Alternate Text

Update 18th Apr 2017

Currently there are quite a number of wild cards in play - however, it's not yet reached onslaught proportions.

Update 8th Mar 2017

Have now got enough information to start coding the new price change rules for players from blanking teams. I just need to find the time to do them !!! The rules for double game week players will have to wait until more data is available.

Update 7th Mar 2017

My guess at a rule for players from blanking teams was close last night. I fudged the displayed Tar% values for the blanking players in line with the guess. Although I correctly predicted Defoe to drop in value, I would have preferred him to fall tonight. Consequently the analysis continues. Also guessing that this change is a direct replacement to the rule used over the last two seasons that throttled back players with exceedingly high transfer rates. I will not make any permanent code or database changes until the new rule conditions are fully understood.

Update 6th Mar 2017

Almost certain now that new price change rules have been implemented for this season. These are currently affecting the transfers for both blank and double game week players. I don't want to make any code or database changes until the new rule conditions are fully understood. Hopefully some illumination may occur from the price changes overnight. Currently, the new rules are impacting players from teams who do not have a game during the next game week. Please treat any player who does not have a game next week and is currently predicted to rise / fall with caution. You have been warned!!!

Update 1st Mar 2017

Investigating a new rule change for this season that impacts players with a blank or double game week.

Update 25th Feb 2017

Another quick note on the impacts of wild cards. Whenever the total volume of wild card transfers for a players exceeds the normal (i.e. effective) transfers for that player, then predictions for that player become almost impossible. In effect, accuracy reduces and overly enthusiastic predictions result for that player. The loss in accuracy is a direct consequence of the statistical mathematics utilised in making the price predictions.

These inaccuracies are i) removed whenever a players transfers are reset (e.g. following a price change for that player, however if more wild transfers are made then the inaccuracies quickly build up again) and ii) reduced (but not completely eliminated) following a game week change.

See this link to understand the mathematical impact of wild cards.

Update 5th Feb 2017

Permanent fix for the 'Statistics data' now in place.

Update 2nd Feb 2017

For the last two days there have been significant issues with data on the 'Statistics' page. These issues are now fully resolved and the page is now fully up to date. A permanent solution is to be developed. The current fix requires a significant degree of support effort to maintain. Work is on-going and a permanent solution will be in place soon.

Apologies if the price predictions over the last two days have impacted anyone. The focus over the last two days was on resolving the statistics issues. No significant errors in the price change predictions occured. However, quite a number of false (i.e. plain stupid) predictions for the price drops were made. The key factor is that there is a strong correlation (Pearson correlation coefficients between 0.73 and 0.97) between wild card spend patterns and a players most recent statistical performance. For example, last night Anichebe was no where near a price drop - however the systems indicated Anichebe was due to drop. These overly enthusiastic predictions were due to the absense of up to date accurate statistical data.

Update 5th Jan 2017

A wild card onslaught is currently underway. Expect some overly enthusiastic predictions over the next few days. The prediction accuracy will consequently reduce whilst the wild card volumes are high.

Update 13th Dec 2016

Off to the Xmas Markets today. Consequently I will be running the game week end / start remotely (What could possibly go wrong? At least its not Friday 13th). Please note that the target% jumps that occur at a game week end, (as the base lines are adjusted), will commence a lot earlier in the day. This is to minimise the work down remotely. The main players impacted are those who have already risen / fallen in the current game week. The screens will get somewhat messy as the new game week commences - apologies for that.

One final caveat to be aware of is that I would expect the individual accuracies to reduce tonight. This is due to the close proximity of the timing of the game week start and tonights price changes. You have been warned!!!

Update 13th Nov 2016

A quick comment on the accuracy values shown on the price predictions page.

These values show the average accuracy following a price change for all the predictions made during the current season.

It is important to note that these values do not reflect the expected accuracy of any individual future predictions. The individual accuracy of any single prediction will vary signifcantly during the season - see the price predictions error page for the current range.

Factors that adversely impact accuracy include: i) high rates of ineffective transfers (e.g. wild cards; ii) low transfer volumes (e.g. during international breaks; iii) predictions made on the day that includes the late start of a new game week; and iv) periods of exceedingly low wild card usage. The final factor is interesting and probably shows the need to retune the four predictions engines that compete to show the final prediction values.

