Community

Points Gained over the Lifetime of a Transfer

A crude and widely-used metric for assessing the success of transfers, either individually or collectively across a season, is immediate points gained from transfers. What I’ll argue we really need is a more detailed measure of transfer success, which I’ve started recording for my own transfers this season.

Basically, the theory is that a zombie team you could create would already produce the greater share of the points you need, perhaps 2200 points if you played your Triple Captain, Bench Boost, and Free Hit chips at logical times, but did not make any transfers or play wildcards. (With hindsight, you can usually produce a zombie team that will win FPL, but a realistic zombie team constructed without hindsight should be expected to produce considerably less.) Playing your wildcards at logical times might gain you an additional 50-100 points. To get to about 2500 points, as most of us want for a high rank and a chance of winning mini-leagues, each transfer outside of wildcards needs to gain us at least a certain number of points V (for value) on average. The purpose of this article isn’t to calculate V precisely, but to explore how we can track whether each transfer has done its job or not. I am going to assume for now that V is probably more than the 4-point cost of a hit. Later we’ll come back and do a few calculations to make a rough estimate of V.

Defining PGLT

To construct our metric, we’ll define Points Gained over the Lifetime of a Transfer (PGLT) as the number of points that the players bought outscore the players sold, minus any hit(s), from the time of the transfer until either a wildcard, or the players brought in are sold. For example, my Gameweek 3 transfer was Harvey Barnes to Said Benrahma for free. I’ve hit the wildcard button in advance of Gameweek 8, so this transfer had 5 gameweeks to justify itself, as long as I didn’t sell Benrahma before the wildcard. (If I keep Benrahma on the wildcard, I consider the calculation to still end with the wildcard, because I could have taken out or brought in any player on the wildcard without spending any transfers. By my theory of mini-seasons, the wildcard might as well be the end of the season for most purposes.) Assuming I kept Benrahma at least until the wildcard, the PGLT is simply the number of points by which Benrahma outscores Barnes in Gameweeks 3-7. It would have been negative if Barnes had outscored Benrahma for that period.

For a double transfer or more, the PGLT would be calculated similarly, although the periods that count may be different for each player bought. For example, in Gameweek 4 I made the transfer of Son Heung-min and Callum Wilson to Demarai Gray and Cristiano Ronaldo for two free transfers. If I held both Gray and Ronaldo for the same length of time (or held both until a wildcard), then the calculation is easy – simply the number of points scored by Gray and Ronaldo, minus the number of points scored by Son and Wilson, over that period of time. If I did not hold them the same length of time, then it is not as straightforward, but to make the calculation possible, I’ll pair up the more expensive player sold with the more expensive player bought, and the cheaper player sold with the cheaper player bought. For example, if I held Gray for 2 gameweeks and Ronaldo for 3 gameweeks, then my calculation would be (Gray points GW4+5 + Ronaldo points GW4+5+6 – Wilson points GW4+5 – Son points GW4+5+6). The players bought, in this case, have a total of 5 player-weeks to justify the two transfers spent.

I’ll have to admit that captaincy does muddy the waters somewhat. If I make two transfers that include bringing in Ronaldo to captain him, then do I adjust PGLT over the entire lifetime of the transfer to account for the value of Ronaldo as a captain option? In theory, I might, but to keep it relatively simple, I’m recording the benefit or harm of captaincy only for the week that I bring the new captain in. I’m making a note before the deadline of who I would have captained in the absence of my new signing (Mohamed Salah in this case), and including the points gained by captaining Ronaldo instead of Salah. Since I didn’t sell Salah and still got his non-captain points, I shouldn’t double those points for captaincy. If I had sold Salah, then I would subtract double Salah’s points. As it turned out, the Gameweek 4 double move was one of my best transfers, producing 25 points gained in the first week alone. Ronaldo as captain returned 26 points, Gray returned 7 points, and Salah got 8 points, so I deduct 8 points lost by not captaining Salah. Son and Wilson were both injured, so I didn’t have to subtract any points that they got, until Son returned from injury and started beating down my numbers.

