R_NZ FPL Blog

[Ruth_NZ]

TOP 12 DETAILS continued

7. GK Averages

The GK averages of the 12 (points per game) were: 5.0; 3.6; 5.5; 4.6; 5.6; 4.3; 4.1; 4.6; 4.1; 3.7; 4.3; 4.5; Average 4.48 ppg.

The players ranked 3rd and 5th got something near to the best score available from GKs with 210 and 212 points respectively in a season where Heaton was the highest-scoring single GK with 149 points. They both did this by having Heaton (4.5m) as their base GK for most of the season and then getting Grant in as 2nd GK (4.1m) when he became established at Stoke. By doing so they had almost perfect cover because Grant's scores when Heaton was injured (GW13, 14 & 27) were 7, 8, 7. They then had both De Gea and Cech for the DGWs at the end of the season. You couldn't reasonably expect to do much better.

It is perhaps worth noting that the difference between best and worst for GKs in the top 12 was worth 74 points. This is a significant number and it's interesting to see that it's quite close to the 87-point swing between best and worst in the top 12 for Captaincy.

8. Defender Averages

The Defender averages (points per game) of the 12 were: 4.2; 4.8; 4.2; 4.1; 3.9; 4.2; 4.3; 4.1; 4.2; 4.3; 4.6; 4.3; Average 4.26 ppg.

Unfortunately I don't have similar numbers for previous seasons but from recall the average here is significantly down on what it was even a couple of seasons back. It is also the first time in my experience where the average GK score at the top is bigger than the average defender score.

Why is this? Well, the fall in the number of clean sheets is one element. From 232 CSs in 2013/14 (30.5%) down to 214 in 16/17 (28%) and it has fallen each season over that period. This affects both GKs and defenders of course, but in the mean time the GKs have had their saves better rewarded in the BPS and are now more likely to get BPs. At the same time the number of penalties rose to a record 106 penalties in 2016/17 with a record-high number of penalty saves (17). Penalty saves are very well rewarded in FPL, both in terms of points and in the BPS. If a GK saves a penalty in achieving a CS they are almost guaranteed 14 points plus any save points.

It would probably be unwise to make any rash judgements on the basis of one season but if the number of CSs continues to fall then it is true that attacking points for defenders become relatively a bit more important than they were; in other words they are likely to be a slightly larger component in percentage terms of a defender's points. It may also be that investing more in defence to get defenders from the best CS teams who also have attack threat (meaning 6.0m+ usually) will be more necessary than was once the case. This crosses over into the "value" study and I shall take it up again there.

In terms of swings, it must be remembered that at least 3 defenders must be played every week. With that in mind (and bearing the BB in mind) I have assumed 120 defender appearances per team over the season. Based on that calculation the swing between the best and worst in the top 12 was 116 points, bigger than the captaincy swing. The low score of the 5th placed team looks like an outlier though, it is significantly below the rest. The difference between the best and the 11th best was a more reasonable 93 points.

The top scorer for defender averages (the 2nd placed team overall) got most of his defender decisions right. He started the season with Coleman, Walker and MacAuley (I thought he might come up somehow ) despite Coleman being injured when the season started. As a result of those successful selections he only needed to make 2 defender transfers in the whole season up until his 2nd wildcard in GW34. And those were to bring Alonso in fairly early and for most of his points (GW15) and then Mawson for most of his later season points (GW25). This was a manager that used 442 quite often, mainly in that mid-season period. So he can be commended for being willing to switch out of the more standard 343/352 when he saw the price/performance landscape warranted it.

Almost there now. The next post should complete this hopefully. Although I then have it in mind to do a similar run comparing the top 12 with the top 20k, we'll see.

[Ruth_NZ]

TOP 12 DETAILS continued

9. Midfielder Averages

The Midfielder averages of the 12 (points per game) were: 5.4; 5.5; 5.8; 5.7; 5.8; 5.4; 5.6; 5.5; 5.6; 5.4; 5.6; 5.7; Average 5.58 ppg.

