Community

Is the Joe Lepper’s Goal Imminent Table the Best Way to Forecast FPL Points

INTRODUCTION

The first four Gameweeks of FPL have been such a roller-coaster for the so-called ‘non-casual’ FPL manager. Ubiquitous information has been a double-bladed sword, ruining transfer plans and overall points. The injury-story of Son Heung-Min, apparently, was the most recent highlight as he scored 18 points when his status in FPL is flagged as 25% chances of playing due to injury. The casual players who ‘forget’ to transfer him out pre-deadline are the ones singing joys, while the more competitive managers often find themselves regretful. In a season where the final scores become more and more unpredictable, what do we have in our pocket to narrow the odds?

The statistics have been very deceiving early this season. All the research on the underlying stats was rarely fruitful, leaving blanks left, right and center on captain selections. Kevin De Bruyne, for example, was totally electric in Gameweek 2. Without any proper pre-season, he singlehandedly destroyed the statistically best defence at the end of the last season. Wolves, who conceded the least amount of shots in the box, big chances, and had the lowest xGC since Gameweek 25 last season until the beginning of this season, was sliced through like butter, and De Bruyne had 6 chances created all by himself, more than half of the whole Wolves’ players managed the whole game. Add that with the most goal attempts in the match, 1 goal, 1 assist, and maximum bonus points, he was definitely a very good captain option coming forward. Since then, he has played against the Leicester City side who had just conceded twice against Burnley and the Leeds United side who conceded 8 times in 3 games pre-Gameweek 4. The result is so obvious in FPL perspective – two successive blanks. There are numerous more examples of this, but the miseries are not what we’re going to talk about in this article.

It’s very hard to jump on players’ haul at the right time, which seems to be key in FPL, particularly in this kind of season. So many points missed, and jumping to players’ bandwagon after their hauls are good, but not best, so to speak. Regarding the prediction on when and which players are gonna haul, the famous tool to forecast such thing is the goals imminent table. Full credit to Joe Lepper who came up with the idea of creating such a powerful table! In this article, I am going to talk about how the table was set initially by Joe, how it has been working in the first four Gameweeks, and analyzing its flexibility in forecasting the next players to haul.

THE GOALS IMMINENT TABLE

Principally, the goals imminent table is constructed to scout the incoming players that are either coming very close to scoring goals or just cannot finish their chances. Using the Opta-stats driven data, Joe filtered the goal threat stats of every single FPL assets based on certain criteria in the last four Gameweeks.

His boundary conditions are set, currently, to select the players who make minimum 10 goal attempts, have at minimum 0.20 xG, but have not scored more than 1 goal through those shots. He assumed that the boundary conditions represent the players good positioning, as they can take the number of shots given, and the quality of their shots is bounded by the minimum xG to filter players like Townsend or Shelvey who likes to shot on sight from everywhere across the pitch. The table is updated each Gameweek, working on a moving average environment. The dynamic updates automatically filter the players who cannot finish their chances and drop their confidence after repetitive failures.

In this article, as we are not even yet into Gameweek 5, I tweak the table’s boundary condition a little bit. For Gameweek 2 only, I decrease the number of required shots to 5 instead of 10. The Gameweek 3 and 4 have the same amount of shots boundary of 10. The data is not collected based on four Gameweeks data, but from gameweek 1 until the latest available data. Let’s see how the table performs.

GAMEWEEK 2 PREDICTION & RESULT

In Gameweek 1, there were 5 players who met the criteria: Raul Jimenez, Richarlison, Harvey Barnes, Timo Werner, and Sadio Mane. Just to remind, each of them had at least 5 goal attempts with a total xG of at least 0.2, but did not score more than 1 goal. Using the data in Gameweek 1, the goals imminent table predicted these 5 players to get goals in Gameweek 2.

