Impact of non effective transfers and why the apparent transfer rate is a poor indicator
(This document was originally written in January 2015. The information outlined within it is still valid.)
There have been a significant increase in questions received at the site. These typically fall into three main categories:
- How do Wild Cards impact the price prediction accuracy?
- What formulae are used to take account of Wild Cards?
- Why don't you publish the transfer rate, isn't this a good indicator?
The purpose of this page is to answer each of the above questions.
The main fantasy site publishes transfer data for each player on the fantasy site. These are used to predict the potential price changes. Unfortunately, in August 2014, the main site complicated matters by adding in additional transfers that do not contribute to the price changes. The additional transfers are known as the non-effective transfers. There are a number of components, however, the main component (by far) are transfers by wild card.
To summarise the transfers published on the main site are given by:
Published Transfers = Effective Transfers + Non-Effective Transfers
To convert the 'Published Transfers' to only contain 'Effective Transfers' we multiply it by the following adjustment formula.
Adjustment Factor =
|Effective Transfers + Non-Effective Transfers
This is the main formula used by the price prediction algorithms. The adjustment factors (sometimes termed the 'transfer probabilities') are calculated for each player and then applied to the price change threshold such that each player now has his own unique threshold. These are updated according to the schedule given in the 'price calculations' link - Click Here
This formula has some interesting properties. If the Effective Transfers are equal to the Non Effective Transfers, but opposite in sign then the adjustment factor will become infinite. If the non-effective transfers are greater in size than the effective transfers and also opposite in sign, then the adjustment factor changes sign. These are important results and demonstrate why the apparent transfer rate is a poor indicator. To obtain the 'real' effective transfer rate from the 'apparent' transfer rate we would multiply it by the above adjustment factor. Reusing, the previous two examples we have shown that there will be times when the apparent transfer rate is close to zero but the real effective transfer rate is actually high. The second example shows that the two transfers rates could be opposte in sign, i.e. a player is gaining transfers when the the apparent transfer rate indicates it is falling. In a lot of cases, though, the two transfer rates will be of the same sign, but very different in magnitude. It is the size of these differences that make the predictions challenging during the transfer window.
These results are best demonstrated in the table below by using a number of real examples from gameweek 22 of the season 14/15. It is also important to note that during the transfer window of season 14/15 the effective transfers rates were lower than before Jan 1st 2015 due to the use of wild cards. Using data submitted to this site at the start of the gameweek 22 (Jan 17th 2015), 80.3% of teams had played their first wild card and 34.5% had played their second card. Reviewing teams that actively contribute to the transfer balance roughly 4.2% are ineffective teams, i.e. do not participate in transfers. Clearly these values will increase as more wild cards are activated. Also much higher values are quoted for non participating teams by other sites, but these are probably taken from the entire fantasy team population. Remember we are only interested in teams with transfers that contribute to the effective transfer balance and hence to the price changes!!!
- To aid comparison the effective and non-effective transfer rates are normalised to 10,000 teams.
- All values are those at the exact moment of the price change. Following a price change a players adjustment factor can change significantly.
- Henderson is the closest player to follow a singlarity trend in the last few days.
- Downing is an example of a player where the effective transfers were rising, whilst the published transfers rates indicated a player who was being sold.
- Hamer was an extremely slow riser with only eight in every hundred transfers contributing to the overall transfer balance.
- The transfer balances for Moore and Adrian had not reset since the start of the season. These are examples of players where no initial baseline is available due to the changes by the main site in August. The prediction models were both predicting Moore would drop. However, due to the potential error in the baseline a 20% allowance was added to be safe. Very few players with unknown baselines remain. A similar story for Adrian.
- Cazerola is an example of a player where the real transfer rate is significantly greater than the apparent rate.
A key point to note is the wide spread of adjustment values. A quick review, as I write this, indicates a current adjustment spread of 14 to -20, with the majority of players between between 1 and -1.
Determining the actual values for the non-effective transfers on the live system is the tricky part. As shown above determining the volume of the non effective transfers is key to making accurate predictions - but that will be covered in 'Part 3'. (Hint .... pattern matching, differentiation and a small amount of quantum mechanics theory is used.)
Please keep the wild card information flowing - especially now during the transfer window. Hopefully the above discussion helps you understand how important this information is to maintaining the prediction accuracy of the site - Click Here to Assist
Donate to this website
The continued operation of this website together with the supporting analyses are funded through the 'Adverts' and the 'Donate' facility.
Thank you !!!