The striking thing about the midfielder averages are how close they are compared to the other positions. There was only a gap of 0.46 ppg between the best and worst teams in the top 12 here and the eventual FPL winner actually achieved the lowest average from his midfielders of all the 12 teams.

The overall difference in points between best and worst for midfielders in the top 12 (based on 160 midfielder appearances) was 74 points. This is exactly the same as the swing evident in the GK position despite the fact that we are talking about 4 or 5 midfielders every week rather than 1 GK. The midfielder slots probably get more attention by most FPL managers (and on sites like FFS) than any other position and also have a lot of FTs spent on them; perhaps that is one contributory reason for this. But beyond that I'll just note the numbers rather than proposing even a tentative explanation.

10. Forward Averages

The Forward averages of the 12 (points per game) were: 6.4; 6.1; 5.6; 6.4; 6.1; 6.5; 6.3; 6.3; 6.5; 6.4; 6.1; 6.6; Average 6.26 ppg.

We are back to a wider range again here. Perhaps it is true that many key midfielders performed quite steadily and well over the season (Sanchez, Hazard, Alli, Eriksen, Coutinho) whereas the strikers were more in and out. Aguero had good and bad runs. So did Lukaku. Same Costa. Even Kane was erratic in the first half of the season. Perhaps timing the roll-on roll-off was harder and a bit more haphazard in general for strikers than for midfielders?

An interesting note is that the forward averages (ppg) don't seem to have been affected by whether a manager was playing mainly with 2 or 3 forwards. The overall average of 6.26 ppg for forwards was almost exactly the same for those 7 teams that mainly played 343 (6.25 ppg).

Assuming 100 forward appearances per team the overall gap between best and worst within the top 12 was 94 points. A similar number to the difference achieved for the defender slots. If we exclude the low outlier of the 3rd placed team the gap between the rest was 53 points.

11. General Observations

So, to summarise, the best-to-worst swings in the top 12 were as follows in the various categories:

• Captaincy: 87 points
• GK: 74 points
• Defenders: 116 points
• Midfielders: 74 points
• Forwards: 94 points

By comparison the best-to-worst swings were much lower for the chips; 18 points for TC and (if we exclude 2 crazy low outliers) 20 points for the BB. We focus a lot of attention on these chips and clearly it's good to get them right. But they don't look to be as important in the scheme of things as they might initially appear.

I have some thinking to do about these results and maybe the top 20k figures (if I do them) will shed some additional light. But as it stands I feel (as I said earlier in the season) that it's the defence where the biggest ground can be made. The best-to-worst gap is biggest there and it is given a lower priority by many (most) managers. My other initial observation is that the GK position is more important than I gave it credit for. FPL have been trying to tweak things so that GKs became more of a factor and it appears to me that they may have succeeded.

What is best for me in this survey is that it sharpens the target. I can see what these guys have managed to do. I can therefore see what is possible. And I can therefore set my sights accordingly.

12. Comments on Luck

Last, just a few comments on the luck issue based on this survey. It is easy to point to one factor (such as captaincy) and say that Manager A has been lucky compared to Manager B. But when you look at the 5 main categories outlined in point 11 you can see that each has a roughly equal importance in terms of differentiation between teams at a similar level. And no team excelled in all 5, even among the top 12. The 12th placed team might claim to be the unluckiest; he equalled or beat the top 12 average in every category apart from captaincy. A few different captain choices and maybe he could have won. But then he didn't get great returns from his chips either and he also took more hits than anyone else.

And so, round in circles it goes. I guess my only observation would be (as I have said before) that there are multiple ways for luck to affect a team every week. So you aren't talking about 38 luck/skill events in a season, you are talking about many hundreds if not thousands. That's certainly one of the awarenesses that doing this survey has emphasised for me anyway.

[gallus]

interesting read. would you be willing to make a ppg per million analysis for all four positions?

[Ruth_NZ]

interesting read. would you be willing to make a ppg per million analysis for all four positions?

That comes into the value/squad structure commentary I started before. I won't be doing that right now.

[Billy Bongo]

Interesting conclusion the defence, in that it's the one area you can gain ground, yet number of cs has deteriorated.