The results were mouthwatering. Seven goals, 1 assist, 1 missed (penalty) assist, and 7 bonus points were collected by these five players in Gameweek 2, combining 49 FPL points among them! Harvey Barnes, Richarlison, and Sadio Mane had double-digit hauls, and Timo Werner was the only blank player from the list.

GAMEWEEK 3 PREDICTION & RESULT

The minimum shots limit was updated back to 10 in Gameweek 3, and there were only 2 players who passed the criteria, Richarlison (again) and Harvey Barnes (again). I am not sure if the table worked in consecutive weeks, but the prediction was the two players should be scoring goals in Gameweek 3 based on the first 2 Gameweeks data.

The results were not as magnificent as the first Gameweek, but considering the freakish Gameweek we had since Gameweek 2 to Gameweek 4, we will take 10 points from two players, right? Harvey Barnes did blank, but Richarlison scored 1 goal and got all 3 bonus points, returning 8 FPL points for his owners.

GAMEWEEK 4 PREDICTION & RESULT

Gameweek 4 was arguably the hardest Gameweek to predict, but on the flip side, the goals imminent table now has 3 Gameweeks worth of data, more than what it previously had. The tables listed Harry Kane, Richarlison (again), Michail Antonio, Harvey Barnes (again), Timo Werner (again), and Tomas Soucek. The players I give the ‘(again)’ sign are now either very close to returning or just cannot score, seriously.

In hindsight, it turned out to be one of the most unpredictable Gameweeks, ever. Manchester United lost 1-6 at home, and the reigning champion, Liverpool, got battered 7-2 by the reinvigorated Aston Villa. But, once again, the goals imminent table did not disappoint. The six players listed combined 5.33 points on average among them, having 3 goals, 1 assist, 1 disallowed goal, and 5 bonus points. That includes the unluckily injured Richarlison with only 1 point return in the mix.

OVERALL PERFORMANCE

Between Gameweek 2 and Gameweek 4, the goals imminent table has had 8 unique players in the list, and so far 6 players have produced at least one attacking return! The only two players who failed to live up to the expectation were Timo Werner (who in reality is indeed a disappointment) and Tomas Soucek (who to be fair is a 5th midfielder material, and his 13 point returns in four Gameweeks is nowhere near shameful considering his kind price).

On average, the players have returned 91 FPL points among them from 11 goals and 2 assists combined. That is a remarkable achievement in a season where each Gameweek, at least two assets among Salah, Mane, Aubameyang, De Bruyne, Sterling, and Bruno Fernandes always blank.

ROOM FOR IMPROVEMENT

As so far the table looked to be working perfectly fine, FPL managers are, sorry to say, hard to satisfy. If there is another way to get more points, we will do – and that’s exactly what I am doing here. I am looking for different boundaries and see how the table can be even more optimized. In short, I decrease the shots lower boundary to 5 goal attempts and lengthened the evaluation period up to the end of Gameweek 4. This way, the table for Gameweek 2 is not necessary to be discussed as the original table was tweaked that way, anyway.

For Gameweek 3 and Gameweek 4, the results are good enough compared to the original table. Obviously, we loosened the criteria here, so there will be more players listed and it brings more uncertainties in FPL perspective. Here is the result of the simulated new boundaries.

There are 10 new players listed going into Gameweek 3: Giorginio Wijnaldum, James Ward-Prowse, Aaron Connolly, Jonjo Shelvey, Denis Praet, Callum Wilson, Michail Antonio, Che Adams, Tomas Soucek, and Jarrod Bowen. Up to the end of Gameweek 4, each of these players has 2 Gameweeks to prove their ‘goal imminency’. Eventually, 5 players have returned more than 6 points in Gameweek 3 and 4 combined, an incredible 73 points in total to be specific, while the other 5 failed. The ‘failure five’ are Wijnaldum, Ward-Prowse, Connolly, Shelvey, and Praet (injured for Gameweek 4).

