 2 playing keepers throughout the season and scored 187 keeper points. I made no goalkeeper transfers outside chips despite 2 injuries. That is 17 points above Pope (who I wouldn’t necessarily have picked at the beginning of the season anyway, albeit I did have him for most of it). I did exceed the base price at times but, well done me, a good result. But does an isolated case really prove anything?

I can look at my own world of outcomes. Out of the 38 weeks, there were 8 where my keepers scored equal points. Let’s ignore them. Of the remaining 30, I picked the keeper who yielded the highest return 18 times. So 60% of my decisions were correct. This seems reasonable enough, I would think I could replicate that on an average year. I’ve not got anything to base that on other than I should be better than random at picking the highest scorer but given it’s not that easy, not dramatically so.

What if we replay the decisions that mattered based on that assumption? 30 correct decisions would have yielded me 215 points and 30 incorrect, 102. To work out the bits in between I simulated 500,000 alternative paths through the 38 weeks from my goalkeeper pairs assuming the correct decision was made 60% of the time to see how my actual picks compared. Turns out 187 was a good score. It’s the 93rd percentile. The mean is 169. So I was a lot better than expected at picking the result correctly when the gap between scores was wide. This makes sense as I should be better than random at predicting which keeper will get a clean sheet and there is also a correlation between clean sheets and bonus points. I’m not really sure where I should expect to sit on that distribution though. I think I should comfortably outperform the mean but I may have got a bit lucky to get out to the 93rd percentile. For reference the 75th percentile was 178.

What if we reduce the probability to random (50%)? The mean falls to 158. So my ‘skill’ at getting the decision the right way added 11 points to the mean. My skill at getting the decision the right way when it mattered added a further 18 points. I suppose it all hinges on the quality of the decision making. At 55% correct, the mean score would fall to 164, and at 65% increase to 175 but I think I’ve convinced myself there is sufficient upside to stick with the rotation strategy if the pairs look right.
Last edited by jabbamunter on 29 Jul 2020, 13:43, edited 1 time in total.

User avatar
Sutter Kane
Dumbledore
Posts: 6227
Joined: 05 Aug 2010, 12:13

Re: Goalkeeper rotation a - case study.

Post by Sutter Kane »

interesting post but lots of things difficult to measure accurately.

Firstly I believe that there will be some statisticians that would argue that you don't have 100m to spend, it's 36mn that is being judge upon above base price but that's another matter I'm not really able to discuss further. It would however mean needing more like 30+ points additionally required.

There is also the fact that the 0.5mn is in your squad for the best part of 35 weeks to be used as needed. This is pretty big imo. Add to that, we can't guarantee by a long shot that there will be two 4.5mn half decent keeper options, in any season.

Also I think, whilst you are correct in your decision making is above average, that you were lucky to get the total you did as you mentioned. So many times, GK scores are random - their variance is so low.

The ceiling seems higher with two playing keepers to rotate but there's been too many examples of people, including myself, picking the wrong ones over and over again when it seems obvious. It's not obvious and so it's not a strategy for me, barring a 4-4.1mn keeper who plays.

User avatar
jabbamunter
Wideboy
Posts: 60
Joined: 02 Oct 2006, 10:34

Re: Goalkeeper rotation a - case study.

Post by jabbamunter »

Thanks for your thoughts. The yield won’t be uniform across the 36m and the 64m but if you want to judge purely on the 36m you’d surely need to judge it against the expected points attributable to the 36m only.

User avatar
math!
Dumbledore
Posts: 5898
Joined: 01 Aug 2015, 03:30
FS Record: Yes

Re: Goalkeeper rotation - a case study.

Post by math! »

I used to rotate two goalkeepers but after getting burned badly for a few weeks I switched to a Pope type and a 4.0m or in some circumstances I would get both keepers for one club which seemed to be fairly common here. If someone like Martinez turns up then he's sure to be the second keeper.

User avatar
Mav3rick
FISO Knight
Posts: 17828
Joined: 20 Jul 2009, 20:35
FS Record: FPL Best: 1082, 1201, 1800, 10203.

