[OUTDATED] The Return of FGei (for v1.32x)

[Keeps]

Yea, more testers and tests will result in the performance of a deck in an average players hand,with less deviation, which is of some value.

[Keeps]

So I started testing and entering some data in the gdocs sheet since, you need more tests...
Something I noticed is FGei isn't doing much adjustment on skips, sometimes no change in FGei. 
Worse is if I enter in a loss that has a TTL greater than the average, and the FGei actually goes up, sometimes by a lot.  Did I just lose that much faster than the average.  What are the calculations you are using.  This is a simple equation.

(((1500 * (Wins * Elemental God Spin Chance %)) + (Electrum Won - (30 * Total Losses))) / (Total Time Played / 3600)  = Electrum/Hour for that god normalize from there.  Add all gods / total gods and there you go.

Anything significantly different, is likely wrong.
I should not be seeing FGei go up when entering a loss or skip, let alone by sometimes 300 points.

[Leodip]

So I started testing and entering some data in the gdocs sheet since, you need more tests...
Something I noticed is FGei isn't doing much adjustment on skips, sometimes no change in FGei. 
Worse is if I enter in a loss that has a TTL greater than the average, and the FGei actually goes up, sometimes by a lot.  Did I just lose that much faster than the average.  What are the calculations you are using.  This is a simple equation.

(((1500 * (Wins * Elemental God Spin Chance %)) + (Electrum Won - (30 * Total Losses))) / (Total Time Played / 3600)  = Electrum/Hour for that god normalize from there.  Add all gods / total gods and there you go.

Anything else, is likely wrong.
I should not be seeing FGei go up when entering a loss or skip, let alone by sometimes 300 points.

I'm not an expert when it comes to FGei, but the FGei goes up when you lose faster than usual because that game would steal more time from your other winning games.
Let's say you hve a deck with a 98% (yeah, lol) WR. If the 2% are insta losses (you see your hand and quit), your FGei would skyrocket. This is because when you lose one game, there are more games you can win than those you will lose.
If the WR was under X (where X is a number I don't know, I'd need to examine the FGei formula first, but it is a number that would get the gains from spins and wins to EQUAL the losses, this means that it is extremely low), the longer the losses, the higher the FGei.

Hopes it makes sense and I was clear (and right) enough.

[Keeps]

Thanks for the explanation; however, that logic is faulty.

Here is why:
The time spent on the next game is already calculated in by adding that time of that next game in.  Any additional calculation is a bad model leading to bad math.

Let's use your example:
A deck with a 98% win rate against Gemini is played 100 times against Gemini, 5 out of 10 games gives you a card, it takes 200 seconds to win each time, each win yields 60 electrum, and 65 when it wins a card, and 10 seconds to lose.

Lets say the first 98 games were wins and the last two where losses for this demonstration

This means :
(((1500 * (98 * 50%)) + ((60*49 + 65*49) - (30 * 0))) / (20000 / 3600) =  79625 /5.556 hours = 14331  /hr

now adding 2 losses

(((1500 * (98 * 50%)) + ((60*49 + 65*49) - (30 * 2))) / (20020 / 3600) =  79565 /5.561 hours = 14307  /hr

The result is a small drop in electrum / hr.
It only makes sense...  you added a small amount of time to your play, cost yourself a little money, you don't just magically get more electrum / hr.

[Leodip]

Hmm, this is pretty hard to explain, because it'll take lots of maths I don't really feel like doing.
Do you know why you should skip while FG farming? That's because however that game ends, it'll be your loss. This way it is better to skip in first place and hope you'll get a better FG you can win, that's why, if the WR is lower than a certain amount, it'll cause a decrease while if higher it will cause an increase.

[Pella]

(((1500 * (Wins * Elemental God Spin Chance %)) + (Electrum Won - (30 * Total Losses))) / (Total Time Played / 3600)  = Electrum/Hour for that god normalize from there.  Add all gods / total gods and there you go.

Anything else, is likely wrong.

If I'm reading this correctly, you're using 1,500 as an average card price.  The true average is much closer to 1,150.  In addition, your formula conveys the right concept yet omits important details.

These facts, combined with your follow-up statement, create an enormous drop in your credibility with me.  Regardless of how many correct things you say after this, I'll have to double- and triple-check any information you provide for a LONG time before I accept ANYTHING you say at face value.  The primary thought in my head will be, "He said what?  Yeah, well, he's the guy who jumped up and declared all my calculations invalid because they didn't match his obviously incorrect formula.  I'll take that with several grains of salt."