Going into Gameweek 4, the new table gives 12 new names: Kevin de Bruyne, Lucas Moura, Mason Mount, Grady Diangana, Josh Brownhill, Giovanni Lo Celso, Andy Carroll, Solly March, Jack Grealish, Leandro Trossard, James Rodriguez, and Ollie Watkins. Unluckily, 3 players did not feature in Gameweek 4; Mount and Carroll due to the surprising team selections and Lo Celso due to injury. From the remaining 9 players that played, 5 of them returned 7 goals and 7 assists combined, returning in total pleasantly 71 points!

ANALYSIS ON THE EFFORT TO IMPROVE THE TABLE

Unavoidable uncertainties and noise appears when the shots’ lower boundary is decreased. The number of failures increases significantly, with only 2 players having blanked on the original table (Werner and Soucek) compared to the 12 on the new table, although 3 of them are unlucky absentees.

On the positive side, the hauls are still detectable, especially through Wilson, Grealish, Rodriguez, and Watkins in Gameweek 4. It looks like more a gamble than the previous list, but the upside is still there.

Most importantly, we can see a pattern on the new table results. The ‘failures’ are mainly the central or box-to-box midfielders playing no further than the number 10 role. Wijnaldum, Ward-Prowse, Shelvey, Praet, Brownhill, and Lo Celso are the names I mean. The strikers/wingers/attacking midfielders, on the other side, are mostly successful.

We can’t deny, though, that the original table gives us more certainties and confidence, although the options are significantly more limited. Additionally, the lower number of shots boundary are more applicable to more advanced players, while the higher number is more widely applicable to any positions.

CLOSURE

In summary, we have been diving into Joe’s goals imminent table. In the first four Gameweeks this season, the table is undeniably one of the most powerful tools to predict hauls. However, its selection and flexibility can be put to test; the more accepting you are to uncertainties, the wider selection you have with higher risk. Whether the table is going to work perfectly the whole season remains to be seen, and don’t forget that as powerful as the goals imminent table has been, you should not just ignore the other tools you have to play your game. Have a good international break and may Gameweek 5 onwards are more kind to us than the first four!

24 Comments Post a Comment
  1. Rotation's Alter Ego
    • Fantasy Football Scout Member
    • Has Moderation Rights
    • 12 Years
    3 years, 5 months ago

    Cheers SRR!

    Really is a fantastic tool, especially when you pair it up with player stats / eye test etc. Fancy a lot of the names on the list this week to move off soon - Werner for one (finally) but keen on Trossard as well.

  2. Dr. Ante Pavlovic
    • 6 Years
    3 years, 5 months ago

    Can you RMWCT please? Thank you

    Mendy, Button
    Chilwell, Semedo, Justin, Mitchell, Branthwaite
    Salah, Sterling, Son, JRod, Zaha
    DCL, Jimenez, Brewster

    Mendy will stay if he is fit, if not I will go for Ryan and KWP in for Mendy and Branthwaite or Mitchell. Mitchell will go out for Ferguson if PVA is back. Zaha will go out for Grealish if he blaks vs Brighton.

    1. Dr. Ante Pavlovic
      • 6 Years
      3 years, 5 months ago

      Blanks*

    2. VincentVega
      • Fantasy Football Scout Member
      • 12 Years
      3 years, 5 months ago

      Do Grealish now imo...far better long term option.

  3. FPL Virgin
    • Fantasy Football Scout Member
    • 7 Years
    3 years, 5 months ago

    No.

    Goals imminent is flawed. It needs columns for shot accuracy and xGOT and shooting goals added. This is because some players have a long history of a below 10% conversion rate and they will stay on the table for a long, long time.

    1. TopMarx
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      3 years, 5 months ago

      I wouldn't say it is flawed. I think it can be enhanced by looking at how well a player converts his chances historically, but this is beyond the scope of the table and is something each FPL manager can do for themselves (research what you can on players new to the league, or add together season totals for players already in the PL). So I would use the Goals Imminent table as a starting point for players to be evaluated.