The stats are dark and full of errors.

Re: Goalkeeper rotation - a case study.

Post by Mav3rick »

It always rather depends on keeper pricing, but I definitely sit more on the side of not using rotating keepers.

This is from a personal subjective experience of keeper rotation just not working for me, and also of seeing various analysis about the cheap keepers from the somewhat worse teams, generally not getting a benefit from strength of schedule.

The issue has always been that the big keeper performances tend to come in games where you'd rotate your keeper out, while at the same time, the chance of a clean sheet doesn't increase to a significant level for 'favourable' fixtures, while save point likelihood reduces.

In the season just gone, a set and forget Pope was about an obvious keeper choice as is possible, and while I agree that if you do have another keeper to rotate you might be able to bag more points from the keeper position, I think that investment of that extra 0.5 elsewhere is likely to yield more points (or save a couple of transfers) over a season.

In fact, last season was a bit of a gift in terms of 4.5 keepers overall with both Henderson and Pope available, and as ever pricing next season will be the question. In recent seasons, FPL have seemingly tried to remove the benefits of keeper rotation through the pricing system and when you get to the point where you're picking between two bottom 6 keepers to rotate or adding in a 5.0 keeper to the mix, then you start to clash with the premium keepers on a pricing basis or the lack of strength of schedule for the cheap keepers.

If a 4.0 keeper were available then rotation would happen certainly, while if someone like DDG was priced at 5.5 and you have to pay 5.0 for the likes of Pope, Fabs or Dubravka with no solid 4.5 keeper/pairing, then keeper rotation will continue to be unfashionable I think.

The problem with even a good looking pair of 4.5 keepers will be that if the pairing is good, then one of them on their own should be decent too, and the argument returns to the beginning with 0.5 invested elsewhere for likely higher gain.

User avatar
Mo Bot
Dumbledore
Posts: 5868
Joined: 15 Oct 2005, 13:59
FS Record: Inaugural Last Man Standing. FISO Fantasy Eurovision champ 2012. World #1 in UFPL for a whole week

Re: Goalkeeper rotation - a case study.

Post by Mo Bot »

I was a keeper rotator for many season and paid a lot of attention to fixture pairing to select my keepers as well. Picking the 'wrong' keeper became a habit and so I went for the set and forget option for most of the season.

I'll probably revert back to rotation next season if there is a decent, cheap pairing on offer. The reason being the possibility of increased postponements due to Covid.

Football Hero
Grumpy Old Man
Posts: 1214
Joined: 19 Aug 2014, 14:05

Re: Goalkeeper rotation - a case study.

Post by Football Hero »

I love threads like these, such meaty content and it is interesting to bring smart minds together to help solve these kind of problems.

I will add my input to this. I don't like this idea of playing the right keepers and getting the decision 55% correct or 65% correct. Surely (if we hope to be great at FPL) we should play the right keeper 100% of the time in an EV sense, and if the sum total of EV exceeds the 30 points that the 0.5M gives you then it is right to rotate keepers. We should be outcome-independent when we analyse this I believe.

I am a little unsure of this 30 points per 0.5M however, as we carry money in the bank a lot so this concept is a bit murkey to me.

Football Hero
Grumpy Old Man
Posts: 1214
Joined: 19 Aug 2014, 14:05

Re: Goalkeeper rotation a - case study.

Post by Football Hero »

jabbamunter wrote:
29 Jul 2020, 13:15
Thanks for your thoughts. The yield won’t be uniform across the 36m and the 64m but if you want to judge purely on the 36m you’d surely need to judge it against the expected points attributable to the 36m only.
I guess Sutter's point is that we could pick a team full of rotters for 64M that would score us 0 points (or close to it) over the season, (so this is our 'base' case or worst case and we look to improve from there), and so we use the extra 36M in the budget to buy players that will actually allow us to score more than 0 points.

User avatar
blahblah
FISO Baron
Posts: 78885
Joined: 13 Oct 2005, 18:46
Location: .. he thinks that he knows something which he doesn't, whereas I am quite concious of my ignorance.
Contact:

Re: Goalkeeper rotation a - case study.

