Many thanks to friend of the site @SoccerSaber for this excellent inght piece into the bonus point system of the FPL. You can read more of his great work at http://soccersaber.wordpress.com
Guest Post: A little something on the new FPL Bonus Points System (BPS)
As any regular on an FPL forum well knows, the new FPL bonus points system (BPS) has been quite the conundrum for fantasy players this season. Unlike the previous incarnation which was calculated by EA Sports behind closed doors, this season’s bonus points are calculated using OptaStats with a publicly accessible formula. The results though have been unexpected, with goalkeepers and defenders factoring in bonus point equations at the expense of their attacking brethren. In today’s post, I analyse the impact of the new BPS system to determine what, if any, real changes have occurred.
The first action was to limit the size of the player pool evaluated to something manageable (lots of manual calculations required in this analysis). Accordingly, I took a look at all non-goalkeepers with at least 3% ownership and 160 minutes played. Arbitrary, yes, but considering we are already dealing with ridiculously small sample sizes I felt it prudent to ensure that the players analysed had seen the pitch a fair amount. Since it’s my study, I get to call the shots; if you fervently disagree I encourage you to start your own blog!
The second step in the analysis was to determine the component parts of each player’s BPS calculations. As stated above, the BPS formula is publicly available and should be replicable for anyone with Excel and access to OptaStats. Note the “should.” In reality, I have found small discrepancies between FPL and my OptaStats source (the superb Fantasy Football Scout). Because of this, I have decided to break the components into two groups, those accessible on FPL’s website and those not. This simplifies calculations and removes any differences in Opta data sources. Those accessible from FPL’s website include appearances, minutes played, goals, assists, clean sheets, penalties missed, own goals, yellow cards, and red cards. Everything else is grouped together into what I have termed the “Opta Component.”
Once we have the components of the BPS system we should be able to easily determine how the system favors each position group, right? Not so fast. Complicating matters is that the start of this EPL season has seen a historically low run of goals. We are trending over 0.8 goals per game below anything the premier league has ever seen. While this may continue, historical data suggests otherwise meaning that we should probably not rely on this season’s attacking performance to gauge potential BPS output going forward. This is especially true for midfielders, who, as a group, have been putrid thus far. To combat this, I have taken the attacking totals from last year for my player pool and assume they will do the same. That point here is that I am trying to simulate a “typical” EPL season rather than the low-scoring league we’ve seen thus far. I am sure that you, my astute reader, have already realized that this means that all new boys are eliminated from the analysis. You are 100 % correct. Nonetheless, I feel this is the appropriate approach.
The next step was to combine the Opta Component with the BPS rate based upon last season’s attack profile for every player in the analysis. This should result in a BPS figure that compares with typical premier league scoring, with obvious sample size caveats applied. The result is a set of interesting data that I feel provides concrete evidence of what we can expect from BPS going forward.
The final step was to compare these projected BPS rates with last year’s EA Sports PPI (ESP) rates. We need this data to have a concrete understanding of what happened last year rather than rely upon anecdotal evidence. With both sets of data, we should have little difficulty identifying what, if anything, has changed vis-a-vis FPL bonus points.
Maybe not too surprising, but the results clearly show that defenders are now much more likely to score bonus points that in last season. Similarly, as expected, midfielders seem to be the primary victims of the change, with forwards minimally impacted. Let’s get to the data.
Notes: % Change represents the change in portion of total BPS/EPS for all three positions. In other words, defenders now get 6.0% more of the total bonus pie than they did last season. Also, BPS does not include appearance points as those are fairly rigid across all positions.
Not at all shocking to see that defenders are the clear winners this year. Even factoring in last year’s attack numbers, the average defender beats the average midfielder and attacker in the evaluation pool. What may be surprising is that forwards, not midfielders, fell most in proportion to last year’s data. This season, forwards are not the clear bonus magnets they have been. That much is clear. Yet, forwards are still right with defenders. Midfielders, on the other hand, are far behind. I am guessing this was not expected from the powers that be over at FPL.
