NOTE: The passage below was added on the 6th of Jan 2006 and contains the current thinking on how price changes in the FPL.com game are calculated.
Beneath this is the original text of the opening post.

THE FISO GUIDE TO PRICE CHANGES

slarty's January update to malefice's initial post in the 'A FISO Guide to Determining Price Changes thread.

This guide is a product of the work being done in finding out the exact formulas used to determine price changes.

Phrases in italics indicate that more research is needed for confirmation, but they are usually working values which are valid in most cases. Once we get confirmation, the italics will be removed.

With that in mind, we proudly present,

The FISO Law of Price Change

Definition:

Net Transfers In (NTI) = Transfers In - Transfers Out
Net Transfers Out (NTO) = Transfers Out - Transfers In

Assumptions:

1. Price changes occur between 00:01 - 01:30 GMT everyday. This marks the start of a new day.
2. The price of a player will not change by more than 0.1 in a day.
3. The player in question is not injured, doubtful, ineligible or suspended, ie his "i" flag is grey
4. The player has not recently been transferred.
5. If a rise follows a fall in the same gameweek - normal rise thresholds apply.


Price Rise:

Player Gameweek Price Rise Threshold
This can be obtained by monitoring the threshold at which players' prices rise. However as this figure will always be over the threshold, equal attention should be paid to the NTI at which players' prices DO NOT RISE. The threshold will lie between the two.

A rough approximation (c.200?) of the price rise threshold may be obtained for any given number of teams in FPL using the formula Threshold=15,500-(0.005 x No. of teams)

1. The NTI for a player need to exceed a certain threshold before his price will rise. The threshold figure seems to be inversely proportional to the number of teams entered in FPL. It started at just over 11,000 mark and has decreased over the weeks. The current minimum NTI required lies in the range 9800-9900.

2. If the NTI is smaller than the threshold at the time of the price change, the figure is brought forward to the new day.

3. If the NTI is greater than or equal to the threshold, his price will rise, then the NTI is reset to zero and the count restarted.

4. The NTI required for subsequent price rises within a gameweek beyond the first is equal to n x threshold, where n = the nth price rise within that gameweek.

For example:

Player X
Threshold for price increase = 11,082

If the NTI = 15,000, his price will rise by 0.1 and his NTI will be reset to 0 for the new day. Therefore, he will need to get more than or equal to 11,082 transfers in for his price to rise in a subsequent gameweek, or 2 x 11,082 to rise again in the same gameweek.

If the NTI = 10,000, his price will remain unchanged and will rise if he gets more than or equal to 1,082 transfers in for the new day.


Price Drop:

1. The NTO for a player needs to exceed a certain threshold before his price will drop. The current figure we are looking at is 9% of the number of teams that have the player at the start of the new day (Due to a statistical error of +/-0.5% for the 'Teams selected by' value, the figure given here can only be used as a rough guide) NB The drop threshold appeared to have risen to 10.5% for a few weeks, but now seems to have reverted to 9%.

2. If the NTO is smaller than the threshold at the time of the price change, the figure is brought forward to the new day.

3. If the NTO is greater than or equal to the threshold, his price will drop, then the NTO is reset to zero and the count restarted.

4. The NTO required for subsequent price rises within a gameweek beyond the first is equal to n x threshold, where n = the nth price fall within that gameweek.

For example:

Player Y
Assuming 780,000 players at the start of the new day.
Teams selected by = 10% = 78,000 (10% of 780,000)
Threshold for price drop = 7020 (9% of 78,000)

If the NTO = 7,500, his price will drop by 0.1 and his NTO will be reset to 0 for the new day. Therefore, he will need to get more than or equal to 7,020 transfers out before his price will drop in subsequent gameweeks, or 14,040 (2 x 7,020 or 18%) to drop again in the same gameweek.

If the NTO = 5,000, his price will remain unchanged and will drop if he gets more than or equal to 2,020 transfers out for the new day.

5. Price drops for players whose price has previously risen
A player whose price has previously risen may still follow the rules above when it comes to dropping if his ownership is low enough. However players where the absolute value of the 9% threshold will be greater than the NTI threshold will fall at the negative NTI threshold.

eg Mellberg gw17 Price was �5.8, up from an initial �5.0
9% threshold for price drop = -19047

Rise threshold = c.10000
ABS(9% threshold) = ABS(-19047) = 19047

19047 is greater than the rise threshold of 10000 therefore price will fall at the negative of the rise threshold:
Actual drop figure = -10194

This continues until the ABS(9%) level is lower than the rise threshold or the player is back at his starting price, whichever is reached first.