Post by blahblah »

Football Hero wrote:
29 Jul 2020, 20:31
jabbamunter wrote:
29 Jul 2020, 13:15
Thanks for your thoughts. The yield won’t be uniform across the 36m and the 64m but if you want to judge purely on the 36m you’d surely need to judge it against the expected points attributable to the 36m only.
I guess Sutter's point is that we could pick a team full of rotters for 64M that would score us 0 points (or close to it) over the season, (so this is our 'base' case or worst case and we look to improve from there), and so we use the extra 36M in the budget to buy players that will actually allow us to score more than 0 points.
His point (not that I may or may not agree with it) is that you have to consider Base Prices for each position eg GK is 4.5m....

User avatar
blahblah
FISO Baron
Posts: 78885
Joined: 13 Oct 2005, 18:46
Location: .. he thinks that he knows something which he doesn't, whereas I am quite concious of my ignorance.
Contact:

Re: Goalkeeper rotation - a case study.

Post by blahblah »

So you have x amount for Base Price in each position and then consider the Marginal stuff above that......

Football Hero
Grumpy Old Man
Posts: 1214
Joined: 19 Aug 2014, 14:05

Re: Goalkeeper rotation - a case study.

Post by Football Hero »

blahblah wrote:
29 Jul 2020, 20:36
So you have x amount for Base Price in each position and then consider the Marginal stuff above that......
Yes, and that is what I am saying too. You have to spend at least 64M, so really it's only the 36M that is flexible in terms of where you allocate it. That puts a much higher premium on each additional 0.5M. However the cash in the bank muddies the waters hugely, as outside the early weeks of the season, often you will have around 1.0M in the bank most of the time.

User avatar
jabbamunter
Wideboy
Posts: 60
Joined: 02 Oct 2006, 10:34

Re: Goalkeeper rotation a - case study.

Post by jabbamunter »

Football Hero wrote:
29 Jul 2020, 20:31
jabbamunter wrote:
29 Jul 2020, 13:15
Thanks for your thoughts. The yield won’t be uniform across the 36m and the 64m but if you want to judge purely on the 36m you’d surely need to judge it against the expected points attributable to the 36m only.
I guess Sutter's point is that we could pick a team full of rotters for 64M that would score us 0 points (or close to it) over the season, (so this is our 'base' case or worst case and we look to improve from there), and so we use the extra 36M in the budget to buy players that will actually allow us to score more than 0 points.
Ok, I see what you’re saying. But if it is priced correctly returns will diminish so how can you judge each marginal 0.5m against 30 points? Adding a 4.5m playing keeper to 2 non-playing 4m goalkeepers to is worth >100 points. Going from 8.5 to 9 just has to be better than adding the final 0.5m increment to one of the other positional budgets.

Everyone has anecdotal evidence about picking the wrong keeper. It happened to me 12/30 times this season. It niggles and you remember it more than when you make the correct pick but has anyone actually gone back and checked the frequency they got it right over the season?

User avatar
blahblah
FISO Baron
Posts: 78885
Joined: 13 Oct 2005, 18:46
Location: .. he thinks that he knows something which he doesn't, whereas I am quite concious of my ignorance.
Contact:

Re: Goalkeeper rotation - a case study.

Post by blahblah »

It is really difficult to be on the right side of GK rotation and belies simple ideas like play the one at home....

User avatar
Mav3rick
FISO Knight
Posts: 17828
Joined: 20 Jul 2009, 20:35
FS Record: FPL Best: 1082, 1201, 1800, 10203.

The stats are dark and full of errors.

Re: Goalkeeper rotation a - case study.

Post by Mav3rick »

jabbamunter wrote:
29 Jul 2020, 21:42
Everyone has anecdotal evidence about picking the wrong keeper. It happened to me 12/30 times this season. It niggles and you remember it more than when you make the correct pick but has anyone actually gone back and checked the frequency they got it right over the season?
It's really dependent on the quality of teams being rotated, so will be seasonal to a point.

