BOTS VS EXPERTS: SELECTING THE BEST TEAM FOR THE WEEK
Good evening and welcome to a new series I
decided to work out on.
Choosing a team for a game week can be
hard. Sentiments and personal preferences can cloud ones judgement. There have
been many FPL sites that try to predict the team for the week but there is
usually a large discrepancy.
My plan in doing this article is to compare
human analysis (totw from experts on different sites) to AI analysis (totw from
bots on different sites).
Now these sites are expected to churn out
different teams every week irrespective of transfer limitations. The difference
is that the experts bring their teams based on strict FPL formation and price
while the bots simply bring out the best set of players from their algorithm
projections.
For this study, I'll be using 5 sites. 3
have bots (fantasy football fix, first and assist) while 2 are human experts
(fantasy football scout and geek).
The picture below is an excel
representation of the data gotten. I didn't delve into stats. Just collated
results by sorting by next game week scores (for the bots) and getting the
teams for the experts.
There is room for 3 goalkeepers, 5
defenders, 7 midfielders and 5 forwards in the initial selection for the bots
and these are given rank 1-7 with 7 being the highest. The players with highest
rank points are then fitted into an FPL formation with a max cost of £86.
Substitutes aren't used, a captain is also selected using the same method.
Another modification made was to reduce the
number of defensive entrants from a team to 2 as most bots select based on
fixtures and can pick all defenders from a team.
Hope you enjoy this piece.
The column in yellow are the top picks based on
rankings by the 3 different bot sites (first 3 columns) and gives you a general
idea of the players that the auto bots in general think will perform. The next
two columns give the experts view from two popular sites. The column in orange
lists the players that were selected by both sites. There is no ranking for the
expert teams. Finally, the column in red collates the data from all 5 sites and
tries to give the best possible combination of teams from both the bots and the
experts. The last column is for players who for cost or maximum number of team
mates missed the cut.
The team cost is 86.1. Pretty high but since we removed
all human sentiment from the judgement and didn’t like the idea of “coverage”,
we had to make do with Amat (the cheapest player listed).
Grant(4.2) was selected by 2 experts and 1 bot to claim
our goal keeper slot. Rose(5.9), reid(5.0) and alonso(6.3) received nominations
from at least 3 sites, more than any other defenders and also came top in the
bots rankings. Sanchez(11.8) was picked by all the sites and he is joined in
midfield by hazard(10.4) and eriksen(8.5) who received nominations from bots
and experts. Up top, kane is the second player to get a pick from all 5 sites
and gets captain nomination from 3 of them. Costa(10.9) and defoe(7.8) complete
the line up.
Feedback is needed on the article so I know if its fun
or useful at all.
Comments
Post a Comment