Flagged Players:

1. Injured, doubtful, ineligible and suspended players will continue to rise or fall. The rise threshold remains as normal, but the fall threshold is increased to around 25%. Players with a prior price increase will fall at the 25% threshold OR 2.5 x negative rise threshold, whichever is lower (see rule 5 above).

2. When a player returns to fitness/eligibility (grey "i") the following conditions apply:

Only having been flagged for one day: Reset to 0 and count begins immediately

Flagged for more than one day: Transfers will not count for seven days, the price will not change during this limbo period. After 7 days the count is reset to 0 and transfers have their normal effect.



Transferred Players:

1. Transfers in and out will not count for seven days, the price will not change during this limbo period. After 7 days transfers have their normal effect with the count starting at 0.


Tips:

Software required
Any spreadsheet, though one compatible with recent versions of Excel will prove useful in other areas of this site
Here's a free excel-compatible spreadsheet for several platforms, though there are several other free ones available if you want to scout around.

Fields used
You may want to use the following, though it's very much up to yourself, there are probably better ways, and you don't need all of these fields if you're only tracking your own players.
T=Text N=Number F=Formula

T Position
T Name
T Team
N Initial price
N No. of teams in FPL (from your "My Team" page)
N Current Price
N Transfers in
N Transfers Out
F Difference: Transfers in - transfers out
N Percentage Ownership
N NTO/NTI at Last Price Change
F Current NTO/NTI: Difference - NTO/NTI at last change
F Percentage Ownership minus 0.5%
F Percentage Ownership as stated by FPL
F Percentage Ownership plus 0.49%
F 9% Fall level for low ownership
F 9% Fall level for medium ownership
F 9% Fall level for high ownership
F Current NTO divided by low ownership level to give high current percentage drop
F Current NTO divided by FPL ownership level to give medium current percentage drop
F Current NTO divided by high ownership level to give low current percentage drop
F NTI/NTO of current drop (only used when price changes)

In addition space under the text fields can be used to note the gameweek of a price change and any flags. Noting the players who have changed price in the current week makes it easier to spot that day's changes.

FPL Figures
1. Use the "Transfers" screen. "Leaders" and "Price Rise (Gameweek)" only does what it says on the tin, but selecting a team and "Price Rise (Gameweek)" will show you both the risers and fallers on one screen. Just hit the "i" button and note the figures for the players you're interested in. The only players whose figures it's important to get around the daily update time are those whose price has changed (so you have accurate count figures), though early morning c.7am GMT is a reasonable second best. Other players can be updated whenever you feel like it, once a week in the case of slow movers.
2. Use "Leaders" and "Transfers In (Gameweek)", AND "Leaders" and "Transfers Out (Gameweek)" daily to save yourself getting caught out by a fast mover.
3. Check for any change of flag status on a daily basis as any missed flag changes, especially when going TO grey, will mean your count is wrong. It is important for getting an accurate starting point when resetting a count to 0.
4. In a normal week, Friday night - Sunday/Monday are the times of greatest transfer activity.
Here's a graph reg produced:


THANKS TO ALL CONCERNED FOR GETTING US THIS FAR

NB
At the start of the 2006-07 season the transfers seem to have begun counting from 11.30am on 19th August (close of gw1 changes), but players prices were not permitted to change until the third night's update (am 22nd August).


Original Text from 'Can We Crack the Code?' - this thread's initial title.

To explain to those new to the game, an eternal enigma of FPL.com Fantasy Football is the way in which prices fluctuate as we managers buy and sell players.

Last season Padrock set off in search of the mystical formulae by which player prices are determined. When he returned, few recognised him, such had his search for The Philospher's Stone affected him.
Whilst his quest had brought him close to grasping the Holy Grail it remained tantalisingly just beyond his reach.

Brothers and Sisters, the time is once again ripe to journey into the unknown!

I'd like to suggest that we organise a panel of members to monitor player transfers once the new season is underway with the aim of establishing the mathamatical equations which FPL.com use.
I say 'equations' as, as far as I understand it, there is one formula which triggers a price rise of 0.1; one which causes a fall of 0.1; and one which applies soley to an injured player's price falling by 0.1 (eventually!).