Last season, rotation of Pope and Henderson is a very different prospect to rotation of two bottom 6 sides, since when you rotate you're looking for clean sheet predictably, which is far less apparent in bottom 6 sides than upper mid table sides.

Could I ask which two keepers you had your success with this year?

User avatar
jabbamunter
Wideboy
Posts: 60
Joined: 02 Oct 2006, 10:34

Re: Goalkeeper rotation a - case study.

Post by jabbamunter »

Mav3rick wrote:
29 Jul 2020, 21:53
jabbamunter wrote:
29 Jul 2020, 21:42
Everyone has anecdotal evidence about picking the wrong keeper. It happened to me 12/30 times this season. It niggles and you remember it more than when you make the correct pick but has anyone actually gone back and checked the frequency they got it right over the season?
It's really dependent on the quality of teams being rotated, so will be seasonal to a point.

Last season, rotation of Pope and Henderson is a very different prospect to rotation of two bottom 6 sides, since when you rotate you're looking for clean sheet predictably, which is far less apparent in bottom 6 sides than upper mid table sides.

Could I ask which two keepers you had your success with this year?
Totally. Although at the point you would have picked your season long keeper, remember the information to hand was that Burnley finished bottom 6 in 18/19 and Sheffield Utd were newly promoted behind Norwich.

1-14: Pope - Heaton
15-29: Pope - Henderson
30: 38: Martinez - Patricio

User avatar
Mav3rick
FISO Knight
Posts: 17828
Joined: 20 Jul 2009, 20:35
FS Record: FPL Best: 1082, 1201, 1800, 10203.

The stats are dark and full of errors.

Re: Goalkeeper rotation - a case study.

Post by Mav3rick »

That is true, although Burnley were suffering from the EL that year, and I think it's fair to say most had identified Pope's price as a gift from the FPL gods.

Pope & Heaton was a typical (and sensible) starting pair if keeper rotation was your thing, how did your rotation work out in those first 14 weeks?

There's a lot of nebulous opportunity cost involved of course, everything from an extra 0.5 on player x over player y for 14 GWs (rather than Heaton over Pope for 7) to the benefit of using that 0.5 to cover price rises in the early weeks when price volatility can easily catch you out.

It almost entirely comes down to seasonal squad circumstances as to where that 0.5 best fits.

User avatar
cymbalrush84
FISOhead
Posts: 806
Joined: 05 Feb 2008, 21:42
Location: Wolverhampton
FS Record: 2nd in Pendulum Swingers Super League 2006/7
Contact:

Re: Goalkeeper rotation - a case study.

Post by cymbalrush84 »

What I've learnt from having two 4.5 keepers is that it's really handy to have this if a first choice keeper at a top 6 club is out of action for the duration and deputy is a steal if the defence in front are pretty stubborn still. Gazzaniga struck gold for me whenever I selected him, especially GW23.

Sent from my VTR-L09 using Tapatalk


User avatar
blahblah
FISO Baron
Posts: 78885
Joined: 13 Oct 2005, 18:46
Location: .. he thinks that he knows something which he doesn't, whereas I am quite concious of my ignorance.
Contact:

Re: Goalkeeper rotation - a case study.

Post by blahblah »

cymbalrush84 wrote:
31 Jul 2020, 10:54
What I've learnt from having two 4.5 keepers is that it's really handy to have this if a first choice keeper at a top 6 club is out of action for the duration and deputy is a steal if the defence in front are pretty stubborn still. Gazzaniga struck gold for me whenever I selected him, especially GW23.

Sent from my VTR-L09 using Tapatalk
Indeed, and I guess Martine would have rocked. Did you try Alison's No2?

User avatar
cymbalrush84
FISOhead
Posts: 806
Joined: 05 Feb 2008, 21:42
Location: Wolverhampton
FS Record: 2nd in Pendulum Swingers Super League 2006/7
Contact:

Re: Goalkeeper rotation - a case study.

