There are two elements to this: multiple transfers made in one go, and transfers spread out across a gameweek. At present, we support neither case.
Multiple transfer support is high on our wishlist. Your donations allow us to spend more time on the project. In the meantime, follow us on Twitter, where we post multiple predictions for high-profile managers.
We usually expect the correct transfer to exist somewhere in the list of predictions, even if it isn't at the top of the list. There is one case, however, where our system is unsuccessful.
We check each team roughly once per hour. Player price changes occur at roughly 1am UTC daily. With that in mind, consider the following scenario, where the events in bold are those seen by Kammy:
In the scenario above, we are unaware of the price changes that occurred shortly after the transfer. We wrongly make our predictions based on a 0.2 team value increase, rather than a 0.1 decrease followed by a 0.3 increase due to price rises. This results in false predictions.
If we can detect that our predictions make no sense, we discard them. This explains why some teams will have no predictions despite only one transfer having been made.
We check each team roughly once per hour. Player price changes occur daily at roughly 1am UTC. We expect to see these changes reflected in team values by 4am latest. If, after 4am, we observe a change in team value that cannot be explained by a transfer, we deduce that a chip has been activated.
Due to limited bandwidth, and a recent increase in tracked teams, we sometimes fail to check all teams quickly enough in order to detect their price changes by 4am. This causes false positives in chip detection.
This is a well-understood issue that we will aim to address.
After each day's price changes, we prune our predictions to those that still make sense with the new team value. After a second transfer is made, we're unable to carry out this pruning (at least until predictions for multiple transfers are supported).
In the interest of providing high-quality predictions, we hide predictions in scenarios where we're unable to continue testing them throughout the week.
In the interest of reducing the number of false positive chip activations, we occasionally remove low-ranked teams from the database. Get in touch if this happened to a team of interest to you, and we'll be able to add it back.
We are experiencing a large backlog of addition requests and appreciate your patience. If you're a donor, get in touch and we'll jump you up the queue.
At present it's only possible to view teams one-by-one by entering their ID. We're working on an improvement to this, which will allow you to see all teams in a league in a single view.
We take numbers deliberately made available to us by the game and make predictions based on these. This, in our view, is fair game. The data we use can be in found plain sight for all to see—we aren't hacking accounts, and we aren't using data hidden away from ordinary players.
FPL Statistics reads data from the FPL website and derives price changes predictions from it. As we see it, we're not doing much different.
At least one website offers transfer predictions as part of their paid service, and has done for some time.