I was wondering whether we could organise half a dozen members each of whom would monitor, say, ten players' Transfers In/Transfers Out daily for a week or so. The pooled information might then provide sufficient data for the likes of Padrock and Wyld to break the code?

Anyone interested?

P.S. If some kind soul can post a link to Padrock's original thread I'd be grateful.

How about we all use our wild card to perform a little experiment.

Gentlemen (and Ladies) synchronise your watches.

Wednesday 20th September 2006

14:00 - All members of FISO with a PL.com team log on to the websites entering their details.

14:10 After waiting ten minutes to ensure that every member has had a chance to log in/re log in/ wait for their boss to stop takling to them, every member takes out one of their cheapest midfield players and we move Sheffield United's 4.5m midfielder Nick Montgomery into that place(can be similar player if need be)

14:10 & 55 seconds All players hover over the second confirm tranfer button

14:11 EXACTLY At the very B of the 14:11 bang all members press the button, thus importing Montgomery in their squad.

14:12 onwards - We sit back and see what happens (watch his price rise?), constantly analysing the stats page for the moment of price change and any other changes.

Sign below if you are willing to participate.


1.Tort

Sorry to bring bad news Tort as it's a wonderful idea but there are two fatal flaws to that experiment:

1) At a hugely optimistic estimate we could get 200 FISOers taking part in the game. I'm sure it will take significantly more than 200 transfers in/out to trigger a price change. (Given that there will be around 1/2mil people taking part in total)

2) Although they claimed that price changes can happen at any time, last season they were only ever adjusted in the middle of the night (5am-ish?).

Bah!

Not worry back to the drawing board


I have noticed that on the Guardian "The Chairman" forum somebody had some price tracking software?

Is tracking the price change that important... surely most should be obvious?... *lack of knowledge showing*

The idea is that you can hopefully predict if/when the price will change.

Example:
It's Friday evening and you're planning your transfer for the weekend but you want to wait until Saturday morning for late team news. However you're right on your budget and if your target goes up 0.1 overnight your plans are up sh|t creek. If we had an expert that could look at the recent transfer activity stats for your player and predict whether a price rise is imminent, that would save you a whole lot of bother and frustration.

having not played this game before do the prices and percentage of teams selected by change much?

What Do We Already Know?

As I understand it (and I could well be wrong!) the prices are regulated like this:

Pre-season, we managers wheel & deal freely, choosing our squads, and the market remains static.
At the very start of GW1 the number of managers who have bought each player is noted. This figure - 'A' - then becomes the basis of that player's price changes.

When X% of A have sold the player or Y% of A have bought the player a decrease, or increase, of 0.1 occurs.

There may be further factors which prevent a player's price from plummeting to 3.2 or lower through mass exodus, or factors which increase/decrease depending on the player's current price, I just don't know!

I don't know about the %ages, Tort, but prices can rise by as much as 0.2 in a gameweek (0.3 may have occured, but in 3 seasons I've not seen it), and by over 2.0 in a season.
(Remember, you will only recover 50% 0f any price rise profit when you sell.)

here's the thread from the old forum at the start of last season by ferg where i reckon we made some good progress into said formula.
and another one

I'm flattered Flyman that you thought it was me who started the whole player price change debate but I'm afraid my intellect nor imagination don't stretch that far! I'm more of a 3 sentence bish bash bosh poster!

In any scientific study, would it not be illuminating to track closely the values of players added to the game once the season is underway? They will start from a very small %age-picked base (zero) and therefore could be subject to quite wild fluctuation compared to established players.

That's how I got some of the data nash, I followed Owen when he was added last season and he was popular enough to generate some worthwhile stats.

My apologies, Padrock. I was confused - but you did discover the 'Bonus Points' page that appears when the boni are released (many thanks!). I think that was the source of my confusion.

So, is anyone up for trying to work this out?

I think it is something to do with how many buy or sell a player

I know what to look for this season so I'll have a stab once the game starts and post some predictions before the prices change.

If the formula is the same as last year then I'll definitely crack it this year.

Hey Flyman, im up for a crack at it.

Are we going to track daily price changes for x no. of players each and then combine stats?