Update 7th Nov 2016

A reminder that I recently added a transfer trend / rate of change column on the far right of the price predictions table. This utilises the effective transfers since the last price change (or game week start) together with the net effective transfers in the last hour. If you cannot see this column then it is likely that you are not using one of the optimised views - please note these views require scripts to be running. This is easily checked by clicking on the logo in the top left hand corner which will select the best view for your device.

An important caveat is that due to threshold changes a player with a positive transfer rate can still drop in price (and vica versa). This is a common occurance for low ownership players and is due to the threshold change being greater than the net transfer rate for that player - you have been warned!!!

Update 15th Oct 2016

Hopefully the plagiarisation of the fplstatistics site content is now resolved.

17 out of the Top 50 managers have gone for Aguero as captain. Sanchez (15), Costa (12), Benteke (3) and de Bruyne, Walcott, Defoe(1).

Update 14th Oct 2016

Unfortunately a separate site has decided to republish the price change information from this site without permission. This has meant that resources have had to be redirected to combat this threat. This is taking resources away from ensuring that the predictions are optimised.

It is important to note that the continued operation of this site is now at risk!!!

Update 9th Oct 2016

There will be a short outage tonight, sometime after 9pm, whilst routine maintenance work is undertaken on the web server. At least the international break has some use!!!

Update 27th Sept 2016

The team selection module for the 'Stats' page is now live on PCs / Laptopa. The functionality was requested by over 300 regular users of this site - so I thought I'd better get it implemented!!!

Additional work on the 'stats' page includes:.

  • Release equivalent functionaility for mobiles, smart phones and tablets.
  • Ability to hold multiple team ids in your account
  • Seperate 'tab' on 'Stats' page to show additional statistics.

Update 26th Sept 2016

The team selection module for the 'Stats' page will be released tomorrow.

Update 14th Sept 2016

Price change prediction systems are finally out of learning mode - Yippee!!!

Work has commenced on adding a team selection module and additional statistics to the 'Stats' page - will be a couple of weeks before its ready.

Update 1st Sept 2016

Apologies for the system failing to update the price changes overnight. The 'Windows 10' machine that provides the twice hourly updates decided to update it self overnight. I think I will revert that machine back to windows 7!!!

Update 30th Aug 2016

Solution found for the low ownership players. Will take time to bed in and the system needs to 'learn' how to make the adjustments.

Update 28th Aug 2016

I'm experimenting with low ownership players today. Ignore any 'silly' tar% drop values. Hopefully there will be none but advising in case something goes wrong.

Update 27th Aug 2016

I'm currently getting the system ready for the next game week. There are a number of short term impacts:

i) For a short period a few players with low ownership will display some extermely large erroneous target% values. These will be corrected shortly after the new game starts and should be fully cleared before 13:00 GMT.

ii) Expect some major Target% jumps as the new Game Week starts and the baselines and adjustment factors are updated for each player.

To maintain the accuracy of the price predictions requires information from you whenever you play your wild card - Please Click Here.

Update 16th Aug 2016

Top 50 page is now up and running

Coutinho should now rise in price - unless i) Wild Cards are interfering; ii) There is a new rule; iii) Price changes are not switched on, or iv) I've coded something wrong or set a parameter incorrectly! Remember the price change algorithms are currently in learning mode.

Because this is game week one price changes can occur at any time - day or night (or maybe not at all)!!!

Finally - For those worrying about the purchase price of a player not being visible. I can see it. So I guess it will become visible later on the main site.

Update 14th Aug 2016

New code, debugging and testing appears complete for the price changes - that is until I find some new issues lurking away waiting to trip me up!!! I will now be moving on to the Top 50 and Stats tools. Remember that the KPI stats will not be optimal until gameweek four or later. The same applies to other sites unless thay have some advanced statistical knowledge that I have missed (or maybe a time machine hidden away.)

The price change prediction algorithms are currently in learning mode. The implications are that the prediction accuracy will be reduced until a number of price changes have occurred (both up and down). Also the main site has hidden certain parameters, consequently, I expect the price prediction accuracy of players with a low ownership to now reduce (most users will probably not care - but I do!). I have created pseudo replacement elements - but only time will tell how effective these are.

Last season the first price changes occured at 7:50am not at 2:20am! - so basically they can occur any time!!!

