Our Moneyball approach to the Fantasy EPL (team_id: 2057677)

If this is the first time you land on one of my Fantasy EPL Blogs, you might want to check out Part1, Part2, Part3, Part5, and Part9 first to get familiar with our overall approach and the improvements we’ve made over time. My partner in crime for this project has been Andrew Sproul, who just like me has lots of passion for Data Science, ML / AI and the Fantasy EPL.
What is new this week?
After getting punished by the Fantasy Gods for yet another week, we decided to introduce a new comforting feature, which is mostly designed to make people like me and Andrew feel good about their terrible GW Stats 🙂
Top10 Fantasy Users’ Stats – Each week we will be plotting how the Top 10 players in the world did that week, including the AVG, Min, and Max score of the Top10 players. This can be used as a meter for our own performance, so if we had a terrible week, but our score was still close to how the Top10 people did, then we can at least feel better about ourselves 🙂 Please, enjoy your first serving of "comfort food" below…(You can thank us for this act of kindness 😛 by giving the article a few claps and shares – it can help push the Blog to the trending section, which helps it reach a larger audience. Thanks in advance!)
Top 10 Players in the World GW9 Stats

So if you scored close to 44pts this week, you can actually feel really good about yourself since this is the AVG Score of the Top10 players in the world. And if you scored 32pts like us 🙁 , just know that there is still one player in the Top10 that scored less points than you – 25pts!
Now that we got you feeling warm and fuzzy, it’s time to face the music of another tough GW9 and look at all the other interesting stats.
Top100 EPL players optimal blended team
While we’re on the topic of the Top 100 players in the world let’s look at who were the Top10 selected players on their teams by position and the most preferred team formations for GW9.
Most Selected Goalkeepers by Top100

Most Selected Defenders by Top100

Most Selected Midfielders by Top100

Most Selected Strikers by Top100

We used this data and the current player prices to create the team below which is a blend of the most selected players by Top100:

Not a bad looking team with 2.1M left in the bank to play around with and test different configurations.
Most selected Team Formation by Top 100 Fantasy Users
Below you can see the formations that the top players picked last week. Seems like most top players opt for a 3–4–3 formation, likely trying to optimize the total points gained from their attacking players, who typically have a higher chance of getting assists, goals, and bonus points overall.

GW9 Team Performance Recap and Overall Stats
We thought that it couldn’t get much worse for us than last GW, but apparently – it can…As you can see below we ended up scoring only 32pts – below the AVG for the week of 37pts. We could have actually done much better, but 4 out of 5 defenders were mysteriously injured and so we only ended up playing with 1 defender, and a team of 9 players total. Given that and the fact that we picked a bad captain (AGAIN!?), we feel like we did OK. At least our investment in Vardy paid off for now 🙂

GW8 Overall Stats to Inform our GW9 Picks
Let’s start with the Adjusted Opponent Difficulty Rating (FDR) for the next three game-weeks below:

It looks like this week we have a lot of teams with a difficult schedule coming up – Villa, Southampton, Liverpool, Crystal Palace, and Spurs. The way our algorithm is written, it will not be recommending buying players from these teams this week. Teams with relatively easy schedules (if there can be such a thing in the EPL :P) are Bournemouth, Man Utd, Norwich, Chelsea and West Ham, so getting players from these teams could pay off in over the next three 2–3 games.
Latest Injury Updates Oct 24th, 11am EST
The data below comes from an independent site that updates latest injuries much more frequently than the Fantasy Website:

Let’s look at the top players by ROI and by the pts_per_90min Stats Next:
Goalkeepers Top 10 by ROI

Goalkeepers Top 10 by pts_per_90min

Defenders Top 10 by ROI

Defenders Top 10 by pts_per_90min

Midfielders Top 10 by ROI

Midfielders Top 10 by pts_per_90min

Strikers Top 10 by ROI

Strikers Top 10 by ROI pts_per_90min

GW9 Algorithmic Picks
Reminder that our algorithm takes in adjusted budget per team formation and tries to maximize spend per position for the main 11 players, and then get you subs which are good ROI value. Most weeks we have been playing a 3–4–3 formation and now that we know that most of Top100 players in the world pick the same formation, we’d be even more inclined to keep using it going forward. Based on current ROI scores, filtering out any team with AVG opponent difficulty (FDR ≥ 3.15) over next three games and removing injured players from the list of available picks, our algorithm picked the team below as currently optimal for spending the full $100MM budget:

We plugged in Mane in there, even though Liverpool’s FDR is high this week, because we had too much budget left over and didn’t want to leave a lot of extra money on the table. The Algorithm’s original choice was Adama Traore for Mane if you’re curious, with about $6M leftover.
As far as our team is concerned, we decided to wait until Friday to decide what subs to make, given the Europe games later this week. We already got burned from 3–4 unexpected injuries last week, so we want to try to at least have 11 healthy players this week and not just 9 like last week.
Team Stats
Looking at Best / Worst defense and offense can be used in a few different ways – for example, if a best offense team is playing a worst defense team, you might want to captain your attacking mid or striker. Also, you might want to give preference to DF and GK from the best defensive teams when looking at the algorithmic recommendations for these positions.
Best 7 Defenses

Worst 7 Defenses

Best 7 Offenses

Worst 7 Offenses

Cumulative Team ROI Stats
Below you can see the teams, sorted by cumulative player ROI. Note that an active player is any player that has played at least 25% of the total possible game time. For example, total possible minutes = num_weeks 90 mins = 9 90 = 810. So, we are counting any player that has played a total of at least 810/4 = 202.5 minutes as active on that team.

The top of the chart is starting to look very interesting with Liecester, Sheffield Utd, Wolves, Burnley, West Ham and Villa. As you can see all of these teams use only 13–14 core players, and are so far over-performing against expectations, which makes getting players from these teams a good investment in terms of long term expected ROI.
The bottom half has some usual suspects, but also teams such as Spurs, Everton, Man Utd, Arsenal, and Chelsea mainly due to higher player rotation 15–16 on AVG and the teams underperforming with lots of overpriced players, which don’t seem to be a good investment so far.
Final Thoughts:
Instead of "Final Thoughts" today we will end with "Final Prayers". We pray to the Fantasy Gods that they show us some mercy this week and bless us with a full squad of 11 healthy players, so we can at least maximize our expectations and end our bad run. Also, we wouldn’t be upset if they bless our captain this week with extra speed, strength and shooting accuracy 🙂
As always – thank you for reading and good luck this weekend!