Yes, reg, that's what I thought.
I think we will need to know how may have bought each player when the season starts.
This information is thought to be integral to any player's subsequent price change, but it is not a stat made available throughout the season.
Consequently, we need to gather the data from as many players' (dynamic) profile pages as possible as swiftly as possible. Dividing the effort between us (Stage 1) will make that much easier and the more speedily it is done, the more accurate the data will be.

Once this info is pooled we enter Stage 2!

I suggest (and it is only a suggestion) that a core group of half a dozen or so FISO bods get together and agree to monitor specific players. We can decide who watches who after the first day's games when it will be apparent whose price is likely to rise and who's may fall (non-appearence/injury/red card).

Members would note the number of transfers in and out at the end of every day* for each of their assigned players. After a week or so that information gets posted here for analysis. I think that 14 days should be sufficient to witness several price rises, hopefully including a +0.2 or two.

I think that if 60 players could be followed we might have sufficient data with which to deduce the formulae.

* It is thought that when prices change they do so between 1 a.m. and 2 a.m. G.M.T. Ideally, data would be collected as closely prior to this time as possible.

Any thoughts, folks?

Ferg, you've already done a lot of work on this matter. Please feel free to criticise/augment my suggestions.

For stage 1 then Flyman, i assume that as close as possible to the GW1 deadline we note the total no. of teams (in the game) and the teams selected by (%) (for each player we are going to track) to work out how many people have bought the players.

This will give us the baseline for the subsequent calculations??

Then (stage 2), at approximately 1 a.m. (if possible) we note the transfers in / out of each player and do this for 14 days, obviously noting price changes during that time.

So for example....

Stage 1

500,000 teams entered. Lampard is selected by 25% meaing 125,000 teams have selected him.

Stage 2

If i recall correctly Ferg identified that price rises occur when 1% of the total players in the game at the start of GW 1 transfer a player in (this is net transfers).

So 1% of 500,000 = 5000. If 5000 teams (net) transfer Lampard in his price will rise by £0.1m ??? Also, am i right in assuming this figure (1% of 500,000 = 5000) will remain static throughout the season as it is based on the total no. of players at GW 1

Have i got that right???

Price drops occur when 10% of the teams who have Lampard transfer him out. So 10% of 125,000 = 12,500. If 12,500 teams transfer Lampard out his price will drop by £0.1m ??

Again, is that right ????

Assuming the above is correct, then it shouldnt be too difficult to prove.

The tricky formula is the price fluctuations for injured players, that will obviously have to be studied over a longer period of time.

If ive got any of the above wrong than can somebody (Ferg?) put me right, cheers!

That was the theory.


Yep, it stayed the same regardless of how many new managers registered.

But if they make any changes to the formula this season, which is plausible as they may be expecting more managers to play or maybe it won't be static this year, then Flyman is following the right method to work out the new formula or prove the old one.


I was less sure about this, it worked in most cases but I got the impression there was an extra factor I wasn't considering.


You don't need 60 samples or 14 days really. You'll know on day two whether 5000-6000 transfers are necessary, then the more data you have the closer you'll be able to get to the precise number.

OK, what price drops - or lack thereof - really confused you and what percentage of the total were these anomalies?

Maybe if we concentrate on a few of those, we can come up with theories about what made those cases different?

The problem with drops is that they're based on percentages rather than a set figure. So you can't just look for a common number/threshold of transfers, you have to look at each player individually then see if the same pattern applies to others.

Then you also have the problem of predicting who is going to be unpopular enough to generate data, you can't base it on them picking up an injury as that changes things again so you have to guess which player a lot of managers are going to drop because they're out of form or missing a week.

Unpopular players aren't volatile enough and the popular players have more protection (due to volume), so I was looking at players in around 4-5% of teams as 2000-2500 transfers out was realistic and if I remember rightly the 10% rule applied to them for the first 0.1 drop only.

After that a further 10% didn't cause a further drop so I presumed a new figure/factor came in to play and never got around to working it out.

reg wrote:

Yes, reg, you are quite right, we need to know the total no. of teams in the game and the teams selected by (%), but not just for the 60-odd players we wish to track .........
As I understand it, each player has his own, personal X-factor, based on the percentage of managers who have him at the off. In order to be able to predict any particular player's pending price change we will need to know his S.P. - his starting percentage, so we need to record as many S.P.s as possible, as close to the market closing for GW 1. (If we can muster 10 volunteers for this stage to record two teams' S.P.s each it shouldn't be too exhausting.)

reg wrote:
Yes, I think it's something like this. As the price change is related to the player's personal S.P. we need to know every player's S.P.

