VAPM

[Mav3rick]

Over on Reddit, there have been a series of posts about the VAPM metric (value added per million) as a short-hand to compare players points scoring potential. The equation is simple enough, and is just

(PPG -2) / Price

That means you're taking away the two appearance points (penalising players who play under 60 mins regularly) and then assigning a number that can in theory be used to compare players across positions and price ranges.

I've used the free projection data available at FPL Review to calculate the projected VAPM for all the players over the first 5 GWs, and have attached the spreadsheet to this post.

Anything above about 0.35 is considered decent, and of course the list is front loaded by defenders and keepers, but there are some high valued players in the top 10, including Salah and Haaland.

Any thoughts? Obviously all metrics are flawed...

[blahblah]

I'll open it properly on laptop when 8 can see it properly...

Anything by App needs multiplying by 38 to give it a sense of reality, or correcting injuries etc.

There is an argument that the Base Price of position needs to be subtracted from the price. This comes from you need the following

4.5 X 2
4.5 X 5
5 X 5
? X 3

[blahblah]

Oh, and it should be minus 1 not 2 as the late sub does get a point, but that punishes regular 90 min players?

I think it's best ignored, and common sense to then eliminate outliers eg blah wins Euromillions, buys a Prem Club and gets 38 points in 38 mins extrapolating to 90 points per match for the whole season

[Neath boy]

So only 10 "decent" players per this?

[blahblah]

I opened this with my Open Office, but it couldn't see the table....

Are you using projected points, rather than last season's?

I prefer playing higher\lower with last season's points, rather than a guess, which is in itself an average of something as players are getting fractions of points.....

[Mav3rick]

Well, by decent I mean "on track to win FPL most seasons" if you could pick 11 players at that level.

It's projections over the first 5 GWs only, no backwards looking data.

Blah - with regards to the subtracting the base price, I don't think that works because it gives you players with infinite value, and when you fill the spot in the squad you do have to spend the full asking price which is hard to compare across positions with different base prices.

What I suppose is interesting is that its basically saying "everyone gets appearance points" so they mean nothing and only your additional points matter, and since they are the "value" you are adding, then that value has a cost per million which can be compared across everyone as a short-hand tool. You often see the likes of Salah way down a value list because the "extra" points carry less weight when everyone's totals, including basic appearance points, are being included, but here he features high on the list, so he adds a lot of "value" even though he's the most expensive player in the game.

[blahblah]

Yes, my base prices need 1 adding, or a similar thang

1+(Price-Position Base Price) or lower all the Base Prices by 0.5 etc

There are pages on this from probably more years ago than I want to remember..

Personally, I don't see the 2 vs 1 Appearance Point as long as it is App not Minutes divided by 90 as there has to be a final stage of guessing whether they go Higher or Lower next season. Arguably this could be used during the latter half of a season, if the Higher or Lower can be applied....

I think the last paragraph is very similar to the Base price argument, ie you have to pay 4.5m for a Defender, so what does the extra get you; and your point that they all get 2 points so what else do they add?

[Sutter Kane]

Salah basically trumps everyone every gameweek so this model, rightly or wrongly, probably implies a one premium setup...

[blahblah]

That's pretty much the olde Henry line, but I'm not convinced with KdB and a few others around and people think he repeats his Uber season.

Did he top 300 points as a outlier?

[Mav3rick]

I think there's no reason not to assume Salah is the best pick, but there would be legitimate concerns about the loss of Mane and the integration of Nunez, who is very different from Firmino.

In the last friendly, when Firmino was on the pitch in the first half, Salah was basically the number 9 and virtually the only presence in the box. I'm not sure he occupies that space with Nunez in the team.

Longer term, he of course has no WC either so I expect a big points tally from him with basically two short seasons and a month off in between.

[blahblah]

Indeed to Nunez being different and I noticed the same thing....

There are different ways of looking at Salah and 1 is "he will almost certainly be the highest scorer, and I want him". The others are more complicated, and involve Salah plus a defender vs KdB plus Robbo or VvD, depending on your 1st 2 Lpool defenders; and this season we have Diaz plus 5m vs Salah....

[BLOCKHEAD]

This metric sounds like a trap and is even worse than xG/xA.

Solid goals, assists and cleansheet stats are all you ever need.

[Mav3rick]

It's just a metric built on projections that are themselves based on xMins/xG/xA/xCS so it's a derivative of them for comparison purposes, not an alternative to them.

[goooner]

Indeed to Nunez being different and I noticed the same thing....

There are different ways of looking at Salah and 1 is "he will almost certainly be the highest scorer, and I want him". The others are more complicated, and involve Salah plus a defender vs KdB plus Robbo or VvD, depending on your 1st 2 Lpool defenders; and this season we have Diaz plus 5m vs Salah....

You also need to take into account 5 subs and one of them for liverpool Will be Tsimikas.
How often Will Robertson be subbed before 60 min?
More tricky to trust defenders playing time this year.

[Sutter Kane]

Indeed to Nunez being different and I noticed the same thing....

There are different ways of looking at Salah and 1 is "he will almost certainly be the highest scorer, and I want him". The others are more complicated, and involve Salah plus a defender vs KdB plus Robbo or VvD, depending on your 1st 2 Lpool defenders; and this season we have Diaz plus 5m vs Salah....

You also need to take into account 5 subs and one of them for liverpool Will be Tsimikas.
How often Will Robertson be subbed before 60 min?
More tricky to trust defenders playing time this year.

Logic suggests, almost never. He was rarely subbed off when we had 3 subs so generally he's not in line to be subbed until 4th / 5th. And that won't be before 60 surely. In fact, I'd say the rule benefits Robertson as sometimes he may be subbed off prior to conceding in the last 10 minutes. Maybe same with Trent if Liv are up a couple goals.

