MoSe wrote: ↑21 Sep 2017, 22:54I never ran such analysis on my own, but since I got here in Sept 2014, all vets posted assuming/reporting that by proof or consensus it was 20p.
I can't quite remember what the general feeling used to be - something like 20-25 points per season for that type of calc seems to ring a bell indeed. I recall that the main problems in using 'value rates' of this kind are that in practice
1) These value rates should vary across positions (i.e., if you naively calculate ppm for every player and put in line of best fits for gks, defs, mids, atts then the lines all cross the y-axis at different places) so a one-size-fits-all number isn't that useful in practice
2) Value rates within a position should vary every season as players change in price, vary between positions during the season as players change in price, and those rates also vary season to season.
3) Calculating a value rate (a slope of a line of best fit) is questionable since value is not linear even within a position, i.e., the difference in price-vs-expected-points is different for more expensive players than cheap players. Indeed, taking into account rotation or captaincy makes it very difficult to apply it to cheap players or expensive players repectively. In fact, the whole method of calculating value purely by a points-per-million calc is questionable.
4) Most such calculations choose an arbitrary 'base-price cut off' to help determine how much 0.5m is 'worth', usually in the ones I've seen over the years 0.5m below the cheapest player (but sometimes at the price of the cheapest player - meaning base pricers have 'infinite' value and are only used as the starting point team rather than potential replacement - and sometimes 'actual 0.0m' giving precisely the PPM stat). The fact that this cutoff is arbitrary and you get a different value rate from every such cutoff is a fundamental flaw with the calcs.
5) If you want to be able to apply it to your team in practice, you want your calc to not only efficiently use your budget, but also to use enough of your budget, and pure value calcs do not work this way. This is because 'value' players tend to easily stand out and we don't need a value stat to identify them. Value calcs somehow tend to tell you about/are based on what your 'average player' at a given price should be getting, but in reality we're only considering a small pool of players at the top end of each price point.
6) In practice, I don't think anyone really uses such numbers to calculate whether a transfer is worth it or not anyway (maybe once or twice as an exercise while it's a hot topic), so these calcs are not readily maintained.
So, to work out whether Hazard in for, say, a Spurs mid is worth it, it's probably better just to work in a more algorithmic manner and simply look at with the combo in mind; so just ask, do I think spending 1m on player x over my player y will get me more points than Hazard will over Alli? So it's the exchange-barter-system using an expected points comparison that ultimately tends to win for me too, as bagan puts it.