So does that suggest you need expensive attacking defenders or pound shop defenders to move points into attack? I need to dwell on that one

Sent from mobile

[Ruth_NZ]

I need to dwell on that one...

Yeah, BB, that's exactly what this is all about for me too. It's stuff to ruminate upon.

[Billy Bongo]

Indeed

I'm over dwelling probably but hey ho can't help it

Sent from mobile

[Sutter Kane]

Well yes I wonder how much sinking money into the defence would limit options elsewhere. And does this in turn mean certain transfers are not possible (without hits) and therefore have an effect on TV. At the start of the season, before the first WC, I'll be picking super-cheap defenders based on pre-season and starting fixtures. Then I'll consider the effect of selecting some expensive ones who have attacking potential in my WC.

The thing that struck me about the top 12 is those captain averages. It's simply unsustainable numbers without divine inspiration.

[Ruth_NZ]

The thing that struck me about the top 12 is those captain averages. It's simply unsustainable numbers without divine inspiration.

Not really, SK. Four of the twelve scored 9.0+ doubled and I agree that those numbers are unsustainably high. But three of the top twelve scored 8.1 or less doubled and that's actually not so unusual this season. I managed 8.2 doubled myself and that was with a (failed) captain punt in the last week, had I been protecting my rank I'd have captained Kane and my average would have been 8.33 doubled.

Even with a range from 9.8 doubled down to 7.6 doubled in the top 12 the swing was only 87 points between best and worst. That's not so huge a gap in the scheme of things, otherwise you wouldn't have been able to find those lower captain averages among the top 12 teams.

[Sutter Kane]

I think it's insanely high as I believe we aim for 7 usually don't we. Unless this was a freak season. I managed slightly over 7 myself which I consider excellent.

For me, 8ppg is ridiculous. Be interesting to get an average from a much higher sample. My guess, possibly wrong, would be that those numbers are preposterous for mere mortals.

[Ruth_NZ]

I think it's insanely high as I believe we aim for 7 usually don't we. Unless this was a freak season. I managed slightly over 7 myself which I consider excellent. For me, 8ppg is ridiculous.

I think that 7.0 doubled number (which I also used to use) is redundant. It has been affected by the TC for one thing. I also think that this season permitted a higher captaincy score than may be usual.

As a cross-check I looked at the top 10 in the FISO league (ranked between 55 and 676 in the OR). The comparative captaincy numbers:

Top 12 Captaincy Averages: 9.8; 9.3; 8.6; 8.6; 9.0; 8.1; 8.7; 8.7; 8.5; 9.3; 7.6; 7.8; Average 8.67.
FISO Top 10 Captaincy Averages: 8.9; 8.4; 9.0; 8.2; 8.1; 8.8; 7.6; 8.3; 9.0; 9.4; Average 8.56.

The FISO top 10 average is slightly below the OR top 12 but not by much. I have started work on a cross-section of the top 20k so we'll see how that compares.

Just as a note, I'd like to say that these details are being recorded to make a picture of what happened last season. I don't expect to extract any hard and fast numbers from the study. It is more to give me (and others if they wish) something to deliberate upon and to use as a reference.

[Gambit]

I got just under 8 doubled from my captains which I'm really pleased with, still didn't stop me having one of my worst season though.

the big problem for me, and it's the one thing I struggle with each year, is my defensive points, I can get the right captains, I can jump on/off key attacking players, but even when I invest more in defence I just can't seem to get a decent return compared to the higher ranked teams. So many GW's where my defence return virtually nothing, and it's not like I pick unusual or risky defenders.

really interested to see the point made a few posts above that the 2nd ranked team only made two defensive transfers for most of the season, I thought I picked a really strong defence at the start this year but I was already ripping it up a month in!

[Ruth_NZ]

TOP 20k DETAILS

So, now the same exercise with a cross-section of the top 20k. Throughout this I will use blue for the top 20k and green for the top 12 to ease the comparison.