Update 13th Aug 2016

Good Luck to everyone for the next nine months!!! Lets hope that the fantasy fairies bring you all lots of points.

It's going to be a busy day later this afternoon - coding wise!!!

We have plans to release new pages on the site - but these will be a few weeks away. As usual all content on this site will continue to be free to access. There are no plans to charge for any information held on this site. All funding for this site is generated through the 'Adverts' and the 'Donate' facility.

Update 9th Aug 2016

All work on the site is progressing well - some minor development and testing to complete. However, it will be a busy day this Saturday!!!

Update 28th July 2016

A means to replicate the missing data elements has been identified and will be coded over the next few days.

Update 27th July 2016

This looks like a great guide to becoming a 'top 1000' fantasy player. Click Here. (PS I'm not the author)

Update 26th July 2016

The web site was updated yesterday with data for the new fantasy season 2016/17. Updates are continuing on the site, but the focus will shortly change back to the desktop systems.

Update 21st July 2016

The updates for the new season have commenced. The main Fantasy Site is now Live, but completely changed. The means to extract the data from the new site has now been identified and coded. The bad news is that the new site changes requires a complete rewrite of the desktop support systems. (The web site systems are unaffected). There are now three missing data elements. Two of these are easily replicated. The third will affect player price changes with a low ownership (the majority of users will not care - but I do so need to create a pseudo statistical element). Data for the next season will be visible later this week. Any errors reported by the web site over the next few weeks will be due to the updates - if you try again after approximately waiting one minute all should be OK.

All content on this site will continue to be free to access. There are no plans to charge for any information held on this site. All funding for this site is generated through the 'Adverts' and the 'Donate' facility.

Update 5th May 2016

The final wild card onslaught for the season is gaining significant momentum. Expect some overly enthusiastic predictions whilst an understanding is gained for the increased non effective transfers volumes together with transient delays due to an increase in point hits. There are additional complications to be taken into account arising from the Ultimate game, the influence of the max, min (cost and points), together with the average leagues.

The predictions will improve (hopefully!!!) once the Liverpool game completes and the volume of effective transfers starts to rise. At the moment we are in a period of low effective and high ineffective transfers.

Update 11th January 2016

I am currently experimenting with a new rule. This experiment only impacts Aké. If he drops today/tomorrow then the rule is correct. Treat the current Aké predictions with care!!!

Update 29th December 2015

Tomorrow 30th December I will be on a long haul flight - total journey time approximately 36 hours. Consequently the price prediction screens may (i.e. will) become somewhat untidy until I have a fast internet connection and have overcome jet lag.

Update 1st October 2015

The automated team (i.e. the fantasy team selected by the computer using the stats page) is currently in position 1,666 overall.

Update 19th September 2015

All baseline adjustments complete. Probalistic run commenced.

Given Pedro's miraculous recovery from injury - he will not drop in price tonight!!!

Update 12th September 2015

13:30 pm: Adjustments complete. It's looking like the price changes will now take place at the usual time.

11:44 am: No delayed price changes have been applied by the main site. I will be making the usual game week start base line adjustments on the main site restart. These adjustments will impact the predictions for Ayew, Gomis and Ibe. I will adjust they back later this afternoon.

8:00 am: The main fantasy site was unavailable, for approximately an hour, from just before midnight last night.

The main consequence is that no price changes took place overnight.

I’m not running the morning cleanup processes in case they implement the price changes before the new game week starts. The consequences of this are that the price prediction screens will look very messy today. If the main site does not run the price changes this morning then a further problem will occur when the base line adjustments are run at the start of the new game week. I will issue an update on this later today.

Update 2nd September 2015

Starting to suspect that Pedro's lock has not been removed. The unlock date for Pedro was very close to the international break (when all locks are usually extended). Clearly wild cards could well be impacting Pedro's transfers. Treat Pedro's potential to rise with care!!!

Update 27th August 2015

Server Migration Complete. No need to do anything special to access the site. If any graphs don't work on a Windows PC press Cntrl and F5 keys together.

Update 26th August 2015

Awaiting the server build before I can start the site migration today. Will advise when migration tests complete, goes live and how best to access. I suspect the quickest route will be through the .com url as opposed to the Will advise by Twitter - see link directly below.