At the end of the season, I’ll be able to look at the PGLT for every transfer and see the average points I gained over the lifetime of the transfers I made. I’ll also be able to total it up separately for times I made a single transfer vs multiple transfers. I may decide to exclude transfers done a very short time before wildcards, because it’s probably unfair to expect a one- or two-week punt made just before a wildcard to pay off as well as a normal transfer for a player who will be held at least four weeks or so.

Estimating V

So now I can’t resist trying to take a stab at estimating V and exploring a few consequences. I think most of what I’m going to say is not overly dependent on the exact value of V, as long as it’s more than 4 points. Some of it may apply even if V is less than 4 points.

As stated at the beginning, I’m going to estimate that you could realistically get to about 2300 points without making any transfers outside of playing wildcards and Free Hit. To get to 2500 points, you’d need to gain about 200 points from transfers. You get 34 free transfers over the course of a season – 38 gameweeks minus two for wildcard gameweeks, one for Free Hit, and one for Gameweek 1 when you don’t make official transfers before the first deadline. Dividing 200/34 makes V about 6 points based on the assumption of no hits. If you take some hits, then you have to make up those points from hits on top of the 200 points, so if you take 16 hits costing 64 points, V becomes 264/50, which is only 5.28. Taking hits thus actually reduces the amount of gain you would need to make on average per transfer, but this benefit may be offset by the difficulty of finding a third transfer that gains you as much as the first and second transfers. In practice, mini-wildcards of three transfers for a -4 often involve two downgrades to fund a single upgrade.

A value of V=6 points may seem high, but looking back at my season report for 2020-21, I gained 267 immediate points from 43 transfers, minus 36 points of hits, for an average of 231/43=5.37 points per transfer. These are immediate points gained in the first week only, so I probably gained significantly more than this over the whole lifetime of my transfers, and I still came up well short of 2500 points. (Admittedly, the immediate points metric as calculated by FPL Statistico does appear to include doubled captain points, while not deducting points from the player who would have been captained if the transfer had not been made, so it’s more generous than PGLT in that respect.)

One hypothesis I intend to test using the PGLT metric is that two transfers are not actually more than twice as good as one. If I use one transfer, I just need to gain 6 points over the lifetime of that transfer to justify it, whereas using two free transfers, I need to gain 12 points. Chances are, when using two free transfers, unless I have substantial money in the bank, one will be a downgrade that will actually lose me points. Therefore, the other one will have to gain me considerably more than 12 points. If that player is only being kept for a few weeks before a wildcard, he might need to gain me 3-4 points per week over the player he replaced. Unless replacing an injured player, even a premium asset is not that likely to outperform a mid-priced asset by so much.

Conclusion

I’m ending the first mini-season on a total of 34 PGLT from 6 transfers, slightly below my target. I anticipate that the remaining mini-seasons will have some blank and double gameweeks to exploit to pull the average up. While it takes a bit more effort to track, hopefully PGLT can provide a more useful metric to evaluate our transfers than simply immediate points gained from transfers.

Major League Shocker Contrarian stock market investor (which might creep into my FPL thinking a bit too much at times). Resisting groupthink. Overcoming cognitive biases to make more rational decisions.”

13 Comments Post a Comment
  1. Rotation's Alter Ego
    • Fantasy Football Scout Member
    • Has Moderation Rights
    • 10 Years
    12 days, 4 mins ago

    Thank you MLS! Enjoy your articles, they always leave me thinking.

    Really interested to hear how this turns out!

    1. Major League Shocker
      • Fantasy Football Scout Member
      • 1 Year
      11 days, 16 hours ago

      Cheers RAE! Appreciate the kind words.

  2. fusen
    • 9 Years
    11 days, 8 hours ago

    Interesting approach. I do find this a bit of a rabbit hole though.

    Immediate points gained is simple/(dumb) but that is why it was used, every time you expand from just that first weeks point difference you start to add more and more asterisks to explain caveats to your advanced formula.