To give a cross-section of the top 20k I have selected 12 teams from the FISO league ranked at (or close to) the following points in the OR: 1k, 2k, 3k, 4k, 5k, 6k, 8k, 10k, 12k, 14k, 17k, 20k. So slightly biased towards the top 5k but hopefully enough to create a useful picture. Once again I have partly used Fusen's tool as well as FPL Discovery (and a fair bit of manual labour).

1. Bench Boost

The BB scores in the top 20k were: 10*; 15; 15; 18; 19; 23; 16; 17; 20; 12; 12; 20; Average 16.4.
The BB scores of the top 12 were: 15; 23; 25; 27; 18; 30; 1*; 18; 17; 3; 35; 18; Average 20.

I guess we will see this in most categories, the top 20k cross-section undershooting the top 12 but not hugely. In fact, 7 of the top 12 were in the 15-25 range for the BB and 9 of the 20k group were in that range. What is missing in the 20k group are the 2 or 3 high outliers. That suggests that the top 12 were more willing to make different decisions to the majority (see below) because you can't get the high outliers without doing something different to most others.

It should be noted that I have again adjusted the BB scores to represent the weakest bench that week in cases where the manager hadn't done so.

2. Triple Captain

The TC scores in the 20k group were: 15; 9; 3; 14; 9; 9; 9; 9; 9; 9; 9; 9; Average 9.40.
The TC scores of the top 12 were: 9; 9; 14; 9; 9; 9; 14*; 17*; 14; 27; 19; 9; Average 13.25.

Coupled with the BB results, this starts to create an interesting picture. Every team in the top 20k group ranked below 4k triple-captained Aguero for 9 points in GW27. Every single one (and that's 8 in total). Whereas if we take the top 12 plus the 1k-4k teams (16 of them) then 9 of them (56%) did something different. Same (but to a lesser degree) with the BB; every one of the teams from 5k-20k used the BB in GW37 but the higher-ranked teams varied from that more (though the majority were GW37 too).

This will be considered to be controversial in some quarters (although I doubt they will be reading this) but I think this is quite telling. There is a herd mentality even among the basically competent managers that achieve top 20k finishes. Aguero TC GW27 was the strong FFS advice. BB in GW37 was the strong FFS advice. And every team in this survey ranked 5k or below followed that advice (or came to the same view themselves). Whereas teams in the top 5k were much more willing to be different or to make a different decision. That is a fascinating indicator in my book.

3. 2nd Wildcard

The 2nd Wildcard was played in the following gameweeks: 36; 36; 36; 30; 36; 36; 36; 21; 34; 36; 36; 35.
The 2nd Wildcard was played in the following gameweeks: 36; 34; 35; 36; 36; 36; 33; 36; 36; 22; 33; 36.

It's a very similar picture for both groups here.

4. Captaincy

The Captaincy averages in the top 20k were: 8.2; 8.5; 7.2; 8.4; 8.0; 6.7; 7.6; 7.7; 7.5; 8.1; 8.7; 7.8; Average 7.86.
The Captaincy averages of the top 12 were: 9.8; 9.3; 8.6; 8.6; 9.0; 8.1; 8.7; 8.7; 8.5; 9.3; 7.6; 7.8; Average 8.67.

Please note that the averages are 1/77th of the total captaincy points to account for the TC. So these are the base scores, not doubled.

I noted before that from my experience the top 12 average was unnaturally high. Having now looked at the top 10 in the FISO league (see above) and the top 20k cross-section, I'm not so sure. There were 3 or 4 high outliers in the top 12 that don't appear in the 20k group (same as with the BB) but apart from those there is not such a huge difference. The swing between the captaincy average of the top 12 and that of the 20k group was worth 65 points.

Edit: in view of gallus' observations below I had a quick look at the back data using 3rd Turd's tableau records to see how many players reached 190+ points over the season (equates to 5 PPG). These might be considered 'easy' captain picks for their high-potential games.

• 2016/17: 8 players (Sanchez, Hazard, Alli, Eriksen, KDB, Kane, Lukaku, Costa)
• 2015/16: 4 players (Mahrez, Vardy, Kane, Özil)
• 2014/15: 5 players (Hazard, Aguero, Sanchez, Kane, Silva)
• 2013/14: 5 players (Suarez, Yaya, Gerrard, Hazard, Sturridge).