While Leodip's explanation is valid, there is a simpler explanation for the phenomenon you observed.  For decks with a minimum of data, you will see the word "NEW" on a bright red background immediately below the FGei number.  That should have been a clue.  In the second post of this thread (the one where you probably looked to find a deck to test), I explain:

(NEW) means fewer than 3 duels per FG have been reported.  This FGei should not be trusted at all.

The specific reason the FGei cannot be trusted at this stage is precisely because of the type of behaviour you observed.  With only a few data points, the calculation can change wildly with each entry.  With more data, the FGei finds its center, and this behaviour lessens until it disappears.

TL;DR  The large swings in FGei at this stage of data collection are normal and expected.  The calculation contains no major errors.

Spoiler for a small piece of (unsolicited) advice:

Remember that what you say is much less important than how you say it.  You could have simply asked questions and reported the behaviour you observed.  While you did those things, you also included statements in which I perceived an attitude approximating:  "It is impossible for a correct calculation to produce this behaviour.  Whatever you're doing is not only wrong, it's seriously broken.  The FGei formula is way too simple for you to have broken it this badly."

In all likelihood, you did not intend to communicate that attitude.  You communicated it regardless.  An old adage states, "You attract more flies with honey than with vinegar."  Usually, a softer, gentler approach produces better results than one that is accusatory or can be construed as such. [/advice]

[Keeps]

Pella

If I'm reading this correctly, you're using 1,500 as an average card price.  The true average is much closer to 1,150.

1500 or the cost to upgrade a card is what I am using, the actual value of that card won varies, are you going to sell it, if you upgraded the card itself you lose it's base value, another level of calculations will be needed to come up with this number, I used this as a stepping stone.

In addition, your formula conveys the right concept yet omits important details.

You are missing the point, the formula is a simple one, didn't include some concepts like skips.  The point of it was, to demonstrate that you don't get Electrum for losing so your efficiency does not go up for losing.

These facts, combined with your follow-up statement, create an enormous drop in your credibility with me. 

You stated one fact that is an "argument of fallacy" and missing the actual point.
Then, another another without stating it, which is well no evidence at all,
Following with the assumption the 2nd is automatically true, "argument of ignorance" probably.
Finishing with a sort of personal attack.
Now who's credibility is in question? 

The large swings in FGei at this stage of data collection are normal and expected.

Although understandable that with limited data the FGei should not be considered accurate.  I am pointing out a logic error, the model is by definition wrong.
Adding a loss, no matter how small amount of time to your play, cost yourself a little money and a little time.  It does not just result in magically getting more electrum / hr.

The calculation contains no major errors.

I just showed you a logical error.  Explain to me how if a person enters a loss, the FGei goes up?

[ColorlessGreen]

Fun experiment: Take the raw data from this spreadsheet, toss it in the generally-accepted-to-be-functioning-correctly StatMasta spreadsheet, compare the results, go from there.

[Pella]

I just showed you a logical error.  Explain to me how if a person enters a loss, the FGei goes up?

Leodip already explained this to you.  You rejected his explanation as illogical.  Whether the explanation makes logical sense to you or anyone else, the explanation is both valid and correct.  In my four college courses in statistics, I learned that numbers often do things that appear highly illogical when the number of data points is low.  A full explanation is beyond the scope of this discussion.

TL;DR You already rejected the correct explanation.  I won't give it to you again just so you can reject it again.

[CuCN]

It is possible for a quick loss to raise the FGei, because of normalization. If you enter a game, win or loss, that reduces the weight of every other game against that FG. However, this should only happen if there are no wins recorded against that FG.

[Keeps]

CuCN,
Thank you CuCN.  That explains my question, didn't realize the sheet was trying to normalize.

[Pella]

didn't realize the sheet was trying to normalize.

srsly?

The author of the original FGei thread talked about normalization at great length.  That's the whole purpose of the FGei in the first place.  If the GDocs sheet weren't trying to normalize, it wouldn't be calculating an FGei.

You deride people's correct answers by talking about logical fallacies, all the while inferring that you completely understand FGei, when in fact you have either missed or ignored its most essential element.  Whatever credibility with me you had remaining after your first post has completely washed away now.  Don't expect any change in that status in the foreseeable future, if ever.