Research On Hattrick's Youth Academy Scouting Pools
Research On Hattrick's Youth Academy Scouting Pools
Research on Hattrick’s Youth Academy Scouting Pools
To gain good young players you can start using the Youth Academy. You can hire (up to three) scouts
looking for talents / your future stars. For each scout you can choose in which region and for what kind of
player type he should look for. Each region has at least one player pool, people assume that there is a pool
for each type of player (goalkeeper, defender, winger, midfielder, forwarder) and region, in which the scouts
are looking for talents. You can also search for any type of player a former research has shown, that the
chance of getting a player of a specific type is (slightly) higher when search for this type. Since players got
pulled from the pool into the youth squads Hattrick has to refill them regularly.
HTTools with Dantekavala and hattrick youthclub with Mackshot have analyzed 645,129 youth players to
identify trends, distributions and frequencies of talented players. We divided the research in three parts. In
the first one, we studied the talent of youth players in general through the last years. In the second part, we
examined if the point of time of scouting a youth player has an important impact on the talent and finally, we
searched for trends in order to see if pool refills are important on scouting.
We would like to note that this is not a guide that one should follow with closed eyes. Since it is a statistical
analysis based on historical data it is likely that if many Hattrick managers change their behavior (due to this
research) the results of a repetition of this study would be different.
Frequency of talented Youth Players
We grouped the players by their sum of skill points (keeper, defending, playmaking, winger, passing, scoring
and set pieces)when the player got promoted from youth to senior squad. For example, a player with as sum
of 17 skill points will be in the group of 1620.
The graph fits to a RightSkewed Normal distribution which seems logical in order to have more bad and
average than super talent players in the database. Additionally we have to take into account that not all
managers can train all of their youth players to perfection.
Groups Frequency
05 0
610 0
1115 13478
1620 68124
2125 331514
2630 213722
3135 17954
3640 329
4145 14
4650 0
5155 0
Importance of the point of time for scouting
Moreover, we have studied the importance of the point of time of scouting players / calling the scouts. We
split the dataset into groups with an interval of 2 hours grouped by weekday (Monday to Sunday). We
calculated the sum, the frequency and finally the mean and standard deviation of each group. One can easily
see that managers call their scouts more often on weekends than on weekdays, which is pretty reasonable
since senior matches take place on the weekend and thus most users are loggedin anyway. Additionally,
managers can call their scouts only once a week and this blockage is reset by Hattrick shortly before the
weekend starts. However, the gap in the sum of skill points on weekends is not higher than 0 to 1.0 skill point
from weekdays. Examining also the standard deviation and the coefficient of variation it seems that the
lowest numbers are at 6:00 to 8:00 every day. This means that during this period, all youth players are
relatively similar with regards to their skill points sum. There aren’t a lot of bad, average and talent players at
this period. In specific, you will only find either a lot of bad or a lot of talented players. Combining this with the
graph of average sums of skill points, we observe a peak every day that time.
This means that (based on this dataset and data analysis) you have a slightly higher probability to
scout a talented player between 6:00 and 8:00 in the morning.
Looking for trends
Using an algorithm for searching for periodicities in data series based on systematic sampling procedures,
we examined if there is any important trend of talented youth players after youth player pools have been
refilled. The algorithm split the dataset in all possible sublists, calculate the mean and standard deviation in
every sublist and check if there is any case that the data seems to be duplicated. If there is a part of the
graph that is repeated, we should have find a trend in the time serie. However, there isn’t anything that could
make us conclude in a trend for the time intervals, thus we believe that the pool refills are dynamic and they
depend from the current available size.
Conclusion and Future Research
In general, there seems to be no certain point of time when all pool get refilled. Thus we form the thesis that
Hattrick refills a pool when the filling level has fallen below a certain threshold (this check may be
implemented as postscouting hook or is called, for example, every hour). We want to note, that to confirm or
reject this thesis it would be helpful to analyze if the player quality differs for any point of time for a single
region or country.