Post by cymbalrush84 »

blahblah wrote:
cymbalrush84 wrote:
31 Jul 2020, 10:54
What I've learnt from having two 4.5 keepers is that it's really handy to have this if a first choice keeper at a top 6 club is out of action for the duration and deputy is a steal if the defence in front are pretty stubborn still. Gazzaniga struck gold for me whenever I selected him, especially GW23.

Sent from my VTR-L09 using Tapatalk
Indeed, and I guess Martine would have rocked. Did you try Alison's No2?
Started with Alisson, so yes. But lost my religion on the clean sheet front, even when he returned.

Sent from my VTR-L09 using Tapatalk


User avatar
blahblah
FISO Baron
Posts: 78885
Joined: 13 Oct 2005, 18:46
Location: .. he thinks that he knows something which he doesn't, whereas I am quite concious of my ignorance.
Contact:

Re: Goalkeeper rotation - a case study.

Post by blahblah »

cymbalrush84 wrote:
31 Jul 2020, 18:08
blahblah wrote:
cymbalrush84 wrote:
31 Jul 2020, 10:54
What I've learnt from having two 4.5 keepers is that it's really handy to have this if a first choice keeper at a top 6 club is out of action for the duration and deputy is a steal if the defence in front are pretty stubborn still. Gazzaniga struck gold for me whenever I selected him, especially GW23.

Sent from my VTR-L09 using Tapatalk
Indeed, and I guess Martine would have rocked. Did you try Alison's No2?
Started with Alisson, so yes. But lost my religion on the clean sheet front, even when he returned.

Sent from my VTR-L09 using Tapatalk
Alison was\is too expensive for this idea, imho.

The GK point, imh(umble)o is to get the most from 9 or 9.5m, and Pecker was well behind his defenders?

jane477
Newbie
Posts: 1
Joined: 20 Jul 2020, 22:12

Re: Goalkeeper rotation - a case study.

Post by jane477 »

cymbalrush84 wrote:
31 Jul 2020, 10:54
What I've learnt from having two 4.5 keepers is that it's really handy to have this if a first choice keeper at a top 6 club is out of action for the duration and deputy is a steal if the defence in front are pretty stubborn still https://admission-writer.com/blog/georg ... ay-prompts. Gazzaniga struck gold for me whenever I selected him, especially GW23.

Sent from my VTR-L09 using Tapatalk
Yeah, this was one in 10000 chance
Last edited by jane477 on 06 Aug 2020, 21:44, edited 1 time in total.

Football Hero
Grumpy Old Man
Posts: 1214
Joined: 19 Aug 2014, 14:05

Re: Goalkeeper rotation - a case study.

Post by Football Hero »

blahblah wrote:
31 Jul 2020, 18:10

Alison was\is too expensive for this idea, imho.

The GK point, imh(umble)o is to get the most from 9 or 9.5m, and Pecker was well behind his defenders?

You're putting weird question marks in again when it appears that you are trying to make a statement.

What were you actually trying to say, either:

(1) Was Pecker well behind his defenders?

OR

(2) Pecker was well behind his defenders.

User avatar
blahblah
FISO Baron
Posts: 78885
Joined: 13 Oct 2005, 18:46
Location: .. he thinks that he knows something which he doesn't, whereas I am quite concious of my ignorance.
Contact:

Re: Goalkeeper rotation - a case study.

Post by blahblah »

I was asking the question of whether he got as many points as the Defenders, even Points per £1m?

Isn't the purpose to get more points with cheaper players?

Football Hero
Grumpy Old Man
Posts: 1214
Joined: 19 Aug 2014, 14:05

Re: Goalkeeper rotation - a case study.

Post by Football Hero »

blahblah wrote:
01 Aug 2020, 16:46
I was asking the question of whether he got as many points as the Defenders, even Points per £1m?

Isn't the purpose to get more points with cheaper players?
He was worse value and he got less points so he was not a great pick.

The purpose is to get more points with all your players, both the cheaper ones and the expensive ones.

View Latest: 1 Day View Your posts
Post Reply

Return to “Fantasy PremierLeague.com (FPL)”