      I am interested to see how it performs once the season settles down, presuming it does settle down!

      1. FPL Virgin
        • Fantasy Football Scout Member
        • 7 Years
        3 years, 5 months ago

        Let's break it down.

        What is it's basic premise?

        Players who have a high volume of goal attempts without scoring many goals. For me, that's not revealing if a player is "due"; it reveals which players aren't clinical and couldn't hit a barn door.

        To reveal if a player is in fact due, you need to go way beyond the last four gameweeks and sanity check 1 to 3 seasons worth of data for the player ... while also looking at post xG data to see if the player's shooting ability is a factor.

        Richarlison is a case in point. He is currently top of goals imminent but his goal conversion rate this season is a measly 7% - and even that was a penalty! Moreover he has been a notoriously bad finisher and has had a below average goal conversion rate for every season he has been at Everton. So all those people who bring in Richarlison because he is top of this famous table are going to get stung because they haven't scratched beneath the surface of the data. And THATS why goals imminent is inherently flawed and that's why Mark eats a pie when Jonty wheels it out.

        1. Kroos Kontrol
          • 11 Years
          3 years, 5 months ago

          I actually looked at this statistically. How can a "notoriously bad finisher" play for a side which has been a disaster for 2 seasons, not even be on pens consistently, play out-wide most of the time, and yet get 13 goals!

          Richarlison has had a conversion rate of 15% in each of the last 2 seasons. It is only slightly below overage when you look at all the players with 10+ goals over the last 3 seasons (19%). He gets away around 90 attempts a season, which is above average average (80 attempts. again, for players who get 10+ goals in a season).

          To complicate this further, Richarlison had 90 attempts, but only 13 big chances. The average ratio is that for most players (again, for all players with 10 or more goals in the last 4 seasons), for every 4.5 shots, you get 1 big chance. Richarlison averages around 6.50 shots per big chance, which might perhaps explain why he has a low conversion rate.

          But here's the kicker to your "he couldn't hit a barn door", according to the xG model used in FFScout members area, Richarlison has actually outperformed his xG by 3.93 last season, and 3.41 the season before that. That's the top 25% of all players with 10+ goals in the last 3 seasons.

          The reason the table works is that season-on-season -- there is a positive correlation with: 1) more attempts = more goals, and 2) more big chances = more goals. So if you get a high number of shots/big chances, it should lead to more goals given the high correlation.

          Are there outliers? Sure. But you have to do it in a way that doesn't make everything confusing with 20 different variables for each players. Richarlison would at worst be an outlier (e.g. Mitrovic 18-19), but how does that make it "inherently flawed"?

          Also, you make it seem like conversion rate and all those things don't change from season-to-season. Harry Kane had a conversion rate of around 16% during 17-18 and 18-19, but it went up to 22% last season. Would look at conversion stats really have helped you last season?

          1. FPL Virgin
            • Fantasy Football Scout Member
            • 7 Years
            3 years, 5 months ago

            +1

  4. Now I'm Panicking
    • 9 Years
    3 years, 5 months ago

    Cheers for this. Would be interested to see how other methods performed too, including a pin and a blindfold.

  5. RedLightning
    • Fantasy Football Scout Member
    • Has Moderation Rights
    • 13 Years
    3 years, 5 months ago

    The Goals Imminent Table may reveal a few players that might otherwise have passed under the radar, but why does it have to exclude players that have actually scored more than one goal already?

    I would like to see it also include these players too so that we can get a fair comparison between those who have scored more than one goal already and those who haven't.

    I would be very surprised if in the majority of cases where players have similar stats for shots and xG as well as similar past and future fixture difficulties, the players who have already scored more than one goal do not usually score at least as many in the next week as the ones who haven't.

    I would only be interested in those players in the table whose stats, fixtures and prices compared favourably with players who had already scored more than once.