It looks as if the consistent higher performance of the premium players in 2016/17 has pushed the captaincy averages up. 2015/16 was probably especially difficult as there was an understandable reluctance to captain Mahrez (starting value 5.5m) and Vardy (starting value 6.0m) for a long time.

In any case, I am now revising my targets. It is obvious from the above that a realistic captaincy score will vary from season to season but as from next season I will use 8.0 undoubled as my initial benchmark for a par score.

To be continued.

[gallus]

I got 8.9 on average from my captain. there was some luck involved, but between Sanchez, Kane and Ibra it was rather easy to get a brace every other week.

edit: Is there any correlation between team value and captain points? It would be interesting to see if low value teams that can only afford 2 premium players can match the captain points of high value teams with 3-4 premium players.

[Ruth_NZ]

edit: Is there any correlation between team value and captain points? It would be interesting to see if low value teams that can only afford 2 premium players can match the captain points of high value teams with 3-4 premium players.

I can't see one really. In my opinion TV is very over-rated. A good manager will very easily be around 105-6m by the time the season is done anyway.

The top 12 had a range for final TV between 103.2m and 107.1m, nothing startling. Most were around the 105-6m mark and I can see no evidence that any of them chased TV (though I didn't go through each team transfer by transfer). The average final TV in the 12 was 105.3m and the 2nd and 3rd best on captaincy (both with 9.3 doubled) were the two lowest on TV (103.2m and 103.3m respectively).

[Ruth_NZ]

TOP 20k DETAILS continued

5. Formation

When one formation was mainly used I have named it and noted how many times it was used; where multiple formations were used I have named the main ones in order.

The formations mainly used by the top 20k group were: 352/343/442; 352/343/442; 343(31); 343(22); 343(23); 343(31); 343(30); 343/352; 343/352; 343(32); 343/433/442; 343(23).

This continues the picture from the top 12; only 7 of the group (58%) used 343 predominantly.

6. Hits

The number of hits taken by the top 20k group were: 7; 16; 1; 13; 9; 16; 1; 26; 18; 15; 9; 11; Average 11.83 [Top 12 Average 8.58].

This is no surprise really, as you go down the ranks among active managers you would probably expect to see more hits. The top 20k group spent 13 points on average more than the top 12 and from 10k on down it gets to be quite large numbers. The 10k team took 104 points in hits. Personally I find it hard to imagine why anyone would need so many.

It is interesting, though, that the top 12 all took between 5 and 14 hits, they were ready to do so if they saw the value. Two teams lower down (ranked 3k and 8k) only took one hit each but it didn't necessarily help them to be so hit averse.

7. GK Averages

The GK averages of the top 20k group (points per game) were: 3.7; 5.0; 3.8; 4.8; 3.5; 3.7; 4.2; 4.7; 5.0; 4.2; 4.9; 4.1; Average 4.30 ppg.
The GK averages of the top 12 (points per game) were: 5.0; 3.6; 5.5; 4.6; 5.6; 4.3; 4.1; 4.6; 4.1; 3.7; 4.3; 4.5; Average 4.48 ppg.

Not much difference here, once again it's only the 2 high outliers in the top 12 group that are missing in the top 20k group really. The difference between the 2 averages only amounts to around 7 points, nothing very significant.

[Ruth_NZ]

TOP 20k DETAILS continued

8. Defender Averages

The Defender averages (points per game) of the top 20k group were: 4.0; 4.1; 4.4; 4.3; 3.9; 4.4; 3.4; 4.0; 4.1; 3.5; 3.9; 3.8; Average 3.98 ppg.
The Defender averages (points per game) of the 12 were: 4.2; 4.8; 4.2; 4.1; 3.9; 4.2; 4.3; 4.1; 4.2; 4.3; 4.6; 4.3; Average 4.26 ppg.