    I don't think there will ever be a proper way to judge a transfer over an entire season as every single move you make creates too many different strands of what ifs

    1. Ginkapo FPL
      • Fantasy Football Scout Member
      • 10 Years
      11 days, 8 hours ago

      Would it help to admit that not every transfer is the same? It then becomes very simple.

      When you make a transfer set out clear expectations for the transfer and record them. Routinely review the transfers against expectations and consider how to adjust decision making when there is a trend of discrepancy good or bad.

      1. Major League Shocker
        • Fantasy Football Scout Member
        • 1 Year
        11 days, 7 hours ago

        Yes, I have thought of eventually refining the system with something like "par" like in golf. For example, par would be a higher PGLT for Son and Wilson out because they were injured and would not contribute any negative PGLT while injured. It would be lower for a move made right before a wildcard that has only one week to pay off.

        It sounds like you have a good system for tracking already. That's good to hear!

        1. Ginkapo FPL
          • Fantasy Football Scout Member
          • 10 Years
          11 days, 7 hours ago

          To be clear my system is just the base level for logic decision making

      2. fusen
        • 9 Years
        10 days, 21 hours ago

        I was referring to coding this sort of thing on a website. What you've said makes sense for a human planning things but I can't add that to FPL Statistico 😛

        1. Major League Shocker
          • Fantasy Football Scout Member
          • 1 Year
          10 days, 19 hours ago

          I completely understand, I would expect tracking PGLT would always require a spreadsheet and a substantial amount of effort from an FPL manager wanting to track it, and probably couldn't be automated.

          I apologise that I didn't realise you were FPL Statistico when I first responded. Your tools are very useful. Tracking captaincy points is particularly helpful.

    2. Major League Shocker
      • Fantasy Football Scout Member
      • 1 Year
      11 days, 7 hours ago

      @Fusen

      Yes, I have thought there could be situations where it becomes very complex. So far this season it has not been, but in a longer mini-season I could imagine situations where I bring in a new player one week after my first wildcard, and then after a while it is hard to know who to compare that player to -- maybe because I've subsequently brought the player back in that I took out. Over the season, I think I'll be able to iron out these details with reasonable solutions. For a manager who takes a lot of hits, it might be even more complex to track than it is for me.

      Some things I want to know by tracking PGLT over the season are things like:

      1. What is the typical gain from 1 vs 2 vs 3+ transfers? Should I prefer one of these options more vs the others?
      2. What percentage of points gained are in the first week?
      3. Am I wasting transfers on a certain kind of move (e.g. defender transfers) that I'm finding has a lower PGLT than other moves?
      4. Do I get an overwhelming percentage of my PGLT around blank and double gameweeks? Are these times when it would actually pay to take more hits even though I don't normally like hits?
      5. Do I waste transfers trying to optimise chips, when it would be better to accept a lower return from the chip but have more transfers available for more productive purposes?

  3. papercarapace
    • Fantasy Football Scout Member
    • 3 Years
    8 days, 5 hours ago

    amazing - we need more analysis like this to drive our understanding of the game forwards

    1. Major League Shocker
      • Fantasy Football Scout Member
      • 1 Year
      8 days, 5 hours ago

      Cheers mate! There will definitely be more to come.

  4. Sharkytect
    • Fantasy Football Scout Member
    • 7 Years
    7 days, 7 hours ago

    This is brilliant.

    Now, could someone devise a formula that can automate these calcs/assessments for us, so we don't have to put in the same amount of work as this author!?

    1. Major League Shocker
      • Fantasy Football Scout Member
      • 1 Year
      7 days, 5 hours ago

      Cheers Sharkytect! It really only takes me about 5 minutes per GW to update the counts in my spreadsheet and the formulas. Basically each row just contains one player and there is a column which is either Bought, Sold, Bought+Captained, Didn't captain (e.g. Salah in the GW4 example), or Sold+Didn't captain. It shouldn't be too hard to set it up so it automatically puts Bought and Bought+Captained rows on the positive side of the ledger, and the others on the negative side.

      The tricky part would be automatically loading in the points for each player for each week.