    1. FPL Virgin
      • Fantasy Football Scout Member
      • 7 Years
      3 years, 5 months ago

      I agree. The one goal max filter can skew the data in some circumstances.

      1. Black Belt Jones
        • Fantasy Football Scout Member
        • 8 Years
        3 years, 5 months ago

        I agree too mate. If the tool showed players who are scoring and likely to keep it up, it would be super useful!
        As it stands, it seems more of a tool for seeking differentials. This has its place, but you're really at risk of bringing in the most frustrating players to watch!

  6. Sgt Frank Drebin
    • 7 Years
    3 years, 5 months ago

    I don't want to be rude but the goals imminent table is based on biased and very flawed reasoning. It's just better to look just at the stats all of players, it doesn't matter really if someone hasn't scored a goal for a month or a year. He still has the same probability of scoring a goal from a given chance. It's like with flipping a coin 8 times and having it on tails 8 times, it doesn't mean that the chances for flipping heads is higher. It's 50/50 no matter what the previous results were. Che Adam's may be the worst finisher on the world or maybe he was just very unlucky so far. But it still doesn't mean he HAS to score soon. He could as well waste another big chance as the chances for him scoring from it are let's say 50/50. Given how many players there are such streaks aren't something extraordinary. There is for example 6% a player will waste 4 big chances in a row (naively assuming 50/50 chances), so roughly 1 in 17 players will happen to do that. Nothing impossible.

    1. Sgt Frank Drebin
      • 7 Years
      3 years, 5 months ago

      Assuming confidence doesn't play a role in that. But that would be another argument against taking goals imminent into consideration. And it may be something in it as recently there were some research that the "hot hand fallacy" may be in fact not true. I would like to see more research on that though.

    2. Sgt Frank Drebin
      • 7 Years
      3 years, 5 months ago

      *it doesn't mean that the chances for flipping heads are higher in the next throw.

  7. Ci Siamo
    • 8 Years
    3 years, 5 months ago

    Is it fair to say that if Dallas doesn't play then it will be Barry Douglas instead?

    Thinking of getting both on WC (4.5M and 3.9m respectively)

  8. SteJ
    • 4 Years
    3 years, 5 months ago

    Has anyone looked at mean reversion in fpl?
    Do players like DCL, who are currently free scoring, often drop off back to more expected levels and vice versa?

    1. FPL Virgin
      • Fantasy Football Scout Member
      • 7 Years
      3 years, 5 months ago

      DCLs goal conversion is currently at 40%. I would expect that to roughly half going forward.

    2. Sgt Frank Drebin
      • 7 Years
      3 years, 5 months ago

      If he susteined his goalscoring rate he would end the season on something like 50 goals so surely he will drop off back to the expected levels (there are many players I would expect to do that currently) but even his expected levels are high imho. Even if he ends up on 20 goals he should be a good value. Even in the terms of xG I would expect him to cool off, he's on 1.16 xG/90 minutes, Hardly anyone in the world posts over 1 xG per 90 minutes over the course of the season.

    3. SteJ
      • 4 Years
      3 years, 5 months ago

      Indeed, and it would be interesting to know if there os a pattern as tonwhen those dropoffs happen or not.

  9. Bobby_Styles
    • Fantasy Football Scout Member
    • 6 Years
    3 years, 5 months ago

    Great article on a great tool!!

  10. davidfromkent
    • Fantasy Football Scout Member
    • 10 Years
    3 years, 5 months ago

    Should be called the goals inevitable table, obviously when the stats are engaging the most active players, whether they score or not, eventually they are going to return after all they are the players most involved in the game tend to be the ones creating and/or taking chances. Statistics are all based on historical outcomes and as with any area of life where they are employed the data can be used in many ways to back a prediction/argument, hence i never have and never will use stats.

    1. davidfromkent
      • Fantasy Football Scout Member
      • 10 Years
      3 years, 5 months ago

      and therefore* tend to be the ones......