However, if you are right here it would mean that no player's price could possibly fall more than 1.0 in any season (0.1 per 10% x 10 = 1.0 for 100%)? Do we have any instances of a player's price falling by more than 1.0 in a season? Please post below if you can recall any example of this.
If there are any instances it might indicate a sliding scale, e.g. 0.1 fall after the first and second 10% of the S.P go, lengthening to 0.1 fall when the next two 7.5% go, etc., .....

reg wrote:
True, but thankfully collecting the data shouldn't be too labour intensive.

ferg wrote:

Interesting points, ferg. I wrote the last bit above before reading your post (working offline here), so am curious when you say that the 10% rule only applied to the 1st drop and it was 10%+ for a further drop as this would limit the actual possible price fall even further!

I don't think it's necessarily 10%+ for a second drop, more like 10% (or a different figure) of the new total of managers who have the player.

e.g. 125,000 have Lampard, 12,500 out, 0.1 drop
112,500 have Lampard, 11,250 out, 0.1 drop
101,250 have Lampard, 10,125 out, etc.

And as the new total is a lot harder to record accurately (your first prediction needs to be correct and then you need to take the reading after an update, which are random times) the margin of error increases.

So it was just that after the first 10% there was no definitive pattern in the brief sample I looked at.

Hi everyone,

I have had some experience through playing this game over the last season. I did some monitoring for price flectuation, and I am interested if you guys seriously wanna form a group in that regard.

So count me in

That makes sense and would mean that a player could reduce by more than £1m per season after all.

If we manage to record an accurate set of data to analyse then we should be able to prove this also.

Gentleman, set your alarms!! 12.55 a.m.!!

Did you find anything to back up the thoughts above BH?

PS - Welcome to the forum!

In terms of recording the player values, it would surely be not too taxing to set up a monitoring machine. I don't know what software you would use but I'm sure you could leave a computer mechanically trawling through the website recording all player values continuously. Then you analyse the results for changes.

If they update this - http://fantasy.premierleague.com/playerlist.html - at the same time as the real values, then that makes our job a whole lot easier!

Hello, Black_Horse, welcome to FISO and to this discussion. Thanks for offering to join us in our quest! Believe me, we are quite serious about this sort of thing and will make a good stab at solving the riddle.

If you are brand new to the FPL.com branch of FISO I'd recommend you take a look around and sign up to The Elite (our private league) and some of the subsiduary games we play alongside the main game - FISO 5 a Side, for example, or the FISO FPL Poll Team. Most of them demand no real effort from the managers but provide good banter and a lot of fun.

And just to re-iterate reg's question .... what conclusions did you come to last season?

Ferg: Thanks for clarifying. The idea that it is 10% of a varying (declining) number of managers makes a lot of sense, especially, as reg points out, because it permits an (almost) indefinite price drop.
However, it is possible that we may be faced with a varying scale of X% depending on starting price. I say this because:

Joe Bloggs - 3.5 - 800 players have him at the start.
If only 80 sold him in Wk1 his price would drop to 3.4. If just 137 people sell him next week his price falls again, but by 0.2 - to 3.2.
Now, correct me if I'm wrong, but in three seasons of FPL.com-ing I can't remember any player falling below 3.3 or 4.3, though I do remember some striker being 5.1 towards the end of the season before last - which actually backs up the idea if it is purely price based and not class based!

Perhaps we will find that for 3.5ers 50% must sell before triggering a price drop while 30% triggers a 5.5er's drop?
Just a thought .....

That could well be the case and maybe the 10% was just common to the players I was looking at. I definitely didn't follow any of the lowest price defenders.


I did see that as a stumbling block but maybe it could also be explained by the stat that his profile would say 'Teams selected by (%) 0', which may negate the whole formula.

Or maybe the formula just holds out because it was such a small number to begin with, such a large percentage of them wouldn't necessarily work in unison.

And have to remember that a single drop would make them the cheapest player in the game so a lot of people might be transferring them in just to free up as much cash as possible (that cycle occured a lot last season).