I noted earlier that the defender averages seem to be significantly down on what they were even a couple of seasons back and commented about the falling CS ratio. I think the notes above (under captaincy) offer an additional perspective. The premium attackers did very well this season, 8 of them broke 190 points and it looks as if captaincy was easier as a result (and a higher average). It would make sense if the counterpart to that was a lower return being available from defenders.

I will explore this further when I return to the commentary about value and squad structure. I do believe that getting the defence right is potentially a powerful differentiation but I'm not sure yet how best to approach it.

One thing that is clear is that there is a gradual falling off of defender averages as you go down the ranks. The top 20k group had as much internal variance as was visible in the top 12 but there was also a pronounced tailing-off with lower ranks - more pronounced than is apparent within the GK, midfielder and captaincy categories.

In terms of swings (assuming 120 defender appearances per team), the top 12 beat the top 20k by 34 points on average from the defender slots. Again it's not a huge difference between the two groups. The bigger swing is within each group because the variance is quite high.

9. Midfielder Averages

The Midfielder averages in the 20k group (points per game) were: 5.4; 5.0; 5.5; 5.2; 5.1; 5.2; 5.2; 5.4; 5.0; 5.2; 4.6; 5.0; Average 5.14 ppg.
The Midfielder averages of the top 12 (points per game) were: 5.4; 5.5; 5.8; 5.7; 5.8; 5.4; 5.6; 5.5; 5.6; 5.4; 5.6; 5.7; Average 5.58 ppg.

Once again, a fairly consistent group of scores in the 20k group apart from the low outlier from the 17k team. And not a very pronounced falling off at lower ranks; even the 10k team achieved a midfielder average that wouldn't have looked out of place in the top 12. This is consistent with my earlier observation that midfielders probably get more attention from FPL managers (and on sites like FFS) than any other. Perhaps as a result if becomes harder to excel (or to do very badly) with midfielders.

Assuming 160 midfielder appearances per team, the points difference between the top 12 average and the top 20k group average was 74 points.

[Ruth_NZ]

ROTATION PAIRS

I'll pop this here while I have it. These are the perfect rotation pairs next season:

Arsenal/Spurs
Bournemouth/Southampton
Brighton/Huddersfield
Burnley/Chelsea
Crystal Palace/Stoke
Everton/Liverpool
Leicester/Newcastle
Man City/Man United
Swansea/WBA
Watford/West Ham

Nothing too exciting there. Bournemouth/Southampton might be OK depending on defender prices. Leicester/Newcastle might also be of interest. Basically you want a pair of teams that are both solid at home and have cheap enough defender options.

[gallus]

Swansea/WBA? Or do you think think WBA won't have a nailed 4.5 starter?

[Ruth_NZ]

Swansea/WBA? Or do you think think WBA won't have a nailed 4.5 starter?

Maybe I am coloured by the lack of WBA clean sheets last season. But Pulis was under some pressure from the new owners to play a more attractive game, especially at home. And as a result (maybe) their strongest defensive performances came in away games, especially against the big 6. That's when Pulis was happy to park the bus.

I wouldn't discount a WBA defender but I'm not sure about their clean sheet probabilities at home and that's what's important for a rotation pair.

[Ruth_NZ]

TOP 20k DETAILS continued

10. Forward Averages

The Forward averages in the 20k group (points per game) were: 6.0; 6.4; 5.1; 5.4; 6.2; 6.1; 6.0; 5.7; 5.8; 6.0; 5.9; 6.0; Average 5.88 ppg.
The Forward averages of the top 12 (points per game) were: 6.4; 6.1; 5.6; 6.4; 6.1; 6.5; 6.3; 6.3; 6.5; 6.4; 6.1; 6.6; Average 6.26 ppg.

As with the top 12, the top 20k group exhibit a wider range again here. But the gap between the two groups is not huge when the averages are taken. Assuming 100 forward appearances over the season (as before) the swing between the two groups is only 38 points, similar to what was apparent for defenders. It was in the midfield category that the biggest difference was to be found.

11. General Observations

I am keen to complete the work on value and squad structure before the stats from last season disappear so I will leave the comments for now. Before the new season begins I will return to this and make any final observations that seem useful.