While these averages tell us something, that aren’t yet conclusive. It might make no difference if the BPS system eliminated all terrible bonus defenders if, in fact, the majority of defenders still score less than the majority of forwards and midfielders. To analyse that I have a set of graphs for review.
These two graphs show the distribution of the player sample based on 2012/3 EPS/90 and projected BPS/90 data. A few notes before we get into further discussion:
- Hopefully it’s implicit in the graphs, but the x-axis represents the BPS/ESP score moving from smaller to larger. The y-axis is the frequency at which players that fall within that range.
- I have tried to apply both sets of data to normal (Gaussian) distributions. It is quite possible (probable?) that the data comes nowhere close to forming a normal distribution.
- The distribution in EPS/90 for forwards last season was massively skewed by two heavyweights, Robin van Persie and Daniel Sturridge. They both garnered over 21.75 ESP/90. No one else in this study garnered as much as 18.10. As such, that distribution is skewed to all Hell.
- The multiple assumptions required to create these graphs means that the statistical error in them is HUGE. I cannot stress this enough. What that means is that one should look at these graphs for trends only not as models for expected values.
So what do the graphs tell us? The first thing that jumped out to me was the relative consistency in distribution among BPS in comparison to ESP. You can see that the graphs of BPS values all seem to have about the same width, indicating a fairly consistent distribution across the position groups. On the opposite end we have the ESP graph in which all three positions have different distribution patterns. To me, that means that we should no longer see the ridiculous outliers of bonus point allocations that we’ve seen in the past. I suspect things will be distributed much more socialistic this season. Our Scandinavian friends would be proud!
The second thing to notice is that the BPS graphs overlap quite a bit. This indicates that all three positions are all somewhat likely to get bonus points in a game. Certainly the data supports the fact that midfielders are somewhat punished, but not so much that they have almost no shot at grabbing bonus points. If they score, they’ll be in the running, clean sheet be dammed (for evidence, see Yaya Touré’s five bonus points in games this season in which his team kept a clean sheet).
The bonus point system has materially changed. Forwards are no longer the kings of bonus points, with defenders faring worse than midfielders. Now, the average defender will do marginally better than the average forward with the average midfielder falling well back. Yet, when it comes to excellent performances, the midfielder has a chance at top bonus points, even in games in which his team keeps a clean sheet.
Here I will offer a couple of opinions that aren’t necessarily proven by the data. Take them for what they are.
- Defenders will have a great shot of racking up points in 0-0, 1-0 and even 2-0 games. In a 0-0 game, it is virtually impossible for a non-defender/goalkeeper to grab a bonus point. In 1-0 games, I would suspect that only a forward scoring and having a good bonus point game will wrestle the top score from a defender (although a forward who scores in a 1-0 game should grab some bonus points). In 2-0 games, it comes down to if the scoring and assisting is concentrated on a few players. It’s conceivable a defender could grab the top BPS score in a 2-0 game even if he doesn’t score.
- While defenders as a whole are better, it’s really centre backs that benefit from the new BPS system. Fullbacks haven’t shown the same BPS potential, although they still score much better than last season.
- Forwards who score once will be in good position to grab a bonus point. A goal and assist and they will be in great position. Forwards will actually be helped if there is no clean sheet as that virtually eliminates a non-scoring defender from consideration.
- Midfielders should really be broken into two groups, attacking mids and holding mids. Attacking mids have much lower Opta Component scores and thus need to score in order to grab high BPS’s. Holding mids, on the other hand, score as well as fullbacks and some centerbacks, making their threshold to bonus points much lower. Touré’s grabbing of three bonus points in a 2-0 game with just one goal is an example of this phenomenon.
- Adding everything up, I think high assist midfielders are probably overpriced this season. Players such as Silva, Coutinho and most likely Ozil may rack up assists but without a similar number of goals. Accordingly, they will probably struggle to justify their prices. Unfortunate, as two of those players are on my roster! It’s probably more prudent to go with players with more goal threat such as Michu, Nolan, and Mirallas.
Guest Post by @SoccerSaber