Other news is that I found a way to improve the accuracy even further - however, I cannot implement this change until the migration tests are fully complete and the DNS's are correctly pointing to the new web server, i.e. a few days away.

Update 25th August 2015

Migrating to a new server tomorrow (Weds 26th Aug). Will advise when live and how best to access.

Update 24th August 2015

Website is currently limping along very slowly. Due to the bandwidth issues some functionality is unavailable. We are aiming to migrate to a new VPS solution over the new few days. Apologies!!

Update 20th August 2015

Three concerns this morning. All big taking each in turn.

i) The main site has made a rule change (I think) - It will be next week before I can investigate this and implement a fourth price change algorithm (what fun!!!). Thank you to a number people around the world who raised this. What is interesting is that the systems are working out rules to get around this (hence why some of the target values are jumping around a lot!!) - but I really need to work out the correct rule.

ii) We got 46,483 individual users of the web page in an hour on the 15th August. The overall traffic to the website is increasing and using up the bandwidth. So basically, I'm going to have to move to a more expensive VPS package. (No !!!!!!!!!!!). As background to myself I only set this website up as a hobby. I'm also a pensioner. So have got to make some decisions next week.

iii) As stated previously there are a lot of wild cards in play. The systems have been quite accurate so far and identified the majority of the price changes (it got Oscar completely wrong though last night). However, in addition to getting the predictions correct it is also making a number of false predictions (that is what I meant by overly enthusiastic in my previous update.). This will probably get worse tonight and on Saturday morning at 2am I expect that it will reduce the accuracy somewhat. I just wanted everyone to be aware of this!!! You have been warned.

Update 17th August 2015

There appear to be a lot of wild cards in play - This will result in some overly enthusiastic price prediction alerts and a much greater number of price fallers.

Update 16th August 2015

A reminder to all that the price predictions algorithms are still in learning mode. However, the prediction accuracy so far has been quite good given the volume of wild cards in play. The probabilistic run for GW2 should complete later today.

Please keep the wild card information flowing. This is helping!!!.

The stats page will be updated once all the gameweek two matches complete. A reminder that the correlation accuracy for the stats page will not be optimised until at least four game weeks have completed.

Update 12th August 2015

The first price changes of the season were interesting - they changed at 7:50am not at 2:20am! The system correctly predicted all the price rises for last night. However, because the prices hadn't changed early this morning the system recalibrated assuming more teams were playing their wild card than originally estimated. Consequently when they finally changed at 7:50am the predictions were out.

Oh well a lesson learnt!

Update 10th August 2015

The price change prediction algorithms are currently in learning mode. The implications are that the prediction accuracy will be reduced until a number of price changes have occurred (both up and down).

Due to numerous requests I will populate the stats page sometime next week. A caveat - remember that the 'correlation accuracy' will not be optimised until the 4th / 5th gameweek.

Finally, remember that the next game week starts this Friday 14th Aug at 18:45 GMT.

Update 8th August 2015

The updates for the new season are complete and implemented. Some changes have been noted on the main site pre season. Will advise on the impacts (if any) once the season is properly underway. Learning mode has now been engaged.

To maintain the accuracy of the price predictions requires information from you whenever you play your wild card - Please Click Here.

Good luck to everyone !!!

Update 30th July 2015

The updates for the new season are almost complete. The main focus is on algorithm improvements for the price changes and statistics.

Important:During the early season use rather than the url to access this site.

All content on this site will continue to be free to access. There are no plans to charge for any information held on this site. All funding for this site is generated through the 'Adverts' and the 'Donate' facility.

We now also have an account on twitter. This will be used for important announcements.

Update 24th May 2015

Thank You !!!

A big thank you to all the users of this site over the past season.

The final prediction accuracies were 98.6% and 97.4% for the price rises and price falls respectively. The individual prediction errors over the last season (warts and all) can be viewed by clicking on this link.

The automated team (i.e. the fantasy team selected by the computer using the stats page) is currently in position 2836 overall. It would have been much higher (i.e. better) without the ocassional ill advised human intervention.

Finally some usage statistics for this site. The peak number of individual users of this site in an hour was ~22,500 and the peak average usage was 7,500 users per hour. If crawlers, bots and scrappers are included then these values are much higher. The global usage activity over the final week can be viewed by clicking on this link.

Again thank you for using this site and see you again next season.