With 5 subs, Conte's wingbacks are most at risk for plenty of reasons imo. Maybe Chelsea's wingbacks too...

[blahblah]

Yep. I'm erring on the side of it being good for Robbo et al.

It could be less good or even bad for precious players who managers may want to keep fresh, like Salah?

[Mav3rick]

Yeah I don't think Roberston goes off before 60 unless he's injured. 75 mins before CL games I could see happening, although I could also see Tsimikas being given the CL group games anyway, similar to how Matip and Konate have shared in the past.

I do partly wonder if Tsimikas is a better fit for Nunez though. Just my feeling (without any statistics to back it up) is his crosses might suit a big number 9 more (tend to be more airborne I think).

Perhaps we might see that sort of change made if a game is being chased, but equally, I'm sure Robertson can be told to cross more in the air

[goooner]

Yes maybe he Will play leauge and tsimiks CL.
I dont think he manage both if not Early subs.
For example:
Weekend Wolves home
13-14/9 CL Group
Weekend Chelsea away

Or

Weekend Arsenal away
11-12/10 CL Group
Weekend ManCity Home

Could be Some interesting times if Klopp pushes Robertson eg. I do think he Will need to rotate more (as everyone)

[Hogmeister]

Thanks Mav, interesting and again highlights the reliance than any such models have on the underlying forecasts of points per player.

I did a quick comparison of the top predicted points scorers in this table against the equivalent on the FFH points prediction page, as below. It seems like most players are predicted to get around 10-20% more points by FFH than by FPL Review. Quite big, and fairly consistent, difference.

I don't know whether this is because FFH anticipate more minutes for the top players, or whether they simply expect the big teams to net bigger hauls over those first 5 (relatively easy) fixtures than FPL Review does.

FFH have said that a) you can tailor such predictions yourself if you want to, including how many minutes you think a player is going to get, and b) that they will be refining the predictions behind the scenes during pre-season, as certain players seem more/less likely to cement a starting place etc.

[Hogmeister]

And just did the same on FFS, they have a less flexible prediction tool so you can only do weeks 1-6, a very quick and cursory scan suggests their predicted points are somewhere in between FH and FPL Review...

[zipnolan]

I'm not sure FFS and FFH would want those numbers to be distributed; don't you have to pay to get them (at least on the FFS site)?

[Hogmeister]

I'm not sure FFS and FFH would want those numbers to be distributed; don't you have to pay to get them (at least on the FFS site)?

Yeah, good point thanks. They were both from the members area of the sites, so yes you had to pay for them.

Not sure the raw data was terribly valuable, but just in case have just edited these posts to remove the images.

Happy to answer any questions on specifics.

[zipnolan]

I confess to having access to those projections, but my eyes tend to mist over after a few moments of reading them.... I personally find the FFH hub ones to be closer to my own flawed assessment of relative numbers than the FFS ones, and the FPLReview ones to be a bit better than either. FFH are bizarrely offering money back if they don't win your league for you..... sorry I mean if you don't win your league (by doing everything they recommend). So they must have some faith in the numbers.

Probably in the end the absolute numbers don't matter as much as the relative ones in any of these tables (including VAPM). I'm struggling to see the extra aha moment that is provided by recalculating VPM by taking two points off each player's projected scores. I'd like to understand how those scores are calculated (i.e. how factors such as probability of playing, fixture difficulty, past history, Bundesliga tax etc are factored into the numbers). Also, how fluid these numbers are, how often they are recalculated / refined through the season. I would expect all of these numbers to be a lot more useful after, say, 6 or 8 weeks, so much more applicable to the first wildcard selections.

[blahblah]

I'm not sure FFS and FFH would want those numbers to be distributed; don't you have to pay to get them (at least on the FFS site)?

Yeah, good point thanks. They were both from the members area of the sites, so yes you had to pay for them.

Not sure the raw data was terribly valuable, but just in case have just edited these posts to remove the images.

Happy to answer any questions on specifics.

As I've posted elsewhere: could you report how accurate these things are eg predicting 100 points and you get 10 etc

My best guess is that they will only get close by luck...

[Hogmeister]

Zip - yes agreed, it’s the relative numbers that are most important, so any of the three could be used and if there’s say a 10-20% uplift across the board in FFH vs Review, that probably doesn’t make much difference to the prediction-driven decisions you’d make.

Blah - it’s a good point re tracking accuracy by comparing predictions vs results. Think I will aim to to do that on “The Hog Blog” this season, by putting my team up each week, with the predicted points for the coming week from FFH (and if I’m feeling enthusiastic) the other two, and the following week checking the accuracy of those predictions. Can’t promise to do that every week, as often my enthusiasm for such “bright ideas” often wanes once the season is underway, but I’ll try to do that anyway.

[blahblah]

We could turn it into a Comp, where we compete against the 3 bots

[dino1980]

This metric sounds like a trap and is even worse than xG/xA.

Solid goals, assists and cleansheet stats are all you ever need.

One of those set of stats is descriptive and one is predictive. I’ll take the latter every time for FPL purposes.

[Neath boy]

41D9C301-C69D-4A16-B777-5E86C144611E.png

This is now getting silly and quite pathetic. What is the point in playing if you just copy someone else’s plan and pay for the privilege!

[zipnolan]

It does rather make me hope they all crash and burn.....

[zipnolan]

Incidentally (bit of a digression), I used that FFH rate-your-team tool yesterday. It's all over the place. I watched a video with "Bigmanbakar" getting a 97% rating for his squad. I then put exactly the same squad into the tool (just changing the order I put in the players) and it came back with 88%! I mean, I'm sure I'm a worse manager than he is, but I am not that bad at data entry....