[Ruth_NZ]

SUMMER RESEARCH - Wrap-Up

OK, time to wrap up my summer researches, which I'll try to do over the next couple of days before settling down to sorting out my GW1 team. Here's an oddity that I wanted to check out...

DO PROMOTED TEAMS START FAST?

It is a perception I have had that promoted teams tend to out-perform themselves early on - basically that they are running on adrenaline and working very hard and can get some decent early results until all the running they are doing starts to catch up. That interests me because I'd normally expect to wildcard by around GW6 or GW8 and the budget is hard stretched early on as well. So, do players from promoted teams offer value?

I have looked back over the last 4 PL seasons and compared the performance of promoted teams over their first 6 games with their performance over the season. Here is a summary of the results.

(a) Clean Sheets

In their first 6 games, the promoted teams cumulatively managed 16 CSs from 72 games over the 4 seasons. That's a 22% CS strike rate. Over the whole season they kept 104 CSs from 456 games. That's a 23% strike rate. No significant difference at all.

Only 2 of the 12 teams looked at really out-performed their season CS numbers in the first 6 weeks: Watford in 15/16 had 3 CSs in their first 6 games (50%) but 11 over the season (29%). Burnley in 14/15 also had 3 CSs from 6 (50%) with 10 over the season (26%). They were the fast starters. But obviously, with the overall numbers being so similar, there were also teams that improved radically later in the season, for example Palace in 13/14 had no CSs at all in their first 6 but 12 over the season (32%). Essentially everything changed for them when Pulis arrived in mid-season.

(b) Goals Conceded

In their first 6 games, the promoted teams cumulatively conceded 108 goals from 72 games over the 4 seasons. That's 1.5 goals conceded per game. Over the whole season they conceded 728 goals from 456 games. That's a 1.6 goals per game. A marginal difference, at best.

It appears, then, that promoted teams don't in general achieve better defensive numbers initially than they do in the longer term. But it should be noted that 7 of the 12 promoted teams managed between 10 and 12 CSs over the season and conceded between 45 and 55 goals. That's a decent, mid-table kind of level. So, taking defenders from promoted teams isn't necessarily a bad decision as such, given that they usually have playing defenders at the 4.5m baseline level.

(c) Goals Scored

Last thing I looked at was goalscoring. In their first 6 games, the promoted teams cumulatively scored 70 goals from 72 games over the 4 seasons. That's 0.97 goals scored per game. Over the whole season they scored 446 goals in 456 games. That's also 0.97 goals per game. No difference at all.

The only specific difference here was that there were 2 teams that were promoted as champions and had scored unusually heavily in winning the Championship: Bournemouth (98 Championship goals in 14/15) and Leicester (83 goals in 13/14). Both of those teams did out-perform their eventual season numbers initially. Bournemouth scored 1.33 goals per game in their first 6 games in 15/16, falling back to 1.18 goals per game over the season; Leicester scored 1.5 goals per game in their first 6 games in 15/16, falling back to 1.21 goals per game over the season. That might bode well for Newcastle, who scored 85 goals in winning the Championship last time. I guess if you wanted an attacker from a promoted team, they'd probably be the one to look at.

Summary

Well, the perception I started with is well and truly squashed. No evidence of a fast start really, not in general terms anyhow. But what this has done is to clarify that cheap defenders from promoted teams can be good picks, especially if you can find one with some threat for attacking points as well.

[Ruth_NZ]

This thread started life as the "R_NZ No Aguero Experiment" and the first pages are an account of that. Times have moved on and we are now approaching the 17/18 season. So I have re-named the thread accordingly. I intend to post bits of research that are of interest to me (as I have done before) and will probably keep an account of my team too, as much for my own interest as anything. As has always been the case, any comments and/or contributions are most welcome.

[Ruth_NZ]

I wanted to try to get to the bottom of the clean sheet issue and maybe I have found a correlation that makes some sense of it at last?

Here are the total number of PL clean sheets over the last 10 seasons, together with the number of teams managing 12+ CSs that season and the number of teams achieving 6 or less CSs over the season.

16/17 - 214, 8, 3
15/16 - 215, 6, 2
14/15 - 224, 8, 1
13/14 - 230, 8, 1
12/13 - 200, 6, 4
11/12 - 206, 9, 4
10/11 - 191, 7, 5
09/10 - 221, 9, 2
08/09 - 247, 10, 1
07/08 - 218, 8, 2

10 Season Average - 215, 7.9, 2.5

Although the CS percentage has fallen in each of the last 3 seasons, last season's number was pretty much bang on the 10-year average. But the other numbers are a bit more interesting. The number of teams managing 12+ CSs averages 7.9 per season and that never seems to vary all that much. Every season has been in the 6-10 range. This is supported when you look at the total number of CSs achieved by the best 6 defensive teams; over the last 8 seasons the average has been 91.5 CSs per season by those teams (slightly better than 15 each) with a pretty close range of between 87 and 97.

So, that suggests that the significant variation isn't at the top.

The picture for the worst defensive teams tells another story. The number of teams managing 6 or less CSs over a season averaged 2.5 per season with a range between 1 and 5 and the correlation looks pretty good to me. The high outlier season when 247 CSs occurred (08/09) was also the season when only one team fell into the 6 and under range (and that team achieved exactly 6). The three poorest seasons for overall CSs were also the ones when the worst teams failed most badly with 4, 4 and 5 of them failing to beat 6 CSs in those seasons. This is supported when you look at the number of CSs achieved by the 6 worst defensive teams; over 8 seasons the average has been 38 CSs per season by those teams (6.33 each) with a relatively much wider range of between 28 and 48.

So, what this suggests to me is that the overall CS % isn't that relevant at all really. At the top the variance is quite small. What varies the most is the degree to which we can rely on cheaper defenders from the weaker teams, in some seasons it is quite a bit better than others. Which doesn't really help much. It basically puts us back to the idea that in general you pretty much get what you pay for with defenders.

[Billy Bongo]

Good stuff

I think the change in bps will help though, you assume defenders from lower teams will make more passes tackles and blocks and more likely to pick up baps

We had scenes last year defenders from losing teams getting baps

And they are more likely to get baps if they won 1-0

I think

Sent from mobile

[Ruth_NZ]

Good stuff, I think the change in bps will help though, you assume defenders from lower teams will make more passes tackles and blocks and more likely to pick up baps. We had scenes last year defenders from losing teams getting baps and they are more likely to get baps if they won 1-0. I think.

I don't think there are any changes to the BPS from last season? I can't see any myself.

Defenders from lower teams don't necessarily get more passes, though, and in any case the number of passes is irrelevant as long as they make more than 30. Then it's down to pass completion %. But in teams that have a tendency to go long the defenders often don't make 30 passes in the whole game and are thus disqualified from any BPS points at all for passing. That was one reason that Heaton did so well on BPs last season; his defenders were often not getting anything for passing and their BPS scores were suppressed as a result.

[Billy Bongo]

I've got four 4.5s at the moment

Sent from mobile

[MGPT]

Hey Ruth, are you still planning to go with 4-3-3 to start the year? I'd be interested to hear what kind of structure you would use, particularly 5th DEF and 4th MID prices.

[baganboy]

Just read this now - good analysis, Ruth.
Tell me if I am wrong, but if one has to fit 11 players, with the ppgs being the way they are, the 3-4-3 does look the most optimal solution in this case, non?
(And I say this as a 4 at the back fan)

Especially, 3 good (~4ppg kind) players at the back seem the ideal solution mathematically - isn't it?

[Sutter Kane]

Just read this now - good analysis, Ruth.
Tell me if I am wrong, but if one has to fit 11 players, with the ppgs being the way they are, the 3-4-3 does look the most optimal solution in this case, non?
(And I say this as a 4 at the back fan)

Especially, 3 good (~4ppg kind) players at the back seem the ideal solution mathematically - isn't it?

I'd predict the optimum